ArticlePDF Available

Medical imaging dose optimisation from ground up: Expert opinion of an international summit

May 2018
Journal of Radiological Protection 38(3)

Authors:

GE Healthcare

As in any medical intervention, there is either a known or an anticipated benefit to the patient from undergoing a medical imaging procedure. This benefit is generally significant as demonstrated by the manner in which medical imaging has transformed clinical medicine. At the same time, when it comes to imaging that deploys ionizing radiation, there is a potential associated risk from radiation. Radiation risk has been recognized as a key liability in the practice of medical imaging, creating a motivation for radiation dose optimization. The level of radiation dose and risk in imaging is varied but is generally low. Thus, from epidemiological perspective, this makes the estimation of the precise level of associated risk highly uncertain. However, in spite of the low magnitude and high uncertainty of this risk, its possibility cannot be easily refuted. Therefore, given the moral obligation of healthcare providers, "first do no harm," there is an ethical obligation to mitigate this risk. How to precisely achieve this goal scientifically and practically within a coherent system has been lacking. To address this need, in 2016, the International Atomic Energy Agency (IAEA) organized a summit to clarify the role of Diagnostic Reference Levels to optimize imaging dose [1,2], summarized into an initial report [3]. Through a consensus building exercise, the summit further concluded that the imaging optimization goal goes beyond dose alone, and should include image quality as a means to include both the benefit and the safety of the exam. The present, second report details the deliberation of the summit on imaging optimization.

Overall patient risk including radiation risk and clinical risk as a function of dose. Dashed lines represent the optimum target. The units of both axes are arbitrary. Two examples of individual imaging procedures, each represented with three corresponding risk value datapoints, demonstrate different degrees of accuracy in meeting the optimisation target.

…

Overall patient risk including radiation risk and clinical risk as a function of dose illustrated for a cohort of patients. Dashed lines represent the optimum target. Each individual imaging procedure is represented with three corresponding risk values. The minimum of the total risk across the population is noted by an arrow, reflecting the accuracy by which the optimisation is achieved for this cohort. The units on the axes are arbitrary. The 25-75 percentile range of total risk values in the cohort encompassing the minimum population risk, demonstrated by the solid horizontal line, represents the precision by which the optimisation is achieved.

…

Figures - uploaded by Jenia Vassileva

Content may be subject to copyright.

Content uploaded by Jenia Vassileva

Content may be subject to copyright.

Journal of Radiological Protection

ACCEPTED MANUSCRIPT

Medical imaging dose optimization from ground up: expert opinion of an

international summit

To cite this article before publication: Ehsan Samei et al 2018 J. Radiol. Prot. in press https://doi.org/10.1088/1361-6498/aac575

Manuscript version: Accepted Manuscript

Accepted Manuscript is “the version of the article accepted for publication including all changes made as a result of the peer review process,

and which may also include the addition to the article by IOP Publishing of a header, an article ID, a cover sheet and/or an ‘Accepted

Manuscript’ watermark, but excluding any other editing, typesetting or other changes made by IOP Publishing and/or its licensors”

During the embargo period (the 12 month period from the publication of the Version of Record of this article), the Accepted Manuscript is fully

protected by copyright and cannot be reused or reposted elsewhere.

As the Version of Record of this article is going to be / has been published on a subscription basis, this Accepted Manuscript is available for reuse

under a CC BY-NC-ND 3.0 licence after the 12 month embargo period.

After the embargo period, everyone is permitted to use copy and redistribute this article for non-commercial purposes only, provided that they

adhere to all the terms of the licence https://creativecommons.org/licences/by-nc-nd/3.0

Although reasonable endeavours have been taken to obtain all necessary permissions from third parties to include their copyrighted content

within this article, their full citation and copyright line may not be present in this Accepted Manuscript version. Before using any content from this

article, please refer to the Version of Record on IOPscience once published for full citation and copyright details, as permissions will likely be

required. All third party content is fully copyright protected, unless specifically stated otherwise in the figure caption in the Version of Record.

View the article online for updates and enhancements.

This content was downloaded from IP address 161.5.6.201 on 29/05/2018 at 07:58

Medical Imaging Dose Optimization from Ground up: Expert Opinion of an

International Summit

Ehsan Samei

, Hannu Järvinen

, Mika Kortesniemi

, George Simantirakis

, Charles Goh

Anthony Wallace

, Eliseo Vano

, Adrian Bejan

, Madan Rehani

, Jenia Vassileva

Abstract

As in any medical intervention, there is either a known or an anticipated benefit to the patient

from undergoing a medical imaging procedure. This benefit is generally significant as

demonstrated by the manner in which medical imaging has transformed clinical medicine. At

the same time, when it comes to imaging that deploys ionizing radiation, there is a potential

associated risk from radiation. Radiation risk has been recognized as a key liability in the

practice of medical imaging, creating a motivation for radiation dose optimization. The level of

radiation dose and risk in imaging is varied but is generally low. Thus, from epidemiological

perspective, this makes the estimation of the precise level of associated risk highly uncertain.

However, in spite of the low magnitude and high uncertainty of this risk, its possibility cannot

be easily refuted. Therefore, given the moral obligation of healthcare providers, “first do no

harm,” there is an ethical obligation to mitigate this risk. How to precisely achieve this goal

Department of Radiology, Duke University, Durham, North Carolina, USA.

Radiation and Nuclear Safety

Authorty (STUK), Helsinki, Finland.

University of Helsinki, Helsinki, Finland.

Greek Atomic Energy Commission,

Paraskevi, Greece.

Singapore General Hospital, Singapore.

Australian Radiation Protection and Nuclear Safety

Agency, Sydney, Australia.

Complutense University, Madrid, Spain.

Department of Mechanical Engineering and

Material Science, Duke University, Durham, North Carolina, USA.

Massachusetts General Hospital, Harvard

University, Cambridge, Massachusetts, USA.

International Atomic Energy Agency, Vienna, Austria.

Page 1 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

scientifically and practically within a coherent system has been lacking. To address this need,

in 2016, the International Atomic Energy Agency (IAEA) organized a summit to clarify the role

of Diagnostic Reference Levels to optimize imaging dose [1,2], summarized into an initial

report [3]. Through a consensus building exercise, the summit further concluded that the

imaging optimization goal goes beyond dose alone, and should include image quality as a

means to include both the benefit and the safety of the exam. The present, second report

details the deliberation of the summit on imaging optimization.

Article Highlights

• Imaging dose optimization is an ethical, professional, and economic necessity to make

imaging safe, consistent, and accurate.

• Imaging dose optimization is only a part of a broader mandate for optimizing patient

care including all sources of risk.

• Imaging optimization involves an intentional balance between patient-informed

surrogates of radiation risk and clinical risk, taking into consideration the image task

and the attributes of the patient, such that the total risk to the patient (including

radiation risk and clinical risk associated with a sub-quality exam) is minimized.

• Imaging optimization requires and is aided by a pragmatic deployment plan, a

specialized workforce, an informatics infrastructure, and intentional regulations and

professional guidelines.

Page 2 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

I. Introduction

Medical imaging has become an indispensable healthcare resource because of its significant

advantages in accuracy, definitiveness, and versatility by which it offers crucial phenotypical

information about the clinical condition of the patient. Medical imaging is now considered

essential in healthcare for both adults and children across a wide spectrum of diseases [4]. The

annual number of imaging examinations currently exceeds 3.6 billion worldwide and is steadily

increasing [5]. Approximately 3-10% of imaging examinations are performed in the pediatric

age group with some children undergoing multiple examinations [6, 7,8].

The magnitude and the expanding use of imaging create a growing concern about radiation

exposure associated with medical imaging. In the U.S., the per capita effective dose has

increased from 3.6 milliSievert (mSv) in the early 1980's to 6.2 mSv in 2006 [9]. Presently CT is

the single largest source of medical radiation exposure, constituting half of the total medical

exposure and 25% of the annual exposure to the U.S. population [9]; in Europe, CT constitutes

on the average 60% of the total medical exposure and in some countries even more than 80%

[10]. Moreover, and despite an abundance of dose reduction technologies such as automatic

exposure control and iterative reconstruction, there is a wide variation in doses to patients for

similar examinations [11], with varied and sometimes excessive irradiation burden imparted.

Such burdens are of particular concern to vulnerable populations, such as pediatric patients

[12], individuals who are genetically more radiosensitive [13, 14, 15], and those who require

repeated imaging of the same anatomic area [16, 17, 18].

There are two types of effects from radiation: tissue reactions (also called deterministic effects,

e.g., radiation burns and skin necrosis from severe overexposure) and stochastic effects (e.g.,

Page 3 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

increased likelihood of inducing cancer in a given patient and potential mutagenic (or

hereditary) burden on the population at large) [19,20]. In routine imaging exams where dose is

not considered overly excessive, the main effect of relevance is stochastic risk. The risk to an

individual patient is likely to be small. However, the magnitude of the risk has been very

difficult to ascertain as the estimates have been derived primarily from non-medical data and

extrapolative projections with large uncertainties at low radiation levels.

While this uncertainty at low radiation levels has given rise to debates [21, 22, 23, 24, 25],

because of concerns about increased utilization of higher dose imaging exams and its impact

on population risk [26, 27], the prevailing efforts of the scientific and medical communities are

increasingly directed towards mitigation. Mitigating radiation risk, however, should ideally

take place at the individual level and be indication specific. That is because, regardless of the

magnitude, individual healthcare requires individual risk consideration. Further, aggregated

data across a population would be accurate only to the extent that the individual estimates

being pooled are accurate themselves.

Computed Tomography (CT) has been a primary radiological modality of concern due to its

large contribution to the cumulative radiation exposure of the population. A recent national

Summit in the US on Management of Radiation Dose in Computed Tomography [27] indicated

a strong consensus among experts and leaders for targeting low effective dose levels for CT

(and by implication for all imaging examinations), below the emblematic 1.0 mSv level. The

sub-mSv target for dose reduction strategies is worthwhile for several reasons: (a) it signifies

that current clinical practice allows doses that are often unnecessarily high; (b) it aims to lower

any potential risk from an imaging dose to levels that are closer to that of less contentious

Page 4 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

procedures such as radiography; (c) it provides a recognizable target threshold for reducing

imaging dose to ranges that are comparable to that of background radiation; (d) it encourages

a consensus in technical and operational advances to reduce and standardize CT doses; (e) it

promotes an understanding that the imaging community must be accountable for, and take

ownership of, any actual or perceived risk of imaging; and (f) it calls for broad and practicable

translation of current and evolving advances in imaging dose and image quality management

technologies to the clinical arena. It also presents an achievable vision for CT optimization

considering the foreseeable technical development.

However, in achieving the goal of reducing imaging dose, it is imperative not to ignore the

value of imaging, embracing both its benefit and its safety. Medical imaging is expected to

provide a diagnostic benefit; patients have imaging examinations based on the premise of

diagnostic benefit and as such dose should not be reduced to levels that may negatively

compromise this benefit. A dose that is too low to provide sufficient diagnostic information is

actually excessive, as it unnecessarily exposes a patient for the sake of questionable medical

benefit. Recent attention from the public and regulatory governmental sectors to medical

radiation and risk has resulted in focusing on dose reduction without due accountability of the

overall benefit of the intervention. For example, a recent symposium on dose monitoring [28]

afforded little discussion on the additional professional responsibility for optimizing diagnostic

quality. A focus on radiation risk alone without the balance of quality is insufficient to practice

evidence-based medicine or to implement the meaningful aspect of a “meaningful use”

moniker. While emphasized by a few recent efforts [29, 30, 31, 32], the benefit aspect of

Page 5 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

imaging has largely been on the periphery of public focus and scientific response, creating a

potentially ineffective perspective from which to achieve proper dose reduction.

In June 2016, the International Atomic Energy Agency (IAEA) organized a summit titled

“Patient Dose Monitoring and the Use of Diagnostic Reference Levels for the Optimization of

Protection in Medical Imaging,” with more than 60 international experts. The summit led to a

first report [3] in the use of Diagnostic Reference Levels [1] to benchmark operational dose

levels and to optimize imaging [2]. Through multi-day group discussions of the international

participants and consolidations of the opinions, the summit further reached a broad consensus

that considering the very reason for performing medical imaging, it is imperative that dose

reduction and optimization would not be approached in such a manner that the clinical benefit,

on an individual patient-by-patient basis, is compromised; the individual patients should be

exposed to the minimum amount of radiation necessary for an adequate and accurate clinical

diagnosis. In that way, the dose reduction and optimization goal can become a portal to a

more comprehensive directive that includes examination quality, a directive that goes beyond

the concept of dose to incorporate the broader concept of patient welfare [33, 34]. The

metrology and the process to achieve this comprehensive directive of individual care were

deliberated at the meeting, generating a second draft consensus document that was refined

into the present form. This paper was designed to both serve as a comprehensive summary

and a detailed strategy for imaging optimization.

II. Definition of optimization of patient dose for medical imaging

Page 6 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

Optimization can be framed as the objective and the process by which the risk of an imaging

procedure (in this case radiation risk) is balanced with respect to its benefit.

This process,

fundamentally, stems from the very purpose of the procedure being performed, i.e., its

justification. In medical imaging, a radiation exposure is performed for the explicit goal of

safely obtaining useful information relevant to a target indication of interest for accurate and

precise management of patient care [20, 35]. To achieve this objective, images should be more

reflective of the state of the patient and less of the particularities of the imaging technique

used. The choice of imaging system or technique applied in an examination should be directed

by the specific diagnostic information sought from the examination. This goal is very difficult

to achieve in practice due to challenges of measuring the clinical outcome or the added value

of the imaging procedure in an objective and personalized manner.

Clinical care activities are generally heterogeneous, compounded, and complex. They involve

varying technological offerings (e.g., makes and models of different systems for a given

modality), each with varying technological parameters (e.g., imaging factors – the

independent parameters that can be independently varied for the purpose of optimization).

The patients themselves come with natural variability, which in a perfect world would be

accommodated by the dynamic adaptation mechanism built into the modern imaging systems

or processes (e.g., higher mA for larger patients). However, this adaptation mechanism is not

perfect leading to variability in the final results [36]. Complicating this heterogeneous state is

the fact that medical care is a human process, subject to their own associated inconsistency,

Optimization is applied towards several objectives in medical imaging including optimization in the

design of imaging equipment. This report pertains to the optimization of the planned and actual

operational use of medical imaging systems [2, 35].

Page 7 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

competing and conflicting interests, and priorities, differently applicable to an array of

different stakeholders: vendors, insurance companies, employers, etc.

The heterogeneity of this landscape leads to sub-optimal and variable image quality and dose

such that the above stated goal of the procedure, “safely obtaining useful information relevant

to a target indication of interest for accurate and precise management of patient care,” is not

achieved. Thus, sub-optimal imaging carries the risk of not achieving the very purpose for

which it was performed. Optimization should provide an assurance that the goal of the

imaging procedure is achieved. This includes consideration of the risk associated with the

application of the ionizing radiation used in the process, the so-called radiation risk. But most

importantly, the likelihood of not delivering the very purpose of imaging, i.e. delivering the

desired benefit, should be recognized as a risk, which in this report, we refer to as clinical risk.

Comprehensive optimization combines these two risks, radiation and clinical, as a unified total

risk estimate (or index) within an indication-informed process. Other sources of risk, such as

the use of contrast medium, can be added to this framework, though that is beyond the scope

of the present paper.

Figure 1 offers a schematic illustration of this mental viewing of ‘optimization’. The figure also

unveils the physical meaning of optimization. Optimization means to make a choice (to opt)

between two designs (procedures, dosages, options), which are represented by the two

asymptotes shown in the figure. The optimal choice turns out to be the tradeoff between the

two, the optimum balance that in the figure is represented by the minimum of the upper curve.

Incidentally, design in nature and in the human realm is often informed by a balance between

two competing options; the design emerges from the competition between extremes. This

Page 8 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

phenomenon is ubiquitous, in both biological and non-biological systems, and is formulated in

the constructal law in physics [37]. As such, similar graphic constructions of optimal design in

Figure 1 appear in the constructal design of speed and frequency in animal locomotion [38],

turbulent eddy size, snowflake design, cooling channels for electronics, Bénard convection, jet

engine size for aircraft, and others [39, 40, 41].

Figure 1. Overall patient risk including radiation risk and clinical risk as a function of dose. Dashed

lines represent the optimum target. The units of axes are arbitrary. Two examples of individual

imaging procedures, each represented with three corresponding risk values-datapoints,

demonstrate different degrees of accuracy in meeting the optimization target.

In Figure 1, increasing dose (in terms of a relevant metric being optimized) in a given imaging

examination increases the radiation risk to the patient. In this illustration, the radiation risk is

assumed to follow the linear-no-threshold model, but any alternative model can also be

considered here. The increase in dose has a corresponding associated influence on clinical risk:

As the dose increases, this resultant image quality increases which in turn improves the

information content available to the clinician reducing the likelihood of sub-optimal diagnosis.

0.0

0.2

0.4

0.6

0.8

1.0

Rela ve

Risk

Dose

Radia on

risk

index

Clinical

risk

index

Total

risk

index

Page 9 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

The shape of this dependency follows an inversion of the asymptotic relationship universally

observed between radiation dose and image quality [42]. The two risk models follow reversing

trends such that the total risk to the patient exhibits a minimum “valley” of lowest risk. The

valley essentially provides a target for the objective of optimization of overall risk. It should be

noted that in the vertical axis, the 1.0 value of relative clinical risk at a zero radiation dose level

does not imply fatal clinical endpoint (of death) but rather an indication dependent measure of

disease or diagnosis related detriment. This may be further normalized to scale with the

radiation risk endpoint (of cancer death) but such normalization requires inclusion of other

clinical factors and possibly estimates related to an extended set of indications and

pathological prevalence.

The above characterization of optimization is consistent with the ALARA principle, which has

been used to represent the principle of optimization efforts. However, it provides a more

granular and targeted definition. By framing the optimization as a balancing act of two similar

quantities, one risk against another, the process involves a comparison of “apples and apples.”

This is a much more comprehensible task than balancing risk and benefit, factors that follow

different scales and correlations that rarely, if ever, demonstrate an optimum “peak or valley”

that optimization processes generally seek. In this framing of optimization, optimization

further becomes a process by which a patient would be positioned to experience the lowest

combined radiation risk and clinical risk and thus highest level of overall welfare, which is the

very purpose of performing a medical procedure.

It should be noted that the concept illustrated in Figure I is broadly applicable to most imaging

modalities utilizing ionizing radiation, except those that use a detector technology with a

Page 10 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

limited, narrow dynamic range, i.e., screen-film. In film radiography, the film speed and the

narrow range of recordable exposures limit the choices for optimization for a given screen-film

system. While the ability to more flexibly optimize imaging is a possibility for digital imaging

technologies, this possibility has not been fully taken advantage of in most clinical applications

today.

Necessary ingredients to realize the above framework are metrics associated with the

radiation risk and the clinical risk. In optimization discourses, the radiation risk has been

extensively recognized and deliberated upon, demonstrated by the fact that the ALARA is

often enacted in terms of “dose optimization” [43, 20, 35]. Clinical risk, in comparison, has

been less recognized. Clinical risk can be defined as the risk associated with lowered diagnostic

confidence and the associated reduced likelihood of accurate interpretation leading to

misdiagnosis [44]. It can be related to either the detection or quantification of the pathology of

interest or affirmation of its absence, which are directly dependent on the indication of

concern. In either case, risk is defined based on the indication and the patient. In dealing with

both radiation risk and clinical risk, optimization is characterized in a patient-centered manner.

While clinical risk is primarily an indication-centric metric, as a secondary consideration, the

clinical risk can also be recognized as the reduced confidence in identifying incidental

pathologies, not related to the targeted indication. This clinical risk is further dependent on the

prevalence and significance of such incidentals, if present. This latter condition is obviously

hard to target and to optimize, as it is not based on a priori known condition. Nevertheless, it

should not be overlooked in the broader context of optimization. Furthermore, the

performance aptitude of the practitioner and differences in different practice settings can

Page 11 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

impact the magnitude of clinical risk – another factor that should be taken into consideration

within the assessment of clinical risk.

Figure 2. Overall patient risk including radiation risk and clinical risk as a function of dose

illustrated for a cohort of patients. Dashed lines represent the optimum target. Each individual

imaging procedure is represented with three corresponding risk values. The minimum of the total

risk across the population is noted by an arrow, reflecting the accuracy by which the optimization

is achieved for this cohort. The units on the axes are arbitrary. The 25-75 percentile range of total

risk values in the cohort encompassing the minimum population risk, demonstrated by the solid

horizontal line, represents the precision by which the optimization is achieved.

Optimization through characterization and minimization of total risk can be applied towards

an individual or a population. In doing so, the concepts of accuracy and precision come into

play. The accuracy (or bias) of optimization pertains to the ability to perform an imaging

procedure at the lowest point of the risk-dose continuum. In Figure 1, the two example cases,

represented by datapoints represented at two different dose levels, exhibit different degrees

of accuracy in terms of their proximity to the optimum, minimum risk target. Each patient is

represented by three associated data points for radiation, clinical, and total risk, respectively.

0.0

0.2

0.4

0.6

0.8

1.0

Rela ve

Risk

Dose

Radia on

risk

index

Clinical

risk

index

Total

risk

index

Page 12 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

In a population of cases, the bias is the difference between the optimum target and the median

of the population distribution. The precision is likewise the deviation of the total risk values

across the population. Figure 2 demonstrates the application of the accuracy and precision of

optimization across a population.

As noted above, optimization can be applied to an individual, or to a population. In either case,

the patient is recognized within a cohort sharing common attributes and imaging conditions.

Those include imaging system specifications, the indication, protocol, patient size, etc. The

risk data are thus stratified according to a combination of the desired conditions. When the

conditions are too narrowly-defined, however, the number of data points representing a

population can be small, affecting the statistics of population-based optimization analyses.

This is a common consideration in pediatric imaging.

In pediatric imaging, the number of cases within a narrow weight, age, or time window might

be small (e.g., abdominal CT cases of 5-year-old patients imaged in the evening shift within

one month on one CT system). Avoiding such data-starved conditions necessitates broader

stratification, pooling data over a wider window of conditions (e.g., abdominal CT cases of 5-

year-old patients imaged in the evening shift within one month across all CT systems). This in

turn reduces the precision of the optimization claim for the population. Thus, the claim of the

optimization is dependent on the inherent variability of the varying conditions represented in

the population, as well as the number of cases represented in the sample.

To achieve optimization, both radiation risk and clinical risk require proper surrogates, framed

as metrics or indices. These metrics are outlined in the following section.

Page 13 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

III. Definition of quantities necessary to achieve optimization

The process of optimization in imaging involves dependent and independent quantities. There

are different layers of dependencies: the first layer is composed of the basic imaging factors

that can be independently varied along either a continuous or a sparse parameter space, such

as mA and kV. Varying these parameters affects dosimetric quantities such as CTDI, DLP [45],

which form the second layer of dependencies. In the example of CT, in the first layer, CTDI is a

dependent parameter. When CTDI is changed, however, the radiation risk and clinical risk are

correspondingly changed. Thus, in this second layer, CTDI effectually acts as an independent

parameter for the dependent parameters of radiation risk and clinical risk.

Imaging optimization is carried out in terms of one or a combination of multiple independent

or dependent parameters [46]. Some of these parameters involve the geometry (e.g., field of

view), timing (e.g., rotation time), or processing of the acquisition (e.g., slice thickness), while

others (i.e., dosimetric quantities of CTDI and DAP) relate to the radiation output of the

imaging system. The latter is often referred to as “dose,” as they tend to be related to patient

dose. Properly speaking, these metrics are not equivalent to patient dose, but provided clear

definition and precise quantification, they can serve as pseudo-independent quantities in the

optimization framework, where they are related to the dependent quantities of radiation risk

and clinical risk, as shown in Figure 1.

III.A. Radiation risk metrics

Page 14 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

Radiation risk requires quantification with metrics related to the risk of the individual patient to

that exposure. There are no ideal metrics available but many more or less reasonable

surrogates. In fact, all radiation “dose” metrics (dosimetric quantities) have often been

unintentionally treated as such surrogates but they fall into a relevance hierarchy in terms of

how well they can be related to the risk of the individual patient. Any surrogate of dose can be

considered as an ”index”, since it does not measure the real patient dose and ensuing risk. Here

we refer to them collectively as radiation risk indices (RIX). They may include those listed in

Table 1.

Table 1. Select radiation risk indices.

Index

Definition

Representing

Reference

DAP

Dose area

product

Machine output in radiography and fluoroscopy

in terms of the

product of radiation output and area of the exposure

[47, 48

49]

Fluoro time

Machine output in fluoroscopy

in terms of the time used to

perform the exam

[47, 48]

Entrance

Exposure

Machine output

in radiography, fluoroscopy, and mammography

in terms of entrance exposure (including skin backscatter)

[47, 48

]

Air KERMA

Machine output

in radiography, fluoroscopy, and mammography

in terms of KERMA in air at a specified distance

[47, 48

]

Administered

activity

Radionuclide activity

used

for a

specific

radiopharmaceutical

nuclear medicine procedure

[50

]

CTDI

CT dose index

Machine output in CT

in terms of dose in a specified phantom

placed at the iso-center

[45

]

DLP

Dose length

product

Machine output

in CT in terms of the product of CTDI

and

the

exposed length

[45

]

SSDE

Size

specific

dose estimate

Machine output in CT in terms of the product of CTDI and a

patient-size adjustment factor

[51]

DLP

size

SSDE x L

Machine output in CT in terms of the product of SSDE and

the

exposed length in CT

[52]

Organ dose(s)

Estimated organ dose values across the patient organs

for the

exact applied imaging condition representing the multiplicity and

granularity associated with the patient’s radiation burden

[53]

Defining

organ dose

Dose to a sensitive organ that is high enough to be used as a

primary indicator of radiation burden

[20]

based

effective dose

Effective Dose calculated based on estimated actual o

rgan doses

of the patient for the exact applied imaging condition

incorporating weighted organ sensitivities

[55]

Page 15 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

based

effective dose

Effective Dose calculated based on estimated organ doses of the

patient for a standardized imaging condition (eg, typical chest

CT)

[103

]

Output

based

effective dose

Effective Dose pre

calculated for a reference phantom based on

the modality output metrics (eg, DLP, DAP, or AA)

[45, 50,

52, 54]

RRI

Age

based

risk index

Radiation Risk Index calculated based on estimated actual organ

doses of the patient for the exact applied imaging condition

incorporating weighted organ and age sensitivities

[55]

RRI

Age and

gender-based

risk index

Radiation Risk Index calculated based on estimated actual organ

doses of the patient for the exact applied imaging condition

incorporating weighted organ, age, and gender sensitivities

[55]

NRRI

ormalized

RRI

normalized to RRI

of a reference patient (eg, 20 year old,

male, ICRP reference patient) undergoing the same exam

A few main characteristics of the above radiation risk indices are presented in Figure 3. The

metrics that take into account organ sensitivity are based on anatomical/geometrical

information on volumes of each organ exposed. The term “physical” in this figure aims to

signify the attribute of a quantity as corresponding to a physical reality, idealistically

measurable or estimatable. In contrast, the quantities that do not carry that label are purely

derived and cannot be measured even under the best of circumstances. Effective dose is such

a quantity. So are Radiation Risk Indices, with the term “index” incorporated in the definition

to highlight that attribute. But even so, Radiation Risk Indices are closest to the patient

radiation burden, incorporating granular information related to patient radio-sensitivity. Their

calculation will require knowledge of the specified attributes of the patient and the imaging

procedure [42]. We note that many of the quantities noted here are not currently readily

available in the clinical settings, motivating developmental and commercial offerings.

Page 16 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

Figure 3. Attributes of varying metrics of dose and radiation risk. The circles represent the

attributes that are largely reflected in a given metric in column 1. The half circle represents partial

reflection of the attribute.

III.B. Clinical risk metrics

The clinical risk is related to the achieved benefit of imaging reflected in the image quality.

Clinical risk or image quality should likewise be characterized with its own set of metrics

reasonably correlated with clinical benefit as clinical risk is inversely proportional to benefit or

image quality. The benefit, translated in terms of its reverse can be made more comparable to

radiation risk. The benefit should consider the indication or pathology of interest, e.g. the

image quality for confident detection of a renal stone will likely be different from that for a

small, non-obstructive embolus in the pulmonary vasculature. In its simplest form, the clinical

risk can be associated with the likelihood of misdiagnosis. In the future, it may extend the

clinical risk beyond misdiagnosis to mortality and to the broader context of combinatorial risk

from multiplicity of sources.

If characterized in terms of misdiagnosis (and associated diagnostic confidence and accuracy),

most image quality metrics can be considered as surrogates. The image quality metrics can be

Metric Physical

Patient Attributes

Scalar

Patient

Size

Patient

anatomy

Patient

age

Patients

Gender

CTDI, DAP, EE,

Activity

SSDE

Organ dose

Effective Dose

RRI

Page 17 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

either phantom based or patient based. Examples of such clinical risk indices (CIX) are listed in

Table 2.

Table 2. Select clinical risk indices.

Index Definition Representing Reference

Preference Acceptance of the radiologist or imaging physician of the applied

image quality; subjective grading of image quality

[56]

CNR Contrast to

noise ratio

The ratio of the contrast of a defined feature of interest to its

background noise

[57]

Az Area under the

ROC curve

Observer detection accuracy for a feature of interest within a

collection of images using ROC-type metrologies

[58]

Observer

recognition

Observer recognition accuracy of anatomy and/or pathology of

interest under investigation

[59]

d’

phantom

Phantom-based

detectability

indices

Detectability indices calculated from phantom images

estimating the likelihood of detection of a feature of interest

based on signal detection theory

[60]

d’

clinical

Clinical-based

detectability

indices

Detectability indices calculated from patient images estimating

the likelihood of detection of a feature of interest based on signal

detection theory

[61]

e’

precision

Classification

precision indices

Classification indices estimating the precision of classification of

a feature of interest based on signal detection theory

[62]

e’

accuracy

Classification

accuracy indices

Classification indices estimating the accuracy of classification of

a feature of interest based on signal detection theory

CRI

Clinical Risk

Index

Overall clinical risk index incorporating the combination of

factors that reduce the likelihood of proper diagnosis for a given

indication (e.g., Az

-1

)

A few main characteristics of the above clinical risk indices are summarized in Figure 4. The last

metric noted above is expected to be the best surrogate for clinical risk but its application

needs specific evaluation from the knowledge of the probability of misdiagnosis vs image

quality. The clinical risk indices can be applied to any imaging modality.

Page 18 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

Figure 4. Attributes of varying metrics of image quality and associated clinical risk. The circles

represent the attributes that are largely reflected in a given metric in column 1. The half circle

represents partial reflection of the attribute.

III.C. The balancing act

Aligned with the growing international call for personalized imaging [63, 64], the above-noted

metrics of radiation risk and clinical risk and their proper balance should be approached in the

context of an individual patient. This is due to the fact that the benefit and risk of an

examination are highly dependent on the specifics of the patient (e.g., anatomy, age,

radiosensitivity, indication): A dose that is proper for one patient may be grossly elevated or

insufficient for another. Thus, in the midst of great variability in clinical practice, appropriate

adjustments in the examination to achieve the balance between quality and dose are most

meaningful when done individually, as with any other medical procedure. The experience of

the clinician interpreting the images also plays a role here, creating another factor to be

eventually considered in this framework.

Page 19 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

The personalization of optimization is in concert with the goal of P4 medicine [64], which aims

for effective clinical outcome in a healthcare process that is personalized, predictive,

preventive, and participatory. The personalized aspect is given in the previous paragraph. The

predictive and preventive approach is closely linked to the personalized aspect since effective

personalized diagnostics and treatment will also improve the odds for proactive measures. And

conversely – the capability to utilize previously acquired data to make projections will help to

trigger preventive curative actions at the individualized level.

The optimization strategy should accommodate two notable features of radiation risk and

clinical risk: the temporal differences and stochastic quality: First, there is a temporal

difference in the two risk concepts: the radiation risk is a long-term concept because the harm

to the patient, a radiation induced cancer, can be manifested (with the given low probability)

over the course of years or tens of years. The risk of misdiagnosis vice versa can be immediate

and result in serious failure of the expected benefit in the on-going medical examination and

treatment of patient’s disease.

Second, risk is fundamentally a stochastic construct and, as a consequence, its understanding

and communication is subject to confusion. Risk magnitudes and estimates, radiation related

or otherwise, are constructed from population statistics. A 1 in 1000 risk is drawn from an

adverse effect to a certain individual from a population of 1000 individuals who share certain

common attributes. When communicating risk to a patient, we ascribe that population-base

likelihood of harm to an individual, not a population. In doing so, the risk is not an actual

“individual” risk, but rather the likelihood of harm to a theoretical population that shares the

same attributes as that of the patient, and that one theoretical individual in that hypothetical

Page 20 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

population (not necessarily the patient) would be harmed. Without understanding this

statistical nature of risk estimates, people tend to gravitate towards a deterministic

interpretation of risk values and ascribe the values deterministically to the potential harm that

their person will “certainly” receive if they undergo the procedure.

To be able to compare radiation risk and clinical risk, to attain the overall lowest patient risk as

demonstrated in Figures 1 and 2, the risks have to be put on the same scale. For example, there

is a need to know which value of SSDE corresponds to a given actual patient dose value (for a

given patient), and which value (how low a value) of CNR or detectability index corresponds to

a given probability of misdiagnosis (for a given patient and indication). Putting the two into the

same scale of mortality, outcome, and days lost enables the comparison of the two risk indices.

Once the latter relationship is known, there is still the need to know the relationship of this

clinical risk index to the corresponding radiation risk index, i.e. the image quality vs patient

dose. This altogether will enable considering the two risks in the same scale thus achieving the

overall lowest patient risk.

For the radiation risk, knowledge on the risk magnitude as a function of patient dose (e.g., in

terms of cancers observed in 1000 patients imaged) can be derived from the risk data

published based on organ doses. A similar relationship for clinical risk is needed, i.e. knowledge

of clinical risk magnitude (e.g. fraction of images leading to misdiagnosis) as a function of

patient dose, obtained through the image quality analysis; this will perhaps be the most

challenging task of the present modern approach for optimization strategy since it is linked to

various sets of clinically relevant data outside of the radiology context.

Page 21 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

The process of optimization, based on the quantities introduced in this section, must then

establish patient dose and image quality profiles for the imaging equipment using a stratified

approach: indication based/ patient size based/ patient age based. These profiles would then

serve as input data for a comprehensive optimization analysis, which also includes knowledge

on radiation risk versus dose, and clinical risk versus image quality. This process integrates the

multiplicity of information and characterizes optimized dose ranges based on desired risk

preferences, i.e., required accuracy and precision in matching the optimum target (Figure 2).

This challenge most probably cannot be completely solved without capability to combine

different silos of big data with deep learning methods [65]. One of the strengths of the deep

learning methods is the flexibility of application to many different tasks related to medical

imaging and optimization. The tasks specifically related to optimization could, for example,

involve tissue classification, structural detection, organ segmentation or image registration.

Data learning analytics seeks optimized endpoint prediction from feature groups to individual

to population. Effective use of deep learning, however, requires meeting multiple challenges:

First, the data must be accessible, taking into account the dispersed technical systems (closed

proprietary databases in hospitals), privacy regulations, and ethical and legislative

considerations. Second, the data need to have sufficient quality in terms of variability and

bias, necessitating rigorous data quality control and validation – poor data leads to poor

prediction. Specifically, the training data requires annotated image data, the ground truth,

provided in a systematic fashion [65]. These main challenges are likely to be solved in the near

future as the technical IT environment develops and the need for large scale data is more

recognized at various levels including legislation. In conjunction with the deep learning

Page 22 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

methods applied in radiomics, the phenotype data provided by medical imaging has a high

potential to provide biomarkers relevant to clinical outcome and clinical risk [66]. However,

upon combining data from various sources, the heterogeneity between clinical and physical

data may need to be mitigated by combining data based on clinical relevance and probabilistic

connections. Third, the data are often smaller than are ideally needed. This can be mitigated

with the use of models and simulated data sets, and with the use of feature extraction

methods to reduce redundant dimensions. Finally, the results will require validation, which

necessitates additional analyses, classification, interpretation, and probabilistic and predictive

modeling of its own.

IV. Practical process and tools for achieving and managing optimization

The success of the optimization process requires practical and capable tools: appropriate

metrics and their integration, stratification of data based on various parameters, and analytics

tools (software) to manage and analyze the data to extract optimized dose ranges. Figure 5

offers a schematic illustration of the components and the process. The figure illustrates how

image quality and dose metrology is informed by clinical data, with their dependencies

ascertained by analytical, empirical, or machine learning methods, to derive the maximization

of the patient benefit.

All metrics used should be recognized in terms of their relevance to the radiation and clinical

risk, as noted in the hierarchy of indices discussed above. The metrics should be clearly

defined and the method of their derivation and parameterization should be clearly stated in

the local QA documentation. For example, if a phantom based index (a surrogate) is used, the

Page 23 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

type of phantom - simple vs anthropomorphic - should be specified. This is the more important

because there can be differences in the definitions of quantities between manufacturers and

between countries and institutions. Unless the metrics are clearly defined and documented,

wrong interpretations can be made in establishing the patient dose and image quality profiles,

and this may eventually lead to non-comparability of the results of optimization.

The initial characterization of the patient dose and image quality profiles needed for

optimization should be achieved during acceptance testing by the medical physicist as part of

an image system commissioning [67]. A tiered selection of the indication based protocols for

optimization is recommended: most common and high dose protocols first, lower dose or less

frequently performed protocols later. A stratified approach is practical: generic profiles for

each indication based imaging protocol could at first be established, then (as the next layers) it

should be considered whether there is a need to distinguish between different groups of

patient size and/or age [42].

Page 24 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

Figure 5. Comprehensive risk modelling - data flow in clinical parameters, image quality, and radiation

dose characterization for imaging optimization. Connection points (circles) illustrates dependencies

that are ascertained by analytical, empirical, or artificial intelligence (including deep learning) methods.

Patient specific net risk

or benefit

Radiation risk

Anamnesis

Clinical status

findings

Prognosis

Lab data

Genetic data

Medication

Histology data

(Previous)

Imaging

findings

Clinical risk

Organ specific

cancer

models

Organ models

3D dose

distribution

Organ doses

Exposure

source model

Irradiation

event model

Patient model

MTF or TTF

(spatial

resolution)

NPS (noise

power

spectrum)

Clinical task

model

Detectability d’

estimability e’

Observer

noise

Eye function

Clinical data Image quality Radiation dose

Page 25 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

For a given protocol, the optimization could be a stepwise process leading to iterative

improvement: Initiating with vendor independent generic optimization parameters, one can

start with more impacting protocol parameters (e.g., mAs, kV, etc.) and continue with other

parameters (e.g. filtration) to reach an optimal normalized figure of merit, based on indication

specific diagnostic task models – including the range of tasks to provide a representation of

relevant diagnostic task variability [68]. That can then continue with the post-processing for

maximum task performance. And finally, the patient specific automatic selection of scan

parameters can be visited so that the diagnostic accuracy will be consistently reached for each

examination, regardless of varying patient morphology.

If the current clinical preferences are far from optimized conditions, gradual changes should be

the safest choice for a manageable and adaptable process of continual improvement. A new

version of the protocol should result from each step in the process. Once a diagnostically

acceptable protocol version is achieved, there should be a regular multidisciplinary stakeholder

review for improved optimization. In particular, the process of characterizing dose and image

quality profiles needs be repeated whenever there are changes or maintenance procedures or

updates in technical features or post-processing which can impact on either the patient dose or

image quality values, or both. Optimized imaging protocols should also be accompanied by

coherent imaging guidelines (for scanner users, and preferably also for referring clinicians and

patients).

For the optimization of each indication-based protocol, image quality profiles require certain

criteria to judge the likelihood of misdiagnosis (the selected descriptor of clinical risk). In the

multiplicity of possible tasks and variable conspicuity, there is a need to define a reference

Page 26 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

task. For example a reference task in liver imaging might be to be able to detect a 10 mm

lesion with 50 HU contrast. Further there is a need to define a limiting task that would be most

limiting in determining the optimum dose setting, e.g., 5 mm lesion in the above example.

For preference and performance based clinical risk indices, there is a need to develop a

reference set for images, i.e., reference images with acceptable image quality providing a

proper range of diagnostic tasks to represent relevant range of variability in diagnostic

definition of the target.

There should be a premeditated strategy to collect the clinical data (the cohort of patients)

needed to establish dose and image quality profiles [69]. Retrospective data acquisition from

image archives and dose management systems and analytics, based on sampled clinical data,

should be complemented by prospective clinical data; this will enable a stepwise process

towards protocol improvement and eventually purely prospective protocol optimization. In

this area we are likely to see significant development from machine learning methods which

may bring a paradigm change not only to optimization but to entire medical imaging.

Once the dose and image quality profiles are available for an indication-based protocol, these

are integrated in an optimization process. The accuracy and precision of the optimization

achieved for a cohort of patients can then be evaluated as shown in Fig. 2. Based on the

desired risk preferences, i.e. the accuracy (the median of the distribution versus target) and

precision for matching the target (e.g. interquartile range), the output would characterize the

recommended dose ranges for optimized protocol.

V. Needed informatics infrastructure for optimization

Page 27 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

The imaging protocol optimization management should be integrated with other

management systems: protocol management system, patient exposure data management

system, general quality management system (quality control and performance management)

and patient information/medical information management systems – to form a comprehensive

medical data and process management system. This would ensure that the protocol

optimization is based on up-to-date information on protocol versions, patient exposure

statistics, machine performance data and patient demography data. For example, when a new

imaging equipment is taken into clinical use, the informatics system can make the equipment

acceptance test data pertaining dose output and image quality readily available. This makes

the data readily available and comparable with corresponding data from the earlier or others

equipment at the facility, facilitating necessary design and modification of the protocols for

optimum use of the new equipment in a timely manner. With this integrated informatics

infrastructure, any feedback obtained from the protocol optimization, patient exposure and

quality management systems can be reciprocally investigated and appropriate corrective

actions implemented for continuous follow-up of the protocol optimization.

As stressed in sections II and IV, well-specified dose and image quality metrics as well as

standardized nomenclature for indication-based protocols are needed throughout the

optimization process including reporting. Radiological examinations and procedures should be

classified using harmonized nomenclature so that patient cohort based data can be uniquely

defined, analyzed and compared with other data [10, 70]. The nomenclature inconsistencies

can be magnified when comparing results of optimization across institutions, both nationally

and internationally.

Page 28 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

The dissemination of the DICOM standard started the significant and rapid development

towards standardized automated patient exposure data management, providing essential

tools for the data collection and management in the optimization process [71]. DICOM data

enabled the use of bitmap images to store dosimetric data as graphic images in the PACS and

attach them to a study report. DICOM headers could be tagged for multiple anatomical series

in one data set, but unfortunately, DICOM tags have been inconsistently used by different

vendors. Since the early development through the standard Modality Performed Procedure

Step (MPPS), the most advanced and comprehensive DICOM standard collation of data so far

is the Radiation Dose Structured Reports (RDSRs), which allow access to procedure data in a

structured and hierarchical digital format.

VI. Needed expertise for optimization

Training and specialization of all stakeholders (radiological medical practitioners, medical

physicists, medical radiation technologists and other health professionals) in medical

exposures is of major importance for the protection and safety of patients [2]. The need for

continuing education and training, especially in the case of the clinical use of new techniques

and technologies is also pointed out [72]. For example, in the case of CT examinations, the

importance of the users’ familiarization with technical aspects of modern systems that affect

patient doses has been discussed in the respective literature [73]. Guidance on optimization of

medical exposures has been provided by professional bodies [74, 75] and international

organizations [76, 77, 78, 79]. IAEA has also taken action by launching an e-learning platform

on radiation dose management on CT [80]. As machine and deep learning methods are taken

Page 29 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

into more general use, an interdisciplinary training infrastructure and curriculum should be

established for various professional groups to keep up to date with the wide technological

breach brought by these methods [81].

In order to realize the optimization strategy described in sections III and IV, the involvement of

all the stakeholders is required and probably in a sense that goes beyond their usual every-day

work and responsibilities. This fact implies that special training on optimization is required.

Respective training material should contain the rationale of the optimization process and the

methodology that would be followed and describe in detail the engagement of each

stakeholder.

The implementation of such an optimization process would require the setup of an

optimization working group that would include consultant radiologists, radiographers and at

least one medical physics expert. During the process, any proposed change of the examination

protocols or radiographic practice would require the image quality and diagnostic efficacy

evaluation of the outcome by the radiologists and the evaluation of the new dose to the

patient by the medical physics expert in order to evaluate total risk.

It is worth mentioning that radiologist’s preferences on the image appearance and reporting

skills are of great importance. For a given examination this process may have to be repeated

several times with gradual changes in order to achieve the optimum examination protocols

and radiographic practice that would minimize the total risk to the patient. The outcome of the

procedure would then have to be communicated to all the involved personnel of the

institution.

Page 30 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

Along with the members of the optimization working group mentioned above, other people

may also have to contribute to the process. IT department support is essential for, among

others, the reporting of imaging quality metrics and of dose metrics as mentioned earlier and

perhaps with the development of total risk estimators’ algorithms. The manufacturer’s support

of the radiological equipment can also make useful contributions. The implementation of the

optimization process would of course require the consent of the administration of the

institution and the allocation of the necessary resources. But most of all, the success of the

effort depends on the involvement of enthusiastic people, implemented using a team

approach and strategy, who believe optimization matters.

VII. Role of guidelines, regulation and professional organizations

To achieve the strategy outlined in this report, the efforts should be driven by international

organizations such as the IAEA in order to minimize potential commercial or institutional bias.

The work can then be pioneered in a small number of centers to demonstrate its utility and

value at which point a strong engagement with the industry may provide efficient resources for

broad implementation across institutions, facility settings, and countries. Encouraging

adoption of these best practices by imaging professionals and referring clinicians will also

require close partnership with international, regional and national professional organizations.

Finally, regulatory bodies, through the enactment of legislation, have a powerful tool to

enforce compliance with established dose optimization strategies.

Development and elaboration of dose optimization tools and strategies by international

organizations better represent the needs and views of experts from a variety of professional

Page 31 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

and cultural backgrounds, ensuring that the tools are relevant and generalizable over a wider

range of practice settings. As such organizations are not for profit and not associated with

particular institutions or companies, they are more able to act unfettered by such potential

influences and remain truly patient centered. Furthermore, through their existing relationships

with member countries, they are able to widely disseminate information and influence policies.

This approach has proved effective with the concept of diagnostic reference levels, which was

initially introduced by the ICRP in 1990 and is now widely endorsed as a valuable tool for dose

optimization [82, 49] and with the International Basic Safety Standards published by the IAEA

[2], which form the basis for national legislation in many of its member states. The Bonn Call

for Action has proved instrumental in galvanizing stakeholders from various nations into

launching their own campaigns to promote radiation protection in medicine, such as Afrosafe,

Eurosafe Imaging and the ISSRT Action Plan to Bonn Call for Action.

Engagement with healthcare professionals via their professional associations is crucial,

because even when proven, validated dose reduction strategies exist, adoption often lags

behind. In the familiar example of low dose CT for renal colic, a 2014 report from the American

College of Radiology National Radiology Data Registry [83] found marked variation in the

mean radiation exposure, with only 2% of institutions using the recommended sub 3 mSv

effective dose. This gap between theory and practice may be due to lack of knowledge or

unwillingness to implement changes to existing imaging practices, both of which may be

addressed through the work of professional organizations.

Initiatives such as Image Gently, Image Wisely, Afrosafe and Eurosafe Imaging [84], driven by

professional organizations, have created awareness and shared knowledge about radiation

Page 32 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

protection with healthcare providers and patients via smart, stakeholder-targeted educational

materials [85]. The value of such tailor made content cannot be underestimated, as these

groups have a deep understanding of the interests and motivations of the professionals they

represent and are able to convey their message in a compelling and relevant manner [86].

Changing existing imaging practices, in terms of clinical indications for ordering imaging

procedures as well as protocols and techniques of performing imaging procedures, demands

strong leadership from professional bodies. Guidelines such as the ACR Appropriateness

Criteria [87], and the European Society of Radiology iGuide [88] provide clear direction to

referring clinicians to align themselves with current best practice. Part of the reluctance to

adopt lower dose protocols may stem from the fear that lower image quality may lead to

missing alternate pathologies that may account for symptoms (for example cholecystitis

mimicking renal colic) or significant incidentals, which, as discussed earlier in this paper, are

legitimate concerns. Here guidelines also play an important role by clearly delineating clinical

scenarios where lower dose protocols are appropriate, and indeed, the expected standard of

care. This may allay fears of litigation, as one key element in medical malpractice claims is

proof of deviation from accepted standards.

Equipment manufacturers have long played an important role in dose reduction by steady

improvements in hardware and software. Through the work of trade associations such as

Global Diagnostic Imaging, Healthcare IT and Radiation Therapy Trade Association (DITTA),

Medical Imaging and Technology Alliance (MITA), and European Coordination Committee of

the Radiological, Electromedical, and Healthcare IT Industry (COCIR), they have also made

vital contributions in the areas of standardization, quality assurance, and dose monitoring, all

Page 33 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

integral elements of dose optimization efforts. Looking forward, we see how collaboration

with industry can assist in the rapid promulgation of new dose metrics, CT imaging protocols

and dose reduction techniques through software and hardware modifications and user

training. Working together with international organizations and professional bodies, industry

can also help build the infrastructure and technical solutions needed for standardized dose

reporting and tracking. However, a significant challenge is the need for effective cooperation

and communication between different manufacturers, a task that the trade associations are

already attempting to accomplish.

Legislation has proved to be a very effective means to raise quality and safety standards in

imaging. In Europe, under the Euratom treaty, member states are uniformly required to

conform to standards laid out in the medical exposure directive [72]. Consequently, 81% of EU

and EFTA countries had established adult x-ray DRLs by the year 2014 [89], a remarkable

accomplishment given the challenges and complexities involved. In the United States, the

passing of the Mammography Quality Standards Act of 1992 brought about a significant

improvement in the quality of mammography, reflected in a significant improvement in first

attempt pass rates for accreditation of mammography units between 1991 and 2003 [90].

However, use of legislation should be judicious and carefully considered in order to avoid

potential regulatory pitfalls including cost of interventions and potential negative impact on

access to and provision of care [91]. Voluntary adherence to standards put forward by

professional organizations is not necessarily inferior to regulatory control and should not be

arbitrarily supplanted by formal legislation.

Page 34 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

Conclusions

Patient risk optimization of medical imaging practice is essential for three reasons. First, it is

an ethical necessity based on the trust relationship between the provider and the patient; it is

the right thing to do. Second, it is a professional necessity to improve consistency in the

practice of medicine, and excel in its drive towards continuous improvement, as mandated in

many countries. Finally, it is an economically responsible necessity to reduce the likelihood of

misdiagnoses due to poor or inconsistent quality, to reduce the litigation risk, and to help

improve equipment lifespan. Optimization in imaging is made possible by first characterizing

the dose and image quality attributes of an exam in terms of surrogates that are ideally as

patient-informed as possible. That follows with a nuanced balance between the two, taking

into consideration the image task and the attributes of the patient, such that the total risk to

the patient (including radiation risk and clinical risk associated with a sub-quality exam) is

minimized. This strategy should accommodate the statistical variability across patients and

across the clinical process. The optimization is enabled by a pragmatic deployment plan, a

specialized workforce including expert medical physicists, and an informatics infrastructure. It

is also highly encouraged and facilitated by intentional and tactful regulations and professional

guidelines.

Acknowledgment

The authors gratefully acknowledge the assistance of Francesco Ria, Luisa Pierotti, and Luca

Curovic with this manuscript.

Page 35 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

Page 36 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

References

1. ICRP, 1990 Recommendations of the International Commission on Radiological Protection,

ICRP publication 60, 1991, Ann. ICRP 21 (1-3).

2. IAEA, Radiation protection and safety of radiation sources: international basic safety

standard, IAEA Safety Standards Series n. GSR Part 3, International Atomic Energy Agency,

2014, Vienna.

3. Järvinen H., Vassileva J., Samei E., Wallace A, Vano E. and Rehani M., Patient dose

monitoring and the use of diagnostic reference levels for the optimization of protection in

medical imaging: current status and challenges worldwide, 2017 Journal of Medical Imaging

4(3), 031214.

4. https://www.acr.org/Quality-Safety/Appropriateness-Criteria (accessed Nov. 11, 2017).

5. UNSCEAR 2008, Sources and Effects of Ionizing Radiation: United Nations Scientific

Committee on the Effects of Atomic Radiation 2008 Report to the General Assembly, UN

2010.

6. Hricak, H., D.J. Brenner, S.J. Adelstein, D.P. Frush, E.J. Hall, R.W. Howell, C.H. McCollough,

F.A. Mettler, M.S. Pearce, O.H. Suleiman, J.H. Thrall, and L.K. Wagner, Managing radiation

use in medical imaging: a multifaceted challenge, Radiology, 2011, 258(3): p. 889-905. PMID:

21163918.

7. Dorfman, A.L., R. Fazel, A.J. Einstein, K. Applegate, H.M. Krumholz, Y. Wang, E.

Christodoulou, J. Chen, R. Sanchez, and B.K. Nallamothu, Use of medical imaging

procedures with ionizing radiation in children: A population-based study. Archives of

Pediatrics & Adolescent Medicine, 2011, 165(5): p. 458-464. PMCID:PMC 3686496.

Page 37 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

8. UNSCEAR 2013, Sources and Effects of Ionizing Radiation United Nations Scientific

Committee on the Effects of Atomic Radiation 2013 Report to the General Assembly, Effects of

radiation exposure of children, UN Vol II Annex B

9. NCRP Report No. 160, Ionizing Radiation Exposure of the Population of the United States,

2009, National Council on Radiation Protection and Measurements: Bethesda, MD. PMID:

19509507

10. Radiation Protection n. 180 part 1, Medical Radiation Exposure of the European Population,

European Commission 2014.

11. Smith-Bindman, R., J. Lipson, R. Marcus, K.P. Kim, M. Mahesh, R. Gould, A. Berrington de

Gonzalez, and D.L. Miglioretti, Radiation dose associated with common computed

tomography examinations and the associated lifetime attributable risk of cancer, Arch Intern

Med, 2009, 169(22): p. 2078-86. PMID: 20008690.

12. National Research Council, Health Risks from Exposure to Low Levels of Ionizing Radiation:

BEIR VII Phase 2 2006.

13. Foray, N., C. Colin, and M. Bourguignon, 100 years of individual radiosensitivity: how we

have forgotten the evidence, Radiology, 2012, 264(3): p. 627-31. PMID: 22919036.

14. ICRP, Genetic susceptibility to cancer, ICRP publication 79, Ann ICRP, 1998, 28(1-2): p. 1-157.

PMID: 10406427.

15. Pijpe, A., N. Andrieu, D.F. Easton, et al., Exposure to diagnostic radiation and risk of breast

cancer among carriers of BRCA1/2 mutations: retrospective cohort study (GENE-RAD-RISK).

BMJ, 2012. 345: p. e5660.

Page 38 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

16. Mathews, J., A. Forsythe, Z. Brady, M. Butler, S. Goergen, G. Byrnes, G. Giles, A. Wallace,

P. Anderson, T. Guiver, P. McGale, T. Cain, J. Dowty, A. Bickerstaffe, and S. Darby, Cancer

risk in 680 000 people exposed to computed tomography scans in childhood or adolescence:

data linkage study of 11 million Australians, BMJ, 2013, 346. PMID: 23694687.

17. Miglioretti, D.L., E. Johnson, A. Williams, R.T. Greenlee, S. Weinmann, L.I. Solberg, H.S.

Feigelson, D. Roblin, M.J. Flynn, N. Vanneman, and R. Smith-Bindman, The use of

computed tomography in pediatrics and the associated radiation exposure and estimated

cancer risk, JAMA Pediatrics, 2013, 167(8): p. 700-707. PMID: 23754213.

18. Pearce, M.S., J.A. Salotti, M.P. Little, K. McHugh, C. Lee, K.P. Kim, N.L. Howe, C.M.

Ronckers, P. Rajaraman, A.W. Craft, L. Parker, and A. Berrington de González, Radiation

exposure from CT scans in childhood and subsequent risk of leukaemia and brain tumours: a

retrospective cohort study, The Lancet, 2012, 380(9840): p. 499-505. PMCID:PMC 3418594.

19. Hall, E.J. and A.J. Giaccia, Radiobiology for the Radiologist, 7th ed. 2012, Philadelphia, PA:

Lippincott Williams & Wilkins.

20. ICRP, The 2007 Recommendations of the International Commission on Radiological

Protection, ICRP publication 103, 2007, Ann ICRP 37:1–332.

21. Smith-Bindman, R., J. Lipson, R. Marcus, K.P. Kim, M. Mahesh, R. Gould, A. Berrington de

Gonzalez, and D.L. Miglioretti, Radiation dose associated with common computed

tomography examinations and the associated lifetime attributable risk of cancer, Arch Intern

Med, 2009, 169(22): p. 2078-86. PMID: 20008690.

22. Brenner, D.J. and E.J. Hall, Computed tomography--an increasing source of radiation

exposure, N Engl J Med, 2007, 357(22): p. 2277-84.

Page 39 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

23. The Joint Commission, The Joint Commission Sentinel Event Alert: Radiation risks of

diagnostic imaging, 2011, PMID: 21887890.

24. Hendee, W.R. and M.K. O'Connor, Radiation risks of medical imaging: separating fact from

fantasy. Radiology, 2012. 264(2): p. 312-21. PMID: 22821690.

25. Tubiana, M., A. Aurengo, D. Averbeck, and R. Masse, The debate on the use of linear no

threshold for assessing the effects of low doses, Journal of Radiological Protection, 2006,

26(3): p. 317. PMID: 16926474.

26. Hendee, W.R., G.J. Becker, J.P. Borgstede, J. Bosma, W.J. Casarella, B.A. Erickson, C.D.

Maynard, J.H. Thrall, and P.E. Wallner, Addressing overutilization in medical imaging,

Radiology, 2010, 257(1): p. 240-5. PMID: 20736333.

27. McCollough, C.H., G.H. Chen, W. Kalender, S. Leng, E. Samei, K. Taguchi, G. Wang, L. Yu,

and R.I. Pettigrew, Achieving Routine Submillisievert CT Scanning: Report from the Summit

on Management of Radiation Dose in CT, Radiology, 2012, 264(2): p. 567-580. PMCID:PMC

3401354.

28. Gilbert W. Beebe Symposium, Tracking Radiation Exposure from Medical Diagnostic

Procedures, 2011, Washington, DC: National Academy of Sciences.

29. ABR third Summit Conference, Safe Use in Medical Imaging: Developing a Systematic and

Patient Centered Approach, 2012, American Board of Radiology Foundation.

30. FDA, Draft Guidance for Industry and Food and Drug Administration Staff: Pediatric

Information for X-ray Imaging Device Premarket Notifications, in Guidance Documents

(Medical Devices and Radiation-Emitting Products), 2012, U.S. Department of Health and

Human Services, Rockville (MD).

Page 40 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

31. Frush DP, Denham CR, Goske MJ, et al. Radiation protection and dose monitoring in medical

imaging: A journey from awareness, through accountability, ability and action... but where

will we arrive?, Journal of Patient Safety, 2013, 8:1-11.

32. American College of Radiology. ACR Imaging 3.0, 2013.

http://www.acr.org/Advocacy/Economics-Health-Policy/Imaging-3.

33. Brink, J.A. and E.S. Amis, Jr., Image Wisely: a campaign to increase awareness about adult

radiation protection, Radiology, 2010, 257(3): p. 601-2. PMID: 21084410.

34. Goske, M.J., K.E. Applegate, J. Boylan, P.F. Butler, M.J. Callahan, B.D. Coley, S. Farley, D.P.

Frush, M. Hernanz-Schulman, D. Jaramillo, N.D. Johnson, S.C. Kaste, G. Morrison, and K.J.

Strauss, Image Gently(SM): a national education and communication campaign in radiology

using the science of social marketing, Journal of the American College of Radiology, 2008,

5(12): p. 1200-5. PMID: 19027684.

35. ICRP, Radiological Protection in Paediatric Diagnostic and Interventional Radiology, ICRP

publication 121, 2013, Ann ICRP 42(2): 1-66

36. F. Ria, J. M. Wilson, Y Zhang, E. Samei, Image noise and dose performance across a clinical

population: Patient size adaptation as a metric of CT performance, Medical Physics, 2017,

44(6): 2141-2147.

37. A. Bejan and S. Lorente, Constructal theory of generation of configuration in nature and

engineering, Journal of Applied Physics, 2006, 100(4), 041301.

38. A. Bejan and J. H. Marden, Unifying constructal theory for scale effects in running, swimming

and flying, Journal of Experimental Biology, 2006, 209, 238-248.

Page 41 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

39. A. Bejan and S. Lorente, Constructal law of design and evolution: Physics, biology,

technology, and society, Journal of Applied Physics, 2013, 113, 151301.

40. A. Bejan, J. D. Charles and S. Lorente, The evolution of airplanes, Journal of Applied Physics,

2014, 116, 044901.

41. A. Bejan, Evolution in thermodynamics, Applied Physics Reviews, 2017, 4, 011305.

42. Samei E., Tian X., Paul Segars W., Frush D.P., Radiation risk index for pediatric CT: a patient-

derived metric, 2017, Pediatric Radiology, 47(13):1737-1744.

43. IAEA, Applying radiation safety standards in diagnostic radiology and interventional

procedures using x rays, International Atomic energy Agency, 2006, Safety Reports Series

no. 39.

44. National Academies of Sciences, Engineering, and Medicine, Improving diagnosis in health

care, 2015, The National Academies Press, Washington DC.

45. AAPM report n. 96, The Measurement, Reporting, and Management of Radiation Dose in CT,

Report of AAPM Task Group 23, 2008, American Association of Physicist in Medicine.

46. McCollough, C. H., Primak, A. N., Braun, N., Kofler, J., Yu, L., & Christner, J., Strategies for

Reducing Radiation Dose in CT. Radiologic Clinics of North America, 2009, 47(1), 27–40.

47. IAEA, Dosimetry in Diagnostic Radiology: An International Code of Practice. International

Atomic Energy Commission Technical Reports Series No. 457, 2007, STI/DOC/010/457;

ISBN:92-0-115406-2.

48. IAEA, Diagnostic Radiology Physics, A Handbook for Teachers and Students, International

Atomic Energy Commission, 2014, STI/PUB/1564, ISBN:978-92-0-131010-1.

Page 42 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

49. ICRP, Diagnostic Reference Levels in Medical Imaging, ICRP Publication 135, 2017, Ann. ICRP

46(1).

50. IAEA, Nuclear medicine physics: A handbook for teachers and students, International Atomic

Energy Commission, 2014, STI/PUB/1617, ISBN 978–92–0–143810–2.

51. AAPM Report no. 204, Size specific dose estimates (SSDE) in pediatric and adult body CT

examinations, Report of AAPM Task Group 204, 2011, American Association of Physicist in

Medicine.

52. Samei E. Christianson O, Zhang Y., Comment on “Comparison of patient specific dose metrics

between chest radiography, tomosynthesis, and CT for adult patients of wide ranging body

habitus”, 2015, Medical Physics 42(4): 02094-2095.

53. Sahbaee P, Segars WP, Samei E., Patient-based estimation of organ dose for adult

population across a wide range of protocols, 2014, Medical Physics 41: 072104.

54. Hart D., Jones D. G., Wall B. F., Estimation of Effective Dose in Diagnostic Radiology from

Entrance Surface Dose and Dose-Area Product Measurements, 1994, National Radiological

Protection Board, R262.

55. Li X., Samei E., Segars W., Sturgeon G., Colsher J., Frush D. P., Patient-specific dose and risk

estimation in pediatric chest CT, 2011, Radiology 259(3): 862-874.

56. Ludewig E., Richter A., Frame M. Diagnostic imaging--evaluating image quality using visual

grading characteristic (VGC) analysis, 2010, Vet Res Commun. 34(5):473-9.

57. Martin C.J., The Importance of Radiation Quality for Optimisation in Radiology, Biomedical

Imaging and Intervention Journal, 2007, 3.2 e38.

Page 43 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

58. Metz C. E., ROC analysis in medical imaging: a tutorial review of the literature, Radiological

physics and technology, 2008, 1(1):2-12.

59. Verdun F. R., Racine D., Ott J. G., et al., Image quality in CT: From physical measurements to

model observers, 2015, Physica Medica 31:23–843.

60. Solomon J., Samei E., Correlation between human detection accuracy and observer model-

based image quality metrics in CT, Journal of Medical Imaging, 2016, 3(3): 035506.

61. Smith T., Solomon J. B,, Samei E., In vivo detectability index: development and validation of

an automated methodology. SPIE International Symposium on Medical Imaging, Orlando,

FA, February 2017, Proc. SPIE Medical Imaging, 2017.

62. Richard S., Samei E., Quantitative imaging in breast tomosynthesis and CT: comparison of

detection and estimation task performance, Medical Physics, 2010, 37(6): 2627-2637.

63. European Society of Radiology, ESR statement on radiation protection: globalisation,

personalised medicine and safety (the GPS approach), Insights Imaging, 2013, p. 737-9.

PMID: 3846938.

64. Hood, L., Friend S. H., Predictive, personalized, preventive, participatory (P4) cancer

medicine, Nature reviews Clinical Oncology, 2011, 8(3):184-187.

65. Litjens G., Kooi T., Bejnordi B. E., et al., A Survey on Deep Learning in Medical Image

Analysis, Medical Image Analysis, 2017, 42:60-68.

66. Oakden-Rayner L., Carneiro G., Bessen T., et al., Precision Radiology: Predicting longevity

using feature engineering and deep learning methods in a radiomics framework, Scientific

Reports, 2017, 7(1):1648.

Page 44 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

67. Radiation Protection n. 162, Criteria for Acceptability of Medical Radiological Equipment used

in Diagnostic Radiology, Nuclear Medicine and Radiotherapy, 2012, European Commission.

68. McCollough, C. H., Primak, A. N., Braun, N., Kofler, J., Yu, L., & Christner, J, Strategies for

Reducing Radiation Dose in CT, 2009, Radiologic Clinics of North America, 47(1), 27–40.

69. Christianson O., Li X., Frush D. P., Samei E., Automated size-specific CT dose monitoring

program: assessing variability in CT dose, 2012, Medical Physics, 32:7131-7138.

70. Radiological Society of North America, RadLex Playbook and the LOINC/RSNA Radiology

Playbook, https://www.rsna.org/RadLex_Playbook.aspx;

http://playbook.radlex.org/playbook/SearchRadlexAction.

71. NEMA, DICOM – Digital Imaging and Communication in Medicine, http://dicom.nema.org.

72. Council Directive 2013/59/EURATOM of 5 December 2013, Laying down basic safety

standards for protection against the dangers arising from exposure to ionising radiation, and

repealing Directives 89/618/Euratom, 90/641/Euratom, 96/29/Euratom, 97/43/Euratom and

2003/122/Euratom, Official Journal of the European Union, 2014, L 13/1.

73. Brix, G., Nagel, H. D., Stamm, G., Veit, R., Lechel, U., Griebel, J. and Galanski, M., Radiation

exposure in multi-slice versus single-slice spiral CT: results of a nationwide survey, European

Radiology, 2003, 13, 1979–1991.

74. Image wisely, Radiation safety in adult medical imaging, http://www.imagewisely.org/

75. Image Gently, The Image Gently alliance, http://www.imagegently.org/

76. ICRP, Managing Patient Dose in Multi-Detector Computed tomography (MDCT), ICRP

publication 102, 2007, Ann. ICRP 37(1).

77. ICRP, Radiological Protection in Medicine, ICRP publication 105, 2007, Ann. ICRP 37(6).

Page 45 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

78. European Commission, European guidelines on quality criteria for computed tomography,

Official Publication of the European Communities, 2000, EUR 16262.

79. Bongartz G., Golding S. J., Jurik A. G., et al., European Guidelines for Multislice Computed

Tomography, Funded by the European Commission, 2004, Contract number FIGM-CT2000-

20078-CT-TIP.

80. IAEA, Radiation dose management in computed tomography, IAEA launches new E-

learning, Radiation Protection of Patients (RPOP),

https://rpop.iaea.org/RPOP/RPoP/Content/News/10-e-learning.htm.

81. Adam N.R., Wieder R., Ghosh D., Data science, learning, and applications to biomedical and

health sciences, 2017, Annals of the New York Academy of Science, 1387(1):5-11.

82. Vassileva J. and Rehani M., Diagnostic reference levels, 2015, American Journal of

Roentgenology 204.1: W1-W3.

83. Lukasiewicz A., Bhargavan-Chatfield M., Coombs L., et al., Radiation dose index of renal

colic protocol CT studies in the United States: a report from the American College of Radiology

National Radiology Data Registry, 2014,Radiology 271(2):445-451.

84. ESR, Eurosafe Imaging, European Society of Radiology, http://www.eurosafeimaging.org/

85. Goske M. J., Applegate K. E., Bulas D., et al., Image Gently: progress and challenges in CT

education and advocacy, Pediatric radiology, 2011, 41-2: 461-466.

86. Ria F., Bergantin A., Vai A. et al., Awareness of medical radiation exposure among patients: A

patient survey as a first step for effective communication of ionizing radiation risks, Physica

Medica, 2017, 43:57-62.

Page 46 of 47AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

87. ACR, Appropriateness Criteria, American College of Radiology https://www.acr.org/Quality-

Safety/Appropriateness-Criteria.

88. ESR, ESR iGuide Clinical decision support using European imaging referral guidelines,

European society of Radiology, https://www.myesr.org/quality-safety/esr-iguide.

89. Radiation Protection n. 180 part 2, Diagnostic Reference Levels in thirty-six European

Cuntries, European Commission 2014.

90. Destouet J. M., Bassett L. W., Yaffe M. J., et al., The ACR’s Mammography Accreditation

Program: ten years of experience since MQSA, 2005, Journal of the American College of

Radiology 2(7):585-594.

91. Harvey H. B., and Pandharipande P. V., The federal government’s oversight of CT safety:

regulatory possibilities, 2012, Radiology 262(2):391-398.

Page 47 of 47 AUTHOR SUBMITTED MANUSCRIPT - JRP-101227.R2

Accepted Manuscript

A method for measuring spatial resolution based on clinical chest CT sequence images

Article

Full-text available

Mar 2025
J APPL CLIN MED PHYS

Purpose This study aimed to develop and validate a method for characterizing the spatial resolution of clinical chest computed tomography (CT) sequence images. Methods An algorithm for characterizing spatial resolution based on clinical chest CT sequence images was developed in Matlab (2021b). The algorithm was validated using CT sequence images from a custom‐made chest automatic tube current modulation (ATCM) phantom and clinically reconstructed chest CT sequence images. A region of interest (ROI) was automatically established at the edges of CT image subject to calculate the edge spread function (ESF). The ESF curves from consecutive CT images within the same sequence were fitted into a curve, and the line spread function (LSF) was derived through differentiation. A Fourier transformation of the LSF curve was conducted to obtain the modulation transfer function (MTF). The method's effectiveness was verified by comparing the 50% MTF and 10% MTF values with those calculated using IndoQCT (22a) software. The method was also applied to clinical CT images to calculate MTF values for various reconstructions, confirming its sensitivity by determining spatial resolution of clinically reconstructed images. Results Validation experiments based on the phantom CT sequence images demonstrated that the MTF values calculated using the proposed method had an average difference of within ± 5% compared to the results obtained with IndoQCT. Validation experiments with clinical CT sequence images indicated that the method effectively reflects differences and variations in spatial resolution of images under different reconstruction kernels, with the MTF values for B10f‐B50f and D10f‐D50f exhibiting a consistent increase. Conclusion A method for measuring spatial resolution using clinical chest CT sequence images was developed. This method provides a direct means of spatial resolution characterization for clinical CT datasets and a more accurate representation of CT imaging quality, effectively reflects variations across different reconstruction convolution kernels, demonstrating its sensitivity.

Measurement/Recording of Patient Radiation Exposure and Dose Tracking Systems

Article

Mar 2025

Optimization of abdominal CT based on a model of total risk minimization by putting radiation risk in perspective with imaging benefit

Article

Full-text available

Dec 2024

Background Risk-versus-benefit optimization required a quantitative comparison of the two. The latter, directly related to effective diagnosis, can be associated to clinical risk. While many strategies have been developed to ascertain radiation risk, there has been a paucity of studies assessing clinical risk, thus limiting the optimization reach to achieve a minimum total risk to patients undergoing imaging examinations. In this study, we developed a mathematical framework for an imaging procedure total risk index considering both radiation and clinical risks based on specific tasks and investigated diseases. Methods The proposed model characterized total risk as the sum of radiation and clinical risks defined as functions of radiation burden, disease prevalence, false-positive rate, expected life-expectancy loss for misdiagnosis, and radiologist interpretative performance (i.e., AUC). The proposed total risk model was applied to a population of one million cases simulating a liver cancer scenario. Results For all demographics, the clinical risk outweighs radiation risk by at least 400%. The optimization application indicates that optimizing typical abdominal CT exams should involve a radiation dose increase in over 90% of the cases, with the highest risk optimization potential in Asian population (24% total risk reduction; 306%

{{CTDI}}_{{vol}}

increase) and lowest in Hispanic population (5% total risk reduction; 89%

{{CTDI}}_{{vol}}

increase). Conclusions Framing risk-to-benefit assessment as a risk-versus-risk question, calculating both clinical and radiation risk using comparable units, allows a quantitative optimization of total risks in CT. The results highlight the dominance of clinical risk at typical CT examination dose levels, and that exaggerated dose reductions can even harm patients.

An inventory of patient-image based risk/dose, image quality and body habitus/size metrics for adult abdomino-pelvic CT protocol optimisation

Article

Full-text available

Aug 2024
PHYS MEDICA

An image quality assessment index based on image features and keypoints for X-ray CT images

Article

Full-text available

Jul 2024
PLOS ONE

Optimization tasks in diagnostic radiological imaging require objective quantitative metrics that correlate with the subjective perception of observers. However, although one such metric, the structural similarity index (SSIM), is popular, it has limitations across various aspects in its application to medical images. In this study, we introduce a novel image quality evaluation approach based on keypoints and their associated unique image feature values, focusing on developing a framework to address the need for robustness and interpretability that are lacking in conventional methodologies. The proposed index quantifies and visualizes the distance between feature vectors associated with keypoints, which varies depending on changes in the image quality. This metric was validated on images with varying noise levels and resolution characteristics, and its applicability and effectiveness were examined by evaluating images subjected to various affine transformations. In the verification of X-ray computed tomography imaging using a head phantom, the distances between feature descriptors for each keypoint increased as the image quality degraded, exhibiting a strong correlation with the changes in the SSIM. Notably, the proposed index outperformed conventional full-reference metrics in terms of robustness to various transformations which are without changes in the image quality. Overall, the results suggested that image analysis performed using the proposed framework could effectively visualize the corresponding feature points, potentially harnessing lost feature information owing to changes in the image quality. These findings demonstrate the feasibility of applying the novel index to analyze changes in the image quality. This method may overcome limitations inherent in conventional evaluation methodologies and contribute to medical image analysis in the broader domain.

Optimization of CT pulmonary angiography for pulmonary embolism using task-based image quality assessment and diagnostic reference levels: A multicentric study

Article

Apr 2024

Variability Between Radiation-Induced Cancer Risk Models in Estimating Oncogenic Risk in Intensive Care Unit Patients

Article

Apr 2025

Purpose: To evaluate the variability of oncogenic risk related to radiation exposure in patients frequently exposed to ionizing radiation for diagnostic purposes, specifically ICU patients, according to different risk models, including the BEIR VII, ICRP 103, and US EPA models. Methods: This was an IRB-approved observational retrospective study. A total of 71 patients (58 male, 13 female; median age, 66 years; interquartile range [IQR], 65–71 years) admitted to the ICU who underwent X-ray examinations between 1 October 2021 and 28 February 2023 were included. For each patient, the cumulative effective dose during a single hospital admission was calculated. Lifetime attributable risk (LAR) was estimated based on the BEIR VII, ICRP 103, and US EPA risk models to calculate additional oncogenic risk related to radiation exposure. The Friedman test for repeated-measures analysis of variance was used to compare risk values between different models. The intraclass correlation coefficient (ICC) was used to assess the consistency of risk values between different models. Results: Different organ, leukemia, and all-cancer risk values estimated according to different oncogenic risk models were significantly different, but the intraclass correlation coefficient revealed a good (>0.75) or even excellent (>0.9) agreement between different risk models. The ICRP 103 model estimated a lower all-cancer (median 69.05 [IQR 30.35–195.37]) and leukemia risk (8.22 [3.02–27.93]) compared to the US EPA (all-cancer: 139.68 [50.51–416.16]; leukemia: 23.34 [3.47–64.37]) and BEIR VII (all-cancer: 162.08 [70.6–371.40]; leukemia: 24.66 [12.9–58.8]) models. Conclusions: Cancer risk values were significantly different between risk models, though inter-model agreement in the consistency of risk values was found to be good, or even excellent.

A strategy for achieving optimisation of radiological protection in digital radiology proposed by ICRP

Article

Nov 2024
J RADIOL PROT

Radiology is now predominantly a digital medium and this has extended the flexibility, efficiency and application of medical imaging. Achieving the full benefit of digital radiology requires images to be of sufficient quality to make a reliable diagnosis for each patient, while minimising risks from radiation exposure, and so involves a careful balance between competing objectives. When an optimisation programme is undertaken, a knowledge of patient doses from surveys can be valuable in identifying areas needing attention. However, any dose reduction measures must not degrade image quality to the extent that it is inadequate for the clinical purpose. The move to digital imaging has enabled versatile image acquisition and presentation, including multi-modality display and quantitative assessment, with post-processing options that adjust for optimal viewing. This means that the appearance of an image is unlikely to give any indication when the dose is higher than necessary. Moreover, options to improve performance of imaging equipment add to its complexity, so operators require extensive training to be able to achieve this. Optimisation is a continuous rather than single stage process that requires regular monitoring, review, and analysis of performance feeding into improvement and development of imaging protocols. The ICRP is in the process of publishing two reports about optimisation in digital radiology. The first report sets out components needed to ensure that a radiology service can carry optimisation through. It describes how imaging professionals should work together as a team and explains the benefits of having appropriate methodologies to monitor performance, together with the knowledge and expertise required to use them effectively. It emphasises the need for development of organisational processes that ensure tasks are carried out. The second ICRP report deals with practical requirements for optimisation of different digital radiology modalities, and builds on information provided in earlier modality specific ICRP publications.

Fully automated measurement of noise, signal-to-noise ratio, and contrast-to-noise ratio on chest CT images: feasibility and efficiency

Article

Oct 2024
ACTA RADIOL

Background Rapid and accurate measurement of computed tomography (CT) image noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) is a clinical challenge. Purpose To explore the feasibility of intelligent measurement of chest CT image noise, SNR, and CNR. Material and Methods A total of 300 chest CT scans were included in the study, which was divided into research dataset, internal test dataset, and external test dataset. Based on the research dataset, automatically segment and measure the average CT values and standard deviation (SD) of CT values for background air and lung field under different thresholds to obtain noise, SNR, and CNR results. Using the results of manual measurements as the reference standard, we determine the optimal threshold with the highest consistency. Using internal and external test datasets, validate the consistency of automated measurements of noise, SNR, and CNR at the optimal CT threshold with reference standards. Results With background air set at −900 HU and lung field at −800 HU as thresholds, the automated measurements of noise, SNR, and CNR demonstrate the highest consistency with the reference standards. At the optimal threshold, the noise, SNR, and CNR measured automatically on both the internal (intraclass correlation coefficient [ICC] = 0.85–0.96) and external (ICC = 0.75–0.85) test datasets exhibit high consistency with their respective reference standards. Conclusion The method we explored can intelligently measure the noise, SNR, and CNR of chest CT images, exhibits high consistency with radiologists, and offers a novel tool for image quality evaluation and analysis.

Optimisation of Protection in Medical Imaging: Necessary, Challenging, and Possible

Article

Jul 2023
Ann ICRP

Awareness of medical radiation exposure among patients: A patient survey as a first step for effective communication of ionizing radiation risks

Article

Full-text available

Oct 2017
PHYS MEDICA

Introduction: The European Directive 2013/59/EURATOM requires patient radiation dose information to be included in the medical report of radiological procedures. To provide effective communication to the patient, it is necessary to first assess the patient's level of knowledge regarding medical exposure. The goal of this work is to survey patients' current knowledge level of both medical exposure to ionizing radiation and professional disciplines and communication means used by patients to garner information. Material and methods: A questionnaire was designed comprised of thirteen questions: 737 patients participated in the survey. The data were analysed based on population age, education, and number of radiological procedures received in the three years prior to survey. Results: A majority of respondents (56.4%) did not know which modality uses ionizing radiation. 74.7% had never discussed with healthcare professionals the risk concerning their medical radiological procedures. 70.1% were not aware of the professionals that have expertise to discuss the use of ionizing radiation for medical purposes, and 84.7% believe it is important to have the radiation dose information stated in the medical report. Conclusion: Patients agree with new regulations that it is important to know the radiation level related to the medical exposure, but there is little awareness in terms of which modalities use X-Rays and the professionals and channels that can help them to better understand the exposure information. To plan effective communication, it is essential to devise methods and adequate resources for key professionals (medical physicists, radiologists, referring physicians) to convey correct and effective information.

Radiation risk index for pediatric CT: a patient-derived metric

Article

Full-text available

Dec 2017
PEDIATR RADIOL

Background: There is a benefit in characterizing radiation-induced cancer risk in pediatric chest and abdominopelvic CT: a singular metric that represents the whole-body radiation burden while also accounting for age, gender and organ sensitivity. Objective: To compute an index of radiation risk for pediatric chest and abdominopelvic CT. Materials and methods: Using a protocol approved by our institutional review board, 42 pediatric patients (age: 0-16 years, weight: 2-80 kg) were modeled into virtual whole-body anatomical models. Organ doses were estimated for clinical chest and abdominopelvic CT examinations of the patients using validated Monte Carlo simulations of two major scanner models. Using age-, size- and gender-specific organ risk coefficients, the values were converted to normalized effective dose (by dose length product) (denoted as the k factor) and a normalized risk index (denoted as the q factor). An analysis was performed to determine how these factors are correlated with patient age and size for both males and females to provide a strategy to better characterize individualized risk. Results: The k factor was found to be exponentially correlated with the average patient diameter. For both genders, the q factor also exhibited an exponential relationship with both the average patient diameter and with patient age. For both factors, the differences between the scanner models were less than 8%. Conclusion: The study defines a whole-body radiation risk index for chest and abdominopelvic CT imaging, that incorporates individual estimated organ dose values, organ radiation sensitivity, patient size, exposure age and patient gender. This indexing metrology enables the assessment and potential improvement of chest and abdominopelvic CT performance through surveillance of practice dose profiles across patients and may afford improved informed communication.

Precision Radiology: Predicting longevity using feature engineering and deep learning methods in a radiomics framework

Article

Full-text available

Dec 2017

Precision medicine approaches rely on obtaining precise knowledge of the true state of health of an individual patient, which results from a combination of their genetic risks and environmental exposures. This approach is currently limited by the lack of effective and efficient non-invasive medical tests to define the full range of phenotypic variation associated with individual health. Such knowledge is critical for improved early intervention, for better treatment decisions, and for ameliorating the steadily worsening epidemic of chronic disease. We present proof-of-concept experiments to demonstrate how routinely acquired cross-sectional CT imaging may be used to predict patient longevity as a proxy for overall individual health and disease status using computer image analysis techniques. Despite the limitations of a modest dataset and the use of off-the-shelf machine learning methods, our results are comparable to previous ‘manual’ clinical methods for longevity prediction. This work demonstrates that radiomics techniques can be used to extract biomarkers relevant to one of the most widely used outcomes in epidemiological and clinical research – mortality, and that deep learning with convolutional neural networks can be usefully applied to radiomics research. Computer image analysis applied to routinely collected medical images offers substantial potential to enhance precision medicine initiatives.

A Survey on Deep Learning in Medical Image Analysis

Article

Full-text available

Feb 2017

Deep learning algorithms, in particular convolutional networks, have rapidly become a methodology of choice for analyzing medical images. This paper reviews the major deep learning concepts pertinent to medical image analysis and summarizes over 300 contributions to the field, most of which appeared in the last year. We survey the use of deep learning for image classification, object detection, segmentation, registration, and other tasks and provide concise overviews of studies per application area. Open challenges and directions for future research are discussed.

Estimating detectability index in vivo: Development and validation of an automated methodology

Article

Dec 2017

This study's purpose was to develop and validate a method to estimate patient-specific detectability indices directly from patients' CT images (i.e., in vivo). The method extracts noise power spectrum (NPS) and modulation transfer function (MTF) resolution properties from each patient's CT series based on previously validated techniques. These are combined with a reference task function (10-mm disk lesion with [Formula: see text] HU contrast) to estimate detectability indices for a nonprewhitening matched filter observer model. This method was applied to CT data from a previous study in which diagnostic performance of 16 readers was measured for the task of detecting subtle, hypoattenuating liver lesions ([Formula: see text]), using a two-alternative-forced-choice (2AFC) method, over six dose levels and two reconstruction algorithms. In vivo detectability indices were estimated and compared to the human readers' binary 2AFC outcomes using a generalized linear mixed-effects statistical model. The results of this modeling showed that the in vivo detectability indices were strongly related to 2AFC outcomes ([Formula: see text]). Linear comparison between human-detection accuracy and model-predicted detection accuracy (for like conditions) resulted in Pearson and Spearman correlation coefficients exceeding 0.84. These results suggest the potential utility of using in vivo estimates of a detectability index for an automated image quality tracking system that could be implemented clinically.

Patient dose monitoring and the use of diagnostic reference levels for the optimization of protection in medical imaging: Current status and challenges worldwide

Article

Oct 2017

Optimization is one of the key concepts of radiation protection in medical imaging. In practice, it involves compromising between the image quality and dose to the patient; the dose should not be higher than necessary to achieve an image quality (or diagnostic information) needed for the clinical task. Monitoring patient dose is a key requirement toward optimization. The concept of diagnostic reference level (DRL) was introduced by the International Commission on Radiological Protection as a practical tool for optimization. Unfortunately, this concept has not been applied consistently worldwide. To review the current strengths and weaknesses worldwide and to promote improvements, the International Atomic Energy Agency organized a Technical Meeting on patient dose monitoring and the use of DRLs on May 2016. This paper reports a summary of the findings and conclusions from the meeting. The strengths and weaknesses were generally different in less-developed countries compared with developed countries. Possible improvements were suggested in six areas: human resources and responsibilities, training, safety and quality culture, regulations, funding, and tools and methods. An overall conclusion was that radiation protection requires a patient-centric approach and a transfer from purely reactive to increasingly proactive optimization, whereby the best outcome is expected from good teamwork.

Evolution in Thermodynamics

Article

Mar 2017

Adrian Bejan

This review covers two aspects of “evolution” in thermodynamics. First, with the constructal law, thermodynamics is becoming the domain of physics that accounts for the phenomenon of evolution in nature, in general. Second, thermodynamics (and science generally) is the evolving add-on that empowers humans to predict the future and move more easily on earth, farther and longer in time. The part of nature that thermodynamics represents is this: nothing moves by itself unless it is driven by power, which is then destroyed (dissipated) during movement. Nothing evolves unless it flows and has the freedom to change its architecture such that it provides greater and easier access to the available space. Thermodynamics is the modern science of heat and work and their usefulness, which comes from converting the work (power) into movement (life) in flow architectures that evolve over time to facilitate movement.

Image noise and dose performance across a clinical population: Patient size adaptation as a metric of CT performance

Article

Feb 2017
MED PHYS

Purpose: Modern CT systems adjust x-ray flux accommodating for patient size to achieve certain image noise values. The effectiveness of this adaptation is an important aspect of CT performance and should ideally be characterized in the context of real patient cases. The objective of this study was to characterize CT performance with a new metric that includes image noise and radiation dose across a clinical patient population. Materials and methods: The study included 1526 examinations performed by three CT scanners (one GE Healthcare Discovery CT750HD, one GE Healthcare Lightspeed VCT, and one Siemens SOMATOM definition Flash) used for two routine clinical protocols (abdominopelvic with contrast and chest without contrast). An institutional monitoring system recorded all the data involved in the study. The dose-patient size and noise-patient size dependencies were linearized by considering a first order approximation of analytical models that describe the relationship between ionization dose and patient size, as well as image noise and patient size. A 3D-fit was performed for each protocol and each scanner with a planar function, and the Root Mean Square Error (RMSE) values were estimated as a metric of CT adaptability across the patient population. Results: The data show different scanner dependencies in terms of adaptability: the RMSE values for the three scanners are between 0.0385 HU(1/2) and 0.0215 HU(1/2) . Conclusions: A theoretical relationship between image noise, CTDIvol and patient size was determined based on real patient data. This relationship may be interpreted as a new metric related to the scanners' adaptability concerning image quality and radiation dose across a patient population. This method could be implemented to investigate the adaptability related to other image quality indexes and radiation dose in a clinical population. This article is protected by copyright. All rights reserved.

Data science, learning, and applications to biomedical and health sciences

Article

Sep 2017
ANN NY ACAD SCI

The last decade has seen an unprecedented increase in the volume and variety of electronic data related to research and development, health records, and patient self-tracking, collectively referred to as Big Data. Properly harnessed, Big Data can provide insights and drive discovery that will accelerate biomedical advances, improve patient outcomes, and reduce costs. However, the considerable potential of Big Data remains unrealized owing to obstacles including a limited ability to standardize and consolidate data and challenges in sharing data, among a variety of sources, providers, and facilities. Here, we discuss some of these challenges and potential solutions, as well as initiatives that are already underway to take advantage of Big Data.

Health risks from exposure to low levels of ionizing radiation: BEIR VII Phase 2

Book

Apr 2006

This book is the seventh in a series of titles from the National Research Council that addresses the effects of exposure to low dose LET (Linear Energy Transfer) ionizing radiation and human health. Updating information previously presented in the 1990 publication, Health Effects of Exposure to Low Levels of Ionizing Radiation: BEIR V, this book draws upon new data in both epidemiologic and experimental research. Ionizing radiation arises from both natural and man-made sources and at very high doses can produce damaging effects in human tissue that can be evident within days after exposure. However, it is the low-dose exposures that are the focus of this book. So-called "late" effects, such as cancer, are produced many years after the initial exposure. This book is among the first of its kind to include detailed risk estimates for cancer incidence in addition to cancer mortality. BEIR VII offers a full review of the available biological, biophysical, and epidemiological literature since the last BEIR report on the subject and develops the most up-to-date and comprehensive risk estimates for cancer and other health effects from exposure to low-level ionizing radiation. © 2006 by the National Academy of Sciences. All rights reserved.