Baiyu Chen

Duke University, Durham, North Carolina, United States

Are you Baiyu Chen?

Claim your profile

Publications (18)16.36 Total impact

  • [Show abstract] [Hide abstract]
    ABSTRACT: For nonlinear iterative image reconstructions (IR), the computed tomography (CT) noise and resolution properties can depend on the specific imaging conditions, such as lesion contrast and image noise level. Therefore, it is imperative to develop a reliable method to measure the noise and resolution properties under clinically relevant conditions. This study aimed to develop a robust methodology to measure the three-dimensional CT noise and resolution properties under such conditions and to provide guidelines to achieve desirable levels of accuracy and precision.
    Medical physics. 07/2014; 41(7):071909.
  • [Show abstract] [Hide abstract]
    ABSTRACT: Purpose: Volume quantifications of lung nodules with multidetector computed tomography (CT) images provide useful information for monitoring nodule developments. The accuracy and precision of the volume quantification, however, can be impacted by imaging and reconstruction parameters. This study aimed to investigate the impact of iterative reconstruction algorithms on the accuracy and precision of volume quantification with dose and slice thickness as additional variables.Methods: Repeated CT images were acquired from an anthropomorphic chest phantom with synthetic nodules (9.5 and 4.8 mm) at six dose levels, and reconstructed with three reconstruction algorithms [filtered backprojection (FBP), adaptive statistical iterative reconstruction (ASiR), and model based iterative reconstruction (MBIR)] into three slice thicknesses. The nodule volumes were measured with two clinical software (A: Lung VCAR, B: iNtuition), and analyzed for accuracy and precision.Results: Precision was found to be generally comparable between FBP and iterative reconstruction with no statistically significant difference noted for different dose levels, slice thickness, and segmentation software. Accuracy was found to be more variable. For large nodules, the accuracy was significantly different between ASiR and FBP for all slice thicknesses with both software, and significantly different between MBIR and FBP for 0.625 mm slice thickness with Software A and for all slice thicknesses with Software B. For small nodules, the accuracy was more similar between FBP and iterative reconstruction, with the exception of ASIR vs FBP at 1.25 mm with Software A and MBIR vs FBP at 0.625 mm with Software A.Conclusions: The systematic difference between the accuracy of FBP and iterative reconstructions highlights the importance of extending current segmentation software to accommodate the image characteristics of iterative reconstructions. In addition, a calibration process may help reduce the dependency of accuracy on reconstruction algorithms, such that volumes quantified from scans of different reconstruction algorithms can be compared. The little difference found between the precision of FBP and iterative reconstructions could be a result of both iterative reconstruction's diminished noise reduction at the edge of the nodules as well as the loss of resolution at high noise levels with iterative reconstruction. The findings do not rule out potential advantage of IR that might be evident in a study that uses a larger number of nodules or repeated scans.
    Medical Physics 11/2013; 40(11):111902. · 2.91 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: OBJECTIVE. The objective of this study was to evaluate the feasibility of using iterative reconstructions in hepatic CT to improve the precision of Hounsfield unit quantification, which is the degree to which repeated measurements under unchanged conditions provide consistent results. MATERIALS AND METHODS. An anthropomorphic liver phantom with iodinated lesions designed to simulate the enhancement of hypervascular tumors during the late hepatic arterial phase was imaged, and images were reconstructed with both filtered back projection (FBP) and iterative reconstructions, such as adaptive statistical iterative reconstruction (ASIR) and model-based iterative reconstruction (MBIR). This protocol was further expanded into various dose levels, tube voltages, and slice thicknesses to investigate the effect of iterative reconstructions under all these conditions. The iodine concentrations of the lesions were quantified, with their precision calculated in terms of repeatability coefficient. RESULTS. ASIR reduced image noise by approximately 35%, and improved the quantitative precision by approximately 5%, compared with FBP. MBIR reduced noise by more than 65% and improved the precision by approximately 25% compared with the routine protocol. MBIR consistently showed better precision across a thinner slice thickness, lower tube voltage, and larger patient, achieving the target precision level at a dose lower (≥ 40%) than that of FBP. CONCLUSION. ASIR blended with 50% of FBP indicated a moderate gain in quantitative precision compared with FBP but could achieve more with a higher percentage. A higher gain was achieved by MBIR. These findings may be used to reduce the dose required for reliable quantification and may further serve as a basis for protocol optimization in terms of iodine quantification.
    American Journal of Roentgenology 05/2013; 200(5):W475-82. · 2.90 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: PURPOSE To characterize the bias and variation in reader measurement of phantom nodule volumes in CT imagery from multiple scanners. METHOD AND MATERIALS The QIBA 1C project is a study of reader measurements of volume and longest in-slice diameter on 6 synthetic nodules placed in an anthropomorphic phantom. The phantom was imaged on 6 CT scanners: a Siemens Sensation 64, a Toshiba Aquilion 64, a Philips Brilliance 64, 2 Philips Brilliance 16s, and a GE VCT 64. The imaging protocol had 2 arms, one arm based on ACRIN 6678; the other arm determined by device-independent measures of resolution and noise in order to constrain image quality variation. In a single session, each of 7 experienced radiologists segmented each lesion, from which the two size measurements were derived. The 6 synthetic nodules appear in 2 shape types, spherical (nodules 1, 2, 3) and spiculated (nodules 4, 5, 6). There are 3 sizes: 5 mm (nodules 1, 4), 10 mm (nodules 2, 5), and 20 mm (nodules 3, 6). Sizes 5, 10 and 20 mm are radii of equivalent volume spheres. True volume was determined from mass measurements and manufacturer density data. Readers and nodule-order were randomized to reduce bias. The percent relative bias in the volume (Vol), 100*[Vol – nominal Vol]/Vol , was tested for effects of scanner, protocol arm, nodule size, nodule shape and reader. Using a t-test, we evaluated the primary hypotheses that the device effects and the protocol effect in the relative bias are each no greater than 15%. RESULTS An equivalence t-test applied to each nodule separately finds scanner equivalence (to a 15% tolerance) on nodules 2, 3, and 6. Equivalence on the spiculated 10-mm nodule (#5) is at a slightly higher tolerance, 15.5%. On 5 mm nodules, the equivalence limit is greater than 20%. For pooled date for nodules, the scanners are equivalent within 15% tolerance (p<0.0001). The two protocols are equivalent. CONCLUSION Relative bias in pooling the 6 nodules is within a 15% tolerance. On individual nodules, scanner equivalence is found only for the larger synthetic lesions (10 mm and 20 mm). Equivalence of the two protocols supports ACRIN 6678. CLINICAL RELEVANCE/APPLICATION The study demonstrates in larger lesions (>=10mm diameter) bias and variance can be approximately 15% or less across lesion types, scanners and protocols; it confirms QIBA CT lesion size guidance.
    Radiological Society of North America 2012 Scientific Assembly and Annual Meeting; 11/2012
  • [Show abstract] [Hide abstract]
    ABSTRACT: The current estimations of risk associated with medical imaging procedures rely on assessing the organ dose via direct measurements or simulation. The dose to each organ is assumed to be homogeneous. To take into account the differences in radiation sensitivities, the mean organ doses are weighted by a corresponding tissue-weighting coefficients provided by ICRP to calculate the effective dose, which has been used as a surrogate of radiation risk. However, those coefficients were derived under the assumption of a homogeneous dose distribution within each organ. That assumption is significantly violated in most medical-imaging procedures. In helical chest CT, for example, superficial organs (e.g. breasts) demonstrate a heterogeneous dose distribution, whereas organs on the peripheries of the irradiation field (e.g. liver) might possess a discontinuous dose profile. Projection radiography and mammography involve an even higher level of organ dose heterogeneity spanning up to two orders of magnitude. As such, mean dose or point measured dose values do not reflect the maximum energy deposited per unit volume of the organ. In this paper, the magnitude of the dose heterogeneity in both CT and projection X-ray imaging was reported, using Monte Carlo methods. The lung dose demonstrated factors of 1.7 and 2.2 difference between the mean and maximum dose for chest CT and radiography, respectively. The corresponding values for the liver were 1.9 and 3.5. For mammography and breast tomosynthesis, the difference between mean glandular dose and maximum glandular dose was 3.1. Risk models based on the mean dose were found to provide a reasonable reflection of cancer risk. However, for leukaemia, they were found to significantly under-represent the risk when the organ dose distribution is heterogeneous. A systematic study is needed to develop a risk model for heterogeneous dose distributions.
    Radiation Protection Dosimetry 10/2012; · 0.91 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The aim of this study was to compare the image quality of abdominal computed tomography scans in an anthropomorphic phantom acquired at different radiation dose levels where each raw data set is reconstructed with both a standard convolution filtered back projection (FBP) and a full model-based iterative reconstruction (MBIR) algorithm. An anthropomorphic phantom in 3 sizes was used with a custom-built liver insert simulating late hepatic arterial enhancement and containing hypervascular liver lesions of various sizes. Imaging was performed on a 64-section multidetector-row computed tomography scanner (Discovery CT750 HD; GE Healthcare, Waukesha, WI) at 3 different tube voltages for each patient size and 5 incrementally decreasing tube current-time products for each tube voltage. Quantitative analysis consisted of contrast-to-noise ratio calculations and image noise assessment. Qualitative image analysis was performed by 3 independent radiologists rating subjective image quality and lesion conspicuity. Contrast-to-noise ratio was significantly higher and mean image noise was significantly lower on MBIR images than on FBP images in all patient sizes, at all tube voltage settings, and all radiation dose levels (P < 0.05). Overall image quality and lesion conspicuity were rated higher for MBIR images compared with FBP images at all radiation dose levels. Image quality and lesion conspicuity on 25% to 50% dose MBIR images were rated equal to full-dose FBP images. This phantom study suggests that depending on patient size, clinically acceptable image quality of the liver in the late hepatic arterial phase can be achieved with MBIR at approximately 50% lower radiation dose compared with FBP.
    Investigative radiology 06/2012; 47(8):468-74. · 4.85 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Current estimation of lung nodule size typically relies on uni- or bi-dimensional techniques. While new three-dimensional volume estimation techniques using MDCT have improved size estimation of nodules with irregular shapes, the effect of acquisition and reconstruction parameters on accuracy (bias) and precision (variance) of the new techniques has not been fully investigated. To characterize the volume estimation performance dependence on these parameters, an anthropomorphic chest phantom containing synthetic nodules was scanned and reconstructed with protocols across various acquisition and reconstruction parameters. Nodule volumes were estimated by a clinical lung analysis software package, LungVCAR. Precision and accuracy of the volume assessment were calculated across the nodules and compared between protocols via a generalized estimating equation analysis. Results showed that the precision and accuracy of nodule volume quantifications were dependent on slice thickness, with different dependences for different nodule characteristics. Other parameters including kVp, pitch, and reconstruction kernel had lower impact. Determining these technique dependences enables better volume quantification via protocol optimization and highlights the importance of consistent imaging parameters in sequential examinations.
    Physics in Medicine and Biology 03/2012; 57(5):1335-48. · 2.70 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Mammography is currently the most widely accepted tool for detection and diagnosis of breast cancer. However, the sensitivity of mammography is reduced in women with dense breast tissue due to tissue overlap, which may obscure lesions. Digital breast tomosynthesis with contrast enhancement reduces tissue overlap and provides additional functional information about lesions (i.e. morphology and kinetics), which in turn may improve lesion characterization. The performance of such techniques is highly dependent on the structural composition of the breast, which varies significantly across patients. Therefore, optimization of breast imaging systems should be done with respect to this patient versatility. Furthermore, imaging techniques that employ contrast require the inclusion of a temporally varying breast composition with respect to the contrast agent kinetics to enable the optimization of the system. To these ends, we have developed a dynamic 4D anthropomorphic breast phantom, which can be used for optimizing a breast imaging system by incorporating material characteristics. The presented dynamic phantom is based on two recently developed anthropomorphic breast phantoms, which can be representative of a whole population through their randomized anatomical feature generation and various compression levels. The 4D dynamic phantom is incorporated with the kinetics of contrast agent uptake in different tissues and can realistically model benign and malignant lesions. To demonstrate the utility of the proposed dynamic phantom, contrast-enhanced digital mammography and breast tomosynthesis were simulated where a ray-tracing algorithm emulated the projections, a filtered back projection algorithm was used for reconstruction, and dual-energy and temporal subtractions were performed and compared.
    Proc SPIE 02/2012;
  • [Show abstract] [Hide abstract]
    ABSTRACT: The quantification of lung nodule volume based on CT images provides valuable information for disease diagnosis and staging. However, the precision of the quantification is protocol, system, and technique dependent and needs to be evaluated for each specific case. To efficiently investigate the quantitative precision and find an optimal operating point, it is important to develop a predictive model based on basic system parameters. In this study, a Fourier-based metric, the estimability index (e') was proposed as such a predictor, and validated across a variety of imaging conditions. To first obtain the ground truth of quantitative precision, an anthropomorphic chest phantom with synthetic spherical nodules were imaged on a 64 slice CT scanner across a range of protocols (five exposure levels and two reconstruction algorithms). The volumes of nodules were quantified from the images using clinical software, with the precision of the quantification calculated for each protocol. To predict the precision, e' was calculated for each protocol based on several Fourier-based figures of merit, which modeled the characteristic of the quantitation task and the imaging condition (resolution, noise, etc.) of a particular protocol. Results showed a strong correlation (R2=0.92) between the measured and predicted precision across all protocols, indicating e' as an effective predictor of the quantitative precision. This study provides a useful framework for quantification-oriented optimization of CT protocols.
    Proc SPIE 02/2012;
  • [Show abstract] [Hide abstract]
    ABSTRACT: The increasing availability of iterative reconstruction (IR) algorithms on clinical scanners is creating a demand for effectively and efficiently evaluating imaging performance and potential dose reduction. In this study, the location- and task-specific evaluation was performed using detectability index (d') by combining a task function, the task transfer function (TTF), and the noise power spectrum (NPS). Task function modeled a wide variety detection tasks in terms of shape and contrast. The TTF and NPS were measured from a physical phantom as a function of contrast and dose levels. Measured d' values were compared between three IRs (IRIS, SAFIRE3 and SAFIRE5) and conventional filtered back-projection (FBP) at various dose levels, showing an equivalent performance of IR at lower dose levels. AUC further calculated from d' showed that compared to FBP, SAFIRE5 may reduce dose by up to 50-60%; SAFIRE3 and IRIS by up to 20-30%. This study provides an initial framework for the localized and task-specific evaluation of IRs in CT and a guideline for the identification of optimal operating dose point with iterative reconstructions.
    Proc SPIE 02/2012;
  • [Show abstract] [Hide abstract]
    ABSTRACT: Digital breast tomosynthesis (DBT) is a novel x-ray imaging technique that provides 3D structural information of the breast. In contrast to 2D mammography, DBT minimizes tissue overlap potentially improving cancer detection and reducing number of unnecessary recalls. The addition of a contrast agent to DBT and mammography for lesion enhancement has the benefit of providing functional information of a lesion, as lesion contrast uptake and washout patterns may help differentiate between benign and malignant tumors. This study used a task-based method to determine the optimal imaging approach by analyzing six imaging paradigms in terms of their ability to resolve iodine at a given dose: contrast enhanced mammography and tomosynthesis, temporal subtraction mammography and tomosynthesis, and dual energy subtraction mammography and tomosynthesis. Imaging performance was characterized using a detectability index d', derived from the system task transfer function (TTF), an imaging task, iodine contrast, and the noise power spectrum (NPS). The task modeled a 5 mm lesion containing iodine concentrations between 2.1 mg/cc and 8.6 mg/cc. TTF was obtained using an edge phantom, and the NPS was measured over several exposure levels, energies, and target-filter combinations. Using a structured CIRS phantom, d' was generated as a function of dose and iodine concentration. In general, higher dose gave higher d', but for the lowest iodine concentration and lowest dose, dual energy subtraction tomosynthesis and temporal subtraction tomosynthesis demonstrated the highest performance.
    Proc SPIE 02/2012; 41(6):7-.
  • [Show abstract] [Hide abstract]
    ABSTRACT: Optimization studies for x-ray-based breast imaging systems using computer simulation can greatly benefit from a phantom capable of modeling varying anatomical variability across different patients. This study aimed to develop a three-dimensional phantom model with realistic and randomizable anatomical features. A voxelized breast model was developed consisting of an outer layer of skin and subcutaneous fat, a mixture of glandular and adipose, stochastically generated ductal trees, masses, and microcalcifications. Randomized realization of the breast morphology provided a range of patient models. Compression models were included to represent the breast under various compression levels along different orientations. A Monte Carlo (MC) simulation code was adapted to simulate x-ray based imaging systems for the breast phantom. Simulated projections of the phantom at different angles were generated and reconstructed with iterative methods, simulating mammography, breast tomosynthesis, and computed tomography (CT) systems. Phantom dose maps were further generated for dosimetric evaluation. Region of interest comparisons of simulated and real mammograms showed strong similarities in terms of appearance and features. Noise-power spectra of simulated mammographic images demonstrated that the phantom provided target properties for anatomical backgrounds. Reconstructed tomosynthesis and CT images and dose maps provided corresponding data from a single breast enabling optimization studies. Dosimetry result provided insight into the dose distribution difference between modalities and compression levels. The anthropomorphic breast phantom, combined with the MC simulation platform, generated a realistic model for a breast imaging system. The developed platform is expected to provide a versatile and powerful framework for optimizing volumetric breast imaging systems.
    Academic radiology 03/2011; 18(5):536-46. · 2.09 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: In hepatic CT imaging, the lesion enhancement after the injection of contrast media is of quantitative interest. However, the precision of this quantitative measurement may be dependent on the imaging techniques such as dose and reconstruction algorithm. To determine the impact of different techniques, we scanned an iodinated liver phantom with acquisition protocols of different dose levels, and reconstructed images with different algorithms (FBP and MBIR) and slice thicknesses. The contrast of lesions was quantified from the images, and its precision was calculated for each protocol separately. Results showed that precision was improved by increasing dose, increasing slice thickness, and using MBIR reconstruction. When using MBIR instead of FBP, the same precision can be achieved at 50% less dose. To our knowledge, this is the first investigation of the quantification precision in hepatic CT imaging using iterative reconstructions.
    Proc SPIE 03/2011;
  • Baiyu Chen, Daniele Marin, Ehsan Samei
    [Show abstract] [Hide abstract]
    ABSTRACT: An iodinated liver phantom is needed for liver CT related studies, such as the quantification of lesion contrast. Prior studies simulated iodinated hepatic lesions with tubes of iodine solution, which involved complications associated with the setup, differences from actual lesion morphology, and susceptibility to iodine sediments. To develop a dedicated liver phantom with anthropomorphic structures and solid lesions, we designed a phantom with iodinated liver inserts and lesions of different sizes and contrasts. The concentration of iodine in liver parenchyma was determined according to the HU measured from clinical images. The concentrations in high and low contrast lesions were selected so as to provide challenging but reasonable detection tasks. The application of the liver phantom was initially validated at different doses and reconstruction settings.
    Proc SPIE 03/2011;
  • [Show abstract] [Hide abstract]
    ABSTRACT: Current lung nodule size assessment methods typically rely on one-dimensional estimation of lesions. While new 3D volume assessment techniques using MSCT scan data have enabled improved estimation of lesion size, the effect of acquisition and reconstruction parameters on accuracy and precision of such estimation has not been adequately investigated. To characterize such dependencies, we scanned an anthropomorphic thoracic phantom containing synthetic nodules with different protocols, including various acquisition and reconstruction parameters. We also scanned the phantom repeatedly with the same protocol to investigate repeatability. The nodule's volume was estimated by a clinical lung analysis software package, LungVCAR. Accuracy (bias) and precision (variance) of the volume assessment were calculated across the nodules and compared between protocols via Generalized Estimating Equation analysis. Results suggest a strong dependence of accuracy and precision on dose level but little dependence on reconstruction thickness, thus providing possible guidelines for protocol optimization for quantitative tasks.
    Proc SPIE 03/2010;
  • [Show abstract] [Hide abstract]
    ABSTRACT: Conventional mammography is largely limited by superimposed anatomy which is alleviated by breast tomosynthesis and CT. Limited acquisition in tomosynthesis can result in significant out of plane artifacts while large angular acquisition span in CT can limit the imaging coverage of the chest wall near the breast. We propose a new breast imaging modality, wide-angle breast tomosynthesis (WBT), aimed to provide a practical compromise between 3D sampling and chest-wall coverage. This study compares lesion detection between conventional digital breast tomosynthesis, WBT, and breast CT (44°, 99°, and 198° total angle range, respectively) under equal patient dose conditions. A Monte Carlo (MC) code based on the Penelope package modeled a virtual flat-panel breast tomosynthesis system. The modalities were simulated at four breast compression levels. Glandular dose to the breast was estimated and the radiation flux was subsequently adjusted to achieve a constant mean glandular dose level of 1.5 mGy, independent of the breast thickness and acquisition geometry. Reconstructed volumes were generated using iterative reconstruction methods. Lesion detectability was estimated using contrast-to-noise-ratio. Results showed improved detection with increased angular span and compression. Evaluations also showed improved performance of WBT over DBT at lower compression levels, therefore highlighting potential for reduced breast compression when using a larger acquisition angle.
    Proc SPIE 03/2010;
  • [Show abstract] [Hide abstract]
    ABSTRACT: The current estimations of risk associated with medical imaging procedures rely on assessing the organ dose via direct measurements or simulation. Each organ dose is assumed to be homogeneous, a representative sample or mean of which is weighted by a corresponding tissue weighting factor provided by ICRP publication 103. The weighted values are summed to provide Effective Dose (ED), the most-widely accepted surrogate for population radiation risk. For individual risk estimation, one may employ Effective Risk (ER), which further incorporates gender- and age-specific risk factors. However, both the tissue-weighting factors (as used by ED) and the risk factors (as used by ER) were derived (mostly from the atomic bomb survivor data) under the assumption of a homogeneous dose distribution within each organ. That assumption is significantly violated in most medical imaging procedures. In chest CT, for example, superficial organs (eg, breasts) demonstrate a heterogeneous distribution while organs on the peripheries of the irradiation field (eg, liver) possess a nearly discontinuous dose profile. Projection radiography and mammography involve an even wider range of organ dose heterogeneity spanning up to two orders of magnitude. As such, mean dose or point measured dose values do not reflect the maximum energy deposited per unit volume of the organ, and therefore, effective dose or effective risk, as commonly computed, can misrepresent irradiation risk. In this paper, we report the magnitude of the dose heterogeneity in both CT and projection x-ray imaging, provide an assessment of its impact on irradiation risk, and explore an alternative model-based approach for risk estimation for imaging techniques involving heterogeneous organ dose distributions.
    Proc SPIE 03/2010;
  • [Show abstract] [Hide abstract]
    ABSTRACT: This study aimed to extend Fourier-based imaging metrics for the modeling of quantitative imaging performance. Breast tomosynthesis was used as a platform for investigating acquisition and processing parameters (e.g., acquisition angle and dose) that can significantly affect 3D signal and noise, and consequently quantitative imaging performance. The detectability index was computed using the modulation transfer function and noise-power spectrum combined with a Fourier description of imaging task. Three imaging tasks were considered: detection, area estimation (in coronal slice), and volume estimation of a 4 mm diameter spherical target. Task functions for size estimation were generated by using measured performance of the maximum-likelihood estimator as training data. The detectability index computed with the size estimation tasks correlated well with precision measurements for area and volume estimation over a fairly broad range of imaging conditions and provided a meaningful figure of merit for quantitative imaging performance. Furthermore, results highlighted that optimal breast tomosynthesis acquisition parameters depend significantly on imaging task. Mass detection was optimal at an acquisition angle of 85° while area and volume estimation for the same mass were optimal at ~100° and 125° acquisition angles, respectively. These findings provide key initial validation that the Fourier-based detectability index extended to estimation tasks can represent a meaningful metric and predictor of quantitative imaging performance.
    Proc SPIE 03/2010;