ArticlePDF Available

Image Processing in Case-Based Reasoning

Authors:
  • Institute of Computer Vision and applied Computer Sciences IBaI
  • Technische Universität Kaiserslautern, University of Calgary, Germany

Abstract

This commentary summarizes case-based reasoning (CBR) research applied to image processing. It includes references to the systems, workshops, and landmark publications. Image processing includes a variety of image formats, from computer tomography images to remote sensing and spatial data sets.
Image Processing in Case-Based Reasoning: Commentary
P. Perner, A. Holt, and M. Richter
Image processing is a challenging field. The unique data (images) and the necessary
computation techniques require extraordinary case-representations, similarity measures and
CBR strategies to be utilised. An overview of the challenges of image processing and image
interpretation by introducing CBR strategies is given in Perner [Pern01].
Image interpretation is the process of mapping the numerical representation of an image into a
logical representation such as suitable for scene description. An image interpretation system
must be able to extract symbolic features from the pixels of an image (for example, irregular
structure inside the nodule, area of calcification, and sharp margin). This is a complex process
as the image passes through several general processing steps before the final symbolic
description is obtained. Image interpretation systems are becoming increasingly popular in
medical and industrial applications. The existing statistical and knowledge-based techniques
lack robustness, accuracy, and flexibility. New strategies are necessary that can adapt to
changing environmental conditions, user needs and process requirements. Introducing CBR
strategies into image interpretation systems can satisfy these requirements. CBR provides a
flexible and powerful method for controlling the image processing process in all phases of an
image interpretation system to derive information of the highest possible quality. Beyond this
CBR offers different learning capabilities, for all phases of an image interpretation system,
that satisfy different needs during the development process of an image interpretation system.
Perner [Per00] proposes a system that uses CBR to optimize image segmentation at the low
level unit according to changing image acquisition conditions and image quality. The
intermediate-level unit extracts the case representation used by the high-level unit employed
to dynamically adapt image interpretation. The system works on different case representations
such as a graph-based representation for the cases of the high-level image description and the
raw image matrix for the low-level image representation. Therefore the system uses different
CBR strategies for the reasoning and learning part one is based on structural similarity and the
other is based on the digital image distance.
Grimnes and Aamodt [GrA96] present a system that integrates CBR into a task-oriented
model-based system for the interpretation of abdominal CT images. A case-based reasoner
working on a segment case-base contains the individual image segments. These cases with
labels are considered indexes for another case-based reasoner working on an organ
interpretation case-base. There system is based on a propose-critique-modify learning cycle.
Jarmulak [Jar98] presents a system for ultra-sonic B-scans that are one-dimensional signals.
He presents a tree-based retrieval strategy. Micarelli et al. [MNS00] applies CBR to scene
recognition. They calculated image properties from images and stored them into a case base.
They used the Wavelet transform because it is scale-independent, but this limits their
similarity measure to consider only object rotation. The application of CBR to image
segmentation for CT images from the brain is described in [Per99]. Different learning
strategies in a hierarchy of structural cases are presented in Perner [Per98] [Per03]. Learning
case representation and improving the system performance by controlling the similarity
measure is described in [PPM02]. There is also a need for mining raw information into more
general cases [PeJ04]. Making object recognition more robust against model variation is
described in [PeB04].
The application of CBR image interpretation to health monitoring and biotechnology is
described in [PGPFE03].
A new challenging application field are geographical information systems. This kind of
application requires special spatial problem solving techniques and spatial similarity
measures. The first application of CBR to geographic information systems was reported by
Holt and Benwell [HoB99]. Their approach consists of combining case-based reasoning with
geographical information systems to form a hybrid system to solve spatial problems in soil
classification. Carswell et al. [CWDB02] described in their paper a spatial similarity measure
for comparing the location objects in different images.
Several systems have been developed which apply CBR for image retrieval and interpretation
at the symbolic level. Usually the images are not processed rather the symbolic terms are
user-specified. There is a system by Haddad et al. [HaA97].for the detection of coronary heart
disease. Another system has been developed by Macura and Macura [MaM95] for retrieval of
radiological images. Jaulent et al [JBZD98] apply the system for the diagnosis of breast
cancer in histopatology.
Literature
A completely different application of CBR to image processing has been described by Ficet-
Cauchard et al. [FPR99].They apply CBR for the development of the image processing steps
of formerly unknown image processing problems by using past experiences and plan
adaption.
Finally, in Perner [Per02] is built a bridge between the work in CBR and the one in
dissimilarity classification which became recently important in pattern recognition.
Most of the work described here has been presented at the ICCBR and ECCBR. There has
also been established a working group for CBR in image processing under the umbrella of
IAPR Technical committee TC 17 Machine Learning and Data Mining (www.ibai-
research.de/html/TC17) that has a major conference called Intern. Conference on Machine
Learning and Data Mining MLDM.
[Jar98] Jarmulak, J. (1998). Case-based classification of ultrasonic B-Scans: Case-base
organisation and case retrieval. In B. Smyth & P. Cunningham (Eds.) Advances in Case-
Based Reasoning (pp. 100-111). Berlin: Springer Verlag.
[GrA96] Grimnes, M. & Aamodt, A.(1996). A two layer case-based reasoning architecture for
medical image understanding, In I. Smith & B. Faltings (Eds.) Advances in Case-Based
Reasoning (pp. 164-178). Berlin: Springer Verlag.
[MNS00] Micarelli, A. Neri, A., & Sansonetti, G. (2000). A case-based approach to image
recognition, In E. Blanzieri & L. Portinale (Eds.) Advances in Case-Based Reasoning (pp.
443-454). Berlin: Springer Verlag.
[FPR99] Ficet-Cauchard, V., Porquet, C., & Revenu, M. (1999). CBR for the reuse of image
processing knowledge: A recursive retrieval/adaption strategy. In K.-D. Althoff, R.
Bergmann, & L.K. Branting (Eds.) Case-Based Reasoning Research and Development (pp.
438-453). Berlin: Springer.
[CWDB02] Carswell, James D., Wilson, David C., and Bertolotto, Michela. (2002). Digital
Image Similarity for Geo-Spatial Knowledge Management. In proceedings of 6th European
Conference on Case-Based Reasoning, Springer Verlag Lecture Notes in Artificial
Intelligence (ECCBR2002), Aberdeen, Scotland, September 2002
[HoB99] Holt A. and G.L. Benwell G.L., Applying Case-Based Reasoning Techniques in
GIS, Journal of Geograpical Information Science, 13(1):9-25, 1999
[HaA97] M. Haddad, K.-P. Adlassnig, G. Porenta, Feasibility analysis of a case-based
reasoning system for automated detection of coronary heart disease from myocardial
scintigrams, Artificial Intelligence in Medicine 9 (1997), 61-78.
[MaM95] R. Macura and K. Macura, MacRad: Radiology Image Resource with a Case-Based
Retrieval System, In: M. Veloso and A. Aamodt (eds.), Case-Based Reasoning: Research and
Development, Springer 1995, p. 43-45.
[JBZD98] Jaulent MC, Le Bozec C, Zapletal E, Degoulet P. Case based diagnosis in
histopathology of breast tumours. Medinfo. 1998;9 Pt 1:544-8.
[Per03] P. Perner, Incremental Learning of Retrieval Knowledge in a Case-Based Reasoning
System, In: K.D. Ashley and D.G. Bridge (Eds.), Case-Based Reasoning Research and
Development, lnai 2689, Springer Verlag, Berlin, 2003, p. 422-436.
[Per02] P. Perner, Are case-based reasoning and dissimilarity-based classification two sides of
the same coin? Journal Engineering Applications of Artificial Intelligence, vol. 15/3, 2002,
pp. 205-216.
[PPM02] P. Perner,H. Perner, and B. Müller, Similarity Guided Learning of the Case
Description and Improvement of the System Performance in an Image Classification System,
In: S. Craw and A.Preece (Eds.), Advances in Case-Based Reasoning, ECCBR2002, Springer
Verlag, lnai 2416, 2002, p. 604-612.
[Per00] P. Perner. CBR Ultra Sonic Image Interpretation. In: E. Blanzieri and Lugio Portinale
(Eds.), Advances in Case-based Reasoning, Springer Verlag 2000, lnai 1898, 2000, p. 479-
481
[Per99] P. Perner, An Architeture for a CBR Image Segmentation System, Journal on
Engineering Application in Artificial Intelligence, Engineering Applications of Artificial
Intelligence, vol. 12 (6), 1999, p. 749-759
[Per98] P. Perner, Different Learning Strategies in a Case-Based Reasoning System for Image
Interpretation, Advances in Case-Based Reasoning, B. Smith and P.
Cunningham (Eds.), LNAI 1488, Springer Verlag 1998, S. 251-261
[Per01] P. Perner, Why Case-Based Reasoning is Attractive for Image Interpretation, D. Aha
and I. Watson (Eds.), Case-Bases Reasoning Research and Developments, Springer Verlag
2001, lnai 2080, p. 27-44
[PGPFE03] P. Perner, Th. Günther, H. Perner, G. Fiss, R. Ernst, Health Monitoring by an
Image Interpretation System - A System for Airborne Fungi Identification, In: Petra Perner,
Rüdiger Brause, Hermann-Georg Holzhütter (Eds.), Medical Data Analysis, Springer Verlag,
lncs 2868, 2003, p. 64-77
[PeJ04] P. Perner and S. Jähnichen, Case Acquisition and Case Mining for Case-Based Object
Recognition, In: P. Funk and P.A. González Calero, Advances in Case-Based Reasoning,
Springer Verlag 2004, lnai 3155, 616-629.
[PeB04] P. Perner and A. Bühring, Case-Based Object Recognition, In: P. Funk and P.A.
González Calero, Advances in Case-Based Reasoning, Springer Verlag 2004, lnai 3155, 375-
388.
... She proposes using CBR on different levels, mostly focusing on the case representation of images and the similarity measure between cases. A series of works follows this line of research [Perner et al., 2005;Wilson and O'Sullivan, 2008;Perner, 2017]. In a recent survey Perner [2017] reviews applications of CBR in parameter selection, image interpretation, incremental prototype-based classification, novelty detection, and 1-D signal representation. ...
Preprint
Full-text available
The case difference heuristic (CDH) approach is a knowledge-light method for learning case adaptation knowledge from the case base of a case-based reasoning system. Given a pair of cases, the CDH approach attributes the difference in their solutions to the difference in the problems they solve, and generates adaptation rules to adjust solutions accordingly when a retrieved case and new query have similar problem differences. As an alternative to learning adaptation rules, several researchers have applied neural networks to learn to predict solution differences from problem differences. Previous work on such approaches has assumed that the feature set describing problems is predefined. This paper investigates a two-phase process combining deep learning for feature extraction and neural network based adaptation learning from extracted features. Its performance is demonstrated in a regression task on an image data: predicting age given the image of a face. Results show that the neural network can successfully learn adaptation knowledge. The CBR system achieves slightly lower performance overall than a baseline deep network re-gressor, but better performance than the baseline on novel queries.
... Case-Based Reasoning has been shown a successful problem-solving method in different applications were generalized knowledge is lacking. CBR has been used to interpret images [31,32], 1-D signals [31,33,34], and text cases [35]. It also has been used for meta-learning of the best parameter of image segmentation [36] and classification methods [37], so that the best processing and classification results can be achieved, although domain knowledge is lacking. ...
... CBR has been successfully applied to a wide range of problems. Among them are signal interpretation tasks [35], medical applications [36], and emerging applications such as geographic information systems, applications in biotechnology and topics in climate research (CBR commentaries) [37]. We are focussing here on hot real-world topics such as meta-learning for parameter selection, image&signal interpretation, prototype-based classification and novelty detection & handling. ...
Chapter
Full-text available
Case-based reasoning (CBR) solves problems using the already stored knowledge, and captures new knowledge, making it immediately available for solving the next problem. Therefore, CBR is seen as a method for problem solving and also as a method to capture new experience and make it immediately available for problem solving. The CBR paradigm has been originally introduced by the cognitive science community. The CBR community aims to develop computer models that follow this cognitive process. Up to now many successful computer systems have been established on the CBR paradigm for a wide range of real-world problems. We will review in this paper the CBR process and the main topics within the CBR work. Hereby we try bridging between the concepts developed within the CBR community and the statistics community. The CBR topics we describe are the similarity, memory organization, CBR learning, and case-base maintenance. Then we will review based on applications the open problems that need to be solved. The applications we are focusing on are meta-learning for parameter selection, image interpretation, incremental prototype-based classification and novelty detection and handling. Finally, we summarize our concept on CBR.
... In conjunction with image processing, CBR has seen a wide array of uses [17], once again with a great deal of research in the medical domain. Despite the wide usage and success of CBR in a variety of medical domains (e.g., [12] [10] [7]), most applications require hand-crafted features (e.g, [1] [13] [11]) generated by Subject Matter Experts (SMEs), a practice which does not scale to the Exascalelevel computation and learning that Deep Learning making possible [8]. ...
Preprint
Full-text available
Deep Learning methods have shown a rapid increase in popularity due to their state-of-the-art performance on many machine learning tasks. However, these methods often rely on extremely large datasets to accurately train the underlying machine learning models. For supervised learning techniques, the human effort required to acquire, encode, and label a sufficiently large dataset may add such a high cost that deploying the algorithms is infeasible. Even if a sufficient workforce exists to create such a dataset, the human annotators may differ in the quality, consistency, and level of granularity of their labels. Any impact this has on the overall dataset quality will ultimately impact the potential performance of an algorithm trained on it. This paper partially addresses this issue by providing an approach, called NOD-CC, for discovering novel object types in images using a combination of Convolutional Neural Networks (CNNs) and Case-Based Reasoning (CBR). The CNN component labels instances of known object types while deferring to the CBR component to identify and label novel, or poorly understood, object types. Thus, our approach leverages the state-of-the-art performance of CNNs in situations where sufficient high-quality training data exists, while minimizing its limitations in data-poor situations. We empirically evaluate our approach on a popular computer vision dataset and show significant improvements to objects classification performance when full knowledge of potential class labels is not known in advance.
... Case-based reasoning has been used for a variety of image processing and computer vision tasks [10]. One application area that has seen particular interest is medical CBR (e.g., [11][12][13]), primarily due to the prevalence of medical imagery in patient files and the need to retrieve similar images to aid in diagnosis. ...
Chapter
Full-text available
The development of Convolutional Neural Networks (CNNs) has resulted in significant improvements to object classification and detection in image data. One of their primary benefits is that they learn image features rather than relying on hand-crafted features, thereby reducing the amount of knowledge engineering that must be performed. However, another form of knowledge engineering bias exists in how objects are labelled in images, thereby limiting CNNs to classifying the set of object types that have been predefined by a domain expert. We describe a case-based method for detecting novel object types using a combination of an image’s raw pixel values and detectable parts. Our approach works alongside existing CNN architectures, thereby leveraging the state-of-the-art performance of CNNs, and is able to detect novel classes using limited training instances. We evaluate our approach using an existing object detection dataset and provide evidence of our approach’s ability to classify images even if the object in the image has not been previously encountered.
... Particularly relevant to this study is the fact that many CBR systems have been applied in medical decision making, such as CBR for medical knowledge-based systems [28], CBR in the health sciences [29], CBR for the prognosis and diagnosis of chronic diseases [30], case-based medical diagnosis, development, and experimentation [31], and a distributed CBR tool for medical prognosis [32]. Recently, CBR has been applied in problem-solving complex domains, including planning [33], law [34], e-learning [35], knowledge management [36], image processing [37], and recommender systems [38]. ...
Article
Full-text available
One of the significant issues in a smart city is maintaining a healthy environment. To improve the environment, huge amounts of data are gathered, manipulated, analyzed, and utilized, and these data might include noise, uncertainty, or unexpected mistreatment of the data. In some datasets, the class imbalance problem skews the learning performance of the classification algorithms. In this paper, we propose a case-based reasoning method that combines the use of crowd knowledge from open source data and collective knowledge. This method mitigates the class imbalance issues resulting from datasets, which diagnose wellness levels in patients suffering from stress or depression. We investigate effective ways to mitigate class imbalance issues in which the datasets have a higher proportion of one class over another. The results of this proposed hybrid reasoning method, using a combination of crowd knowledge extracted from open source data (i.e., a Google search, or other publicly accessible source) and collective knowledge (i.e., case-based reasoning), were that it performs better than other traditional methods (e.g., SMO, BayesNet, IBk, Logistic, C4.5, and crowd reasoning). We also demonstrate that the use of open source and big data improves the classification performance when used in addition to conventional classification algorithms.
... CBR has been successfully applied to a wide range of problems. Among them are signal [33] and image interpretation tasks [40], medical applications [15], and emerg- ...
Conference Paper
Full-text available
Over the years, image mining and knowledge discovery gained importance to solving problems. They are used in developing systems for automatic signal analysis and interpretation. The issues of model building and adaption, allowing an automatic system to adjust to the changing environments and moving objects, became increasingly important. One method of achieving adaptation in model building and model learning is Case-Based Reasoning (CBR). Case-Based Reasoning can be seen as a reasoning method as well as an incremental learning and knowledge acquisition method. In this paper we provide an overview of the CBR process and its main features: similarity, memory organization, CBR learning, and case-base maintenance. Then we review, based on applications, what has been achieved so far. The applications we are focusing on are meta-learning for parameter selection, image interpretation, incremental prototype-based classification, novelty detection and handling, and 1-D signal interpretation represented by a 0_1 sequence. Finally, we will summarize the overall concept of CBR usage for model development and learning.
Chapter
Old documents suffer from the passage of time: the paper is becoming more yellow, the ink is fading and the handling of these documents can still cause them to be damaged (e.g. stains may appear). The online edition of such documents can have profit of a cleaning step, the aim of which is to improve its readability. Some image filtering systems exist that require proper parameters to perform a cleaning. This article presents {\textsc {Georges}}, a CBR system designed to predict these parameters, for images of texts written in French, with several variants based on approximation, interpolation and extrapolation, based respectively of similarity, betweenness, and analogical proportion relations. The images are characterized by a dirtiness index, with the assumption that two images with similar dirtiness indexes would require similar sets of parameters for being cleaned. This index is based on the detection of the occurrences of a frequent French word on the pages and of averaging these occurrences. Human and automatic evaluations show that the proposed approach (with its variants) provides high-quality results.Keywordscase-based reasoningapproximationinterpolationextrapolationimage processingimage cleaning
Chapter
This chapter describes a few problems and methods combining artificial intelligence, pattern recognition, computer vision and learning. The intersection between these domains is growing and gaining importance, as illustrated in this chapter with a few examples. The first one deals with knowledge guided image understanding and structural recognition of shapes and objects in images. The second example deals with code supervision, which allows designing specific applications by exploiting existing algorithms in image processing, focusing on the formulation of processing objectives. Finally, the third example shows how different theoretical frameworks and methods for learning can be associated with the constraints inherent to the domain of robotics.
Chapter
Deep Learning methods have shown a rapid increase in popularity due to their state-of-the-art performance on many machine learning tasks. However, these methods often rely on extremely large datasets to accurately train the underlying machine learning models. For supervised learning techniques, the human effort required to acquire, encode, and label a sufficiently large dataset may add such a high cost that deploying the algorithms is infeasible. Even if a sufficient workforce exists to create such a dataset, the human annotators may differ in the quality, consistency, and level of granularity of their labels. Any impact this has on the overall dataset quality will ultimately impact the potential performance of an algorithm trained on it. This paper partially addresses this issue by providing an approach, called NOD-CC, for discovering novel object types in images using a combination of Convolutional Neural Networks (CNNs) and Case-Based Reasoning (CBR). The CNN component labels instances of known object types while deferring to the CBR component to identify and label novel, or poorly understood, object types. Thus, our approach leverages the state-of-the-art performance of CNNs in situations where sufficient high-quality training data exists, while minimizing its limitations in data-poor situations. We empirically evaluate our approach on a popular computer vision dataset and show significant improvements to object classification performance when full knowledge of potential class labels is not known in advance.
Article
Full-text available
This paper outlines an approach in applying artificial intelligence techniques to solve spatial problems. The approach consists of combining casebased reasoning with geographical information systems to form a hybrid system to solve spatial problems. In this paper, the hybrid is applied to the problem of soil classification which is solved by searching the case base for other spatial cases similar to the problem case. The knowledge from the retrieved case is used to formulate an answer to the problem case. The paper compares the results from this new approach with a traditional method of soil classification.
Conference Paper
Full-text available
The existing image interpretation systems lack robustness and accuracy. They cannot adapt to changing environmental conditions or to new objects. The application of machine learning to image interpretation is the next logical step. Our proposed approach aims at the development of dedicated machine learning techniques at all levels of image interpretation in a systematic fashion. In this paper we propose a system which uses Case-Based Reasoning (CBR) to optimize image segmentation at the low level according to changing image acquisition conditions and image quality. The intermediate-level unit extracts the case representation used by the high-level unit for further processing. At the high level, CBR is employed to dynamically adapt image interpretation.
Conference Paper
Full-text available
The development of image interpretation systems is concerned with tricky problems such as a limited number of observations, environmental influence, and noise. Recent systems lack robustness, accuracy, and flexibility. The introduction of case-based reasoning (CBR) strategies can help to overcome these drawbacks. The special type of information (i.e., images) and the problems mentioned above provide special requirements for CBR strategies. In this paper we review what has been achieved so far and research topics concerned with case-based image interpretation. We introduce a new approach for an image interpretation system and review its components.
Article
Brings emphasis to the plausible concept of case-based reasoning being integrated with spatial information systems, and the adaptation of artificial intelligence techniques to improve the analytical strength in spatial information systems. This adaptation of artificial intelligence techniques may include examples of expert systems, fuzzy logic, hybrid connection systems and neural networks all integrated with spatial information systems. The unique process of case-based reasoning is described and research into the possible integration of case-based reasoning and spatial information systems is outlined. The benefits of a case-based reasoning spatial information hybrid system are discussed.
Conference Paper
. The development of an Image Processing (IP) application is a complex activity, which can be greatly alleviated by user-friendly graphical programming environments. Our major objective is to help IP experts reuse parts of their applications. A first work towards knowledge reuse has been to propose a suitable representation of the strategies of IP experts by means of IP plans (trees of tasks, methods and tools). This paper describes the CBR module of our interactive system for the development of IP plans. After a brief presentation of the overall architecture of the system and its other modules, we explain the distinction between an IP case and an IP plan, and give the selection criteria and functions that are used for similarity calculation. The core of the CBR module is a search/adaptation algorithm, whose main steps are detailed: retrieval of suitable cases, recursive adaptation of the selected one and memorization of new cases. The system’s implementation is presently completed; its functioning is described in a session showing the kind of assistance provided by the CBR module during the development of a new IP application.
Article
Case-based reasoning (CBR) is used when generalized knowledge is lacking. The method works on a set of cases formerly processed and stored in the case base. A new case is interpreted based on its similarity to cases in the case base. The closest case with its associated result is selected and presented as output of the system. Recently, dissimilarity-based classification (DSC) has been introduced due to the curse of dimensionality of feature spaces and the problem arising when trying to make image features explicitly. The approach classifies samples based on their dissimilarity value to all training samples. In this paper we are reviewing the basic properties of these two approaches. We show the similarity of dissimilarity-based classification to case-based reasoning. Finally, we conclude that dissimilarity-based classification is a variant of case-based reasoning and that most of the open problems in dissimilarity-based classification are research topics of case-based reasoning.
Article
Image Segmentation is a crucial step if extracting information from a digital image. It is not easy to set up the segmentation parameter so that it fits best over the entire set of images, which should be segmented. In the paper, we propose a novel architecture for image segmentation method based on CBR, which can adapt to changing image qualities and environmental conditions. We describe the whole architecture, the methods used for the various components of the systems and show how it performs on medical images.
Article
A tool and a methodology for data mining in picture-archiving systems are presented. It is intended to discover the relevant knowledge for picture analysis and diagnosis from the data base of image descriptions. Knowledge-engineering methods are used to obtain a list of attributes for symbolic image descriptions. An expert describes images according to this list and stores descriptions in the data base. Digital-image processing can be applied to improve imaging of specific image features, or to get expert-independent feature evaluation. Decision-tree induction is used to learn the expert knowledge, presented in the form of image descriptions in the data base. A constructed decision tree presents effective models of decision-making, which can be learned to support image classification by the expert. A tool for data mining and image processing is presented and its application to image mining is shown on the task of Hep-2 cell-image classification. However, the tool and the methodology are generic and can be used for other image-mining tasks. We applied the developed methodology of data mining in other medical tasks, such as in lung-nodule diagnosis in X-ray images, lymph-node diagnosis in MRI and investigation of breast MRI.
Conference Paper
Case-base maintenance typically involves the addition, removal or revision of cases, but can also include changes to the retrieval knowledge. In this paper, we consider the learning of the retrieval knowledge (organization) as well as the prototypes and the cases as case-based maintenance. We address this problem based on cases that have a structural case representation. Two approaches for organizing the case base are proposed. Both are based on approximate graph subsumption.
Conference Paper
We have compiled a case-based retrieval system for radiology, MacRad, that is structured around descriptors for radiology image findings. Our goal is to provide a feature-coded image resource that allows the user to formulate image content-based queries when searching for reference images. MacRad is implemented as a relational database with an image archive. Each image in the library is indexed according to its radiologic content. We structured an index for coding image content as a hierarchical image description index using the relational format. The hierarchical index of radiologic findings allows multilevel query formulation that depends upon the user's level of experience. The system uses rules to control the search direction within the case library and generate lists of diagnostic hypotheses for decision support. Rules are embedded within the database structure. At present, the case library consists of 300 cases and 3,000 images that present intracranial masses on skull X-rays, CTs, MRIs, and angiograms.