Computer-aided Diagnosis Applications Research Articles

Statistical 4D graphs for multi-organ abdominal segmentation from multiphase CT.

The interpretation of medical images benefits from anatomical and physiological priors to optimize computer-aided diagnosis applications. Diagnosis also relies on the comprehensive analysis of multiple organs and quantitative measures of soft tissue. An automated method optimized for medical image data is presented for the simultaneous segmentation of four abdominal organs from 4D CT data using graph cuts. Contrast-enhanced CT scans were obtained at two phases: non-contrast and portal venous. Intra-patient data were spatially normalized by non-linear registration. Then 4D convolution using population training information of contrast-enhanced liver, spleen and kidneys was applied to multiphase data to initialize the 4D graph and adapt to patient-specific data. CT enhancement information and constraints on shape, from Parzen windows, and location, from a probabilistic atlas, were input into a new formulation of a 4D graph. Comparative results demonstrate the effects of appearance, enhancement, shape and location on organ segmentation. All four abdominal organs were segmented robustly and accurately with volume overlaps over 93.6% and average surface distances below 1.1mm.

Contralateral lesions detected by preoperative MRI in patients with recently diagnosed breast cancer: Application of MR CAD in differentiation of benign and malignant lesions

To retrospectively investigate the added value of kinetic features measured by computer-aided diagnosis (CAD) for differentiating benign and malignant contralateral breast lesions detected by preoperative MRI in breast cancer patients. This study was approved by our institutional review board, and the requirement for informed consent was waived. Fifty-two breast MR images and their CAD kinetic features were obtained for 52 consecutive breast cancer patients with contralateral breast lesions detected by preoperative MRI and confirmed by excision (23 cancers and 29 benign lesions). Three experienced radiologists independently reviewed the MR images without CAD information and assessed probabilities of malignancy. Four weeks later, these probabilities were reanalyzed using stored CAD data. Diagnostic performances and detection rates of delayed washout components were compared between interpretations without and with CAD for each reader. Use of MR CAD increased detection of washout component by 2.4- to 3.7-fold than visual assessment for enhancing contralateral lesions, which increased sensitivity (91% vs. 87% in reader 1; 96% vs. 74% in reader 2; 91% vs. 70% in reader 3) and decreased specificity, but statistical significance was only found for decreased specificity in one reader (52% vs. 28%, P = 0.039), and overall performance (areas under ROC curves 0.672 vs. 0.616 in reader 1; 0.624 vs. 0.603 in reader 2; 0.706 vs. 0.590 in reader 3) remained unimproved. Addition of MR CAD increased sensitivity and decreased specificity than radiologist's assessment alone for differentiating benign and malignant contralateral lesions in breast cancer patients and overall performance remained unimproved.

Modeling and Analysis of Shape with Applications in Computer-Aided Diagnosis of Breast Cancer

Malignant tumors due to breast cancer and masses due to benign disease appear in mammograms with different shape characteristics: the former usually have rough, spiculated, or microlobulated contours, whereas the latter commonly have smooth, round, oval, or macrolobulated contours. Features that characterize shape roughness and complexity can assist in distinguishing between malignant tumors and benign masses. In spite of the established importance of shape factors in the analysis of breast tumors and masses, difficulties exist in obtaining accurate and artifact-free boundaries of the related regions from mammograms. Whereas manually drawn contours could contain artifacts related to hand tremor and are subject to intra-observer and inter-observer variations, automatically detected contours could contain noise and inaccuracies due to limitations or errors in the procedures for the detection and segmentation of the related regions. Modeling procedures are desired to eliminate the artifa ts in a given contour, while preserving the important and significant details present in the contour. This book presents polygonal modeling methods that reduce the influence of noise and artifacts while preserving the diagnostically relevant features, in particular the spicules and lobulations in the given contours. In order to facilitate the derivation of features that capture the characteristics of shape roughness of contours of breast masses, methods to derive a signature based on the turning angle function obtained from the polygonal model are described. Methods are also described to derive an index of spiculation, an index characterizing the presence of convex regions, an index characterizing the presence of concave regions, an index of convexity, and a measure of fractal dimension from the turning angle function. Results of testing the methods with a set of 111 contours of 65 benign masses and 46 malignant tumors are presented and discussed. It is shown that shape modeling and a alysis can lead to classification accuracy in discriminating between benign masses and malignant tumors, in terms of the area under the receiver operating characteristic curve, of up to 0.94. The methods have applications in modeling and analysis of the shape of various types of regions or objects in images, computer vision, computer graphics, and analysis of biomedical images, with particular significance in computer-aided diagnosis of breast cancer. Table of Contents: Analysis of Shape / Polygonal Modeling of Contours / Shape Factors for Pattern Classification / Classification of Breast Masses

Semantic Sparse Representation of Disease Patterns

Semantic Sparse Representation of Disease PatternsSparse data representation is discussed in a context of useful fundamentals led to semantic content description and extraction of information. Disease patterns as semantic information extracted from medical images were underlined because of discussed application of computer-aided diagnosis. Compressive sensing rules were adjusted to the requirements of diagnostic pattern recognition. Proposed methodology of sparse disease patterns considers accuracy of sparse representation to estimate target content for detailed analysis. Semantics of sparse representation were modeled by morphological content analysis. Subtle or hidden components were extracted and displayed to increase information completeness. Usefulness of sparsity was verified for computer-aided diagnosis of stroke based on brain CT scans. Implemented method was based on selective and sparse representation of subtle hypodensity to improve diagnosis. Visual expression of disease signatures was fixed to radiologist requirements, domain knowledge and experimental analysis issues. Diagnosis assistance suitability was proven by experimental subjective rating and automatic recognition.

Extraction of media and plaque boundaries in intravascular ultrasound images by level sets and min/max flow

: Estimation of the plaque area in intravascular ultrasound images after extraction of the media and plaque–lumen interfaces is an important application of computer-aided diagnosis in medical imaging. This paper presents a novel system for fully automatic and fast calculation of plaque quantity by capturing the surrounding ring called media. The system utilizes an algorithm that consists of an enhanced technique for noise removal and a method of detecting different iso levels by sinking the image gradually under zero level. Moreover, an important novelty with this technique is the simultaneous extraction of media and lumen–plaque interfaces at satisfactory levels. There are no higher dimensional surfaces and evolution of contours, stopping at high image gradients. Thus, the system runs really fast with curvature velocity only and has no complexity. Experiments also show that this shape-recovering curvature term not only removes the noisy behaviour of ultrasound images but also strengthens very weak boundaries and even completes the missing walls of the media. In addition, the lumen–plaque interface can be detected simultaneously. For validation, a new and very useful algorithm is developed for labelling of intravascular ultrasound images, taken from video sequences of 15 patients, and a comparison-based verification is done between manual contours by experts and the contours extracted by our system.

Computer-aided diagnosis for contrast-enhanced ultrasound in the liver

Computer-aided diagnosis (CAD) has become one of the major research subjects in medical imaging and diagnostic radiology. The basic concept of CAD is to provide computer output as a second opinion to assist radiologists' image interpretations by improving the accuracy and consistency of radiologic diagnosis and also by reducing the image-reading time. To date, research on CAD in ultrasound (US)-based diagnosis has been carried out mostly for breast lesions and has been limited in the fields of gastroenterology and hepatology, with most studies being conducted using B-mode US images. Two CAD schemes with contrast-enhanced US (CEUS) that are used in classifying focal liver lesions (FLLs) as liver metastasis, hemangioma, or three histologically differentiated types of hepatocellular carcinoma (HCC) are introduced in this article: one is based on physicians' subjective pattern classifications (subjective analysis) and the other is a computerized scheme for classification of FLLs (quantitative analysis). Classification accuracies for FLLs for each CAD scheme were 84.8% and 88.5% for metastasis, 93.3% and 93.8% for hemangioma, and 98.6% and 86.9% for all HCCs, respectively. In addition, the classification accuracies for histologic differentiation of HCCs were 65.2% and 79.2% for well-differentiated HCCs, 41.7% and 50.0% for moderately differentiated HCCs, and 80.0% and 77.8% for poorly differentiated HCCs, respectively. There are a number of issues concerning the clinical application of CAD for CEUS, however, it is likely that CAD for CEUS of the liver will make great progress in the future.

On Concise 3-D Simple Point Characterizations: A Marching Cubes Paradigm

The centerlines of tubular structures are useful for medical image visualization and computer-aided diagnosis applications. They can be effectively extracted by using a thinning algorithm that erodes an object layer by layer until only a skeleton is left. An object point is "simple" and can be safely deleted only if the resultant image is topologically equivalent to the original. Numerous characterizations of 3-D simple points based on digital topology already exist. However, little work has been done in the context of marching cubes (MC). This paper reviews several concise 3-D simple point characterizations in a MC paradigm. By using the Euler characteristic and a few newly observed properties in the context of connectivity-consistent MC, we present concise and more self-explanatory proofs. We also present an efficient method for computing the Euler characteristic locally for MC surfaces. Performance evaluations on different implementations are conducted on synthetic data and multidetector computed tomography examination of virtual colonoscopy and angiography.

Computer‐Aided Diagnosis in Thoracic Computed Tomography

SummaryCurrent computed tomography (CT) technology allows for isotropic, sub‐millimetre resolution acquisition of the thorax in a few seconds. These thin‐slice chest scans have become indispensable in thoracic radiology, but have also increased the time and effort required from radiologists for reporting. Industry has focused on the development of computer‐aided diagnosis (CAD) tools to facilitate the interpretation of thoracic CT data. In this paper I discuss the three most ‘senior’ CAD applications for chest CT: nodule detection, nodule volumetry and quantification of emphysema. Are these applications ready for widespread application?

Multiple-Instance Learning Algorithms for Computer-Aided Detection

Many computer-aided diagnosis (CAD) problems can be best modelled as a multiple-instance learning (MIL) problem with unbalanced data, i.e., the training data typically consists of a few positive bags, and a very large number of negative instances. Existing MIL algorithms are much too computationally expensive for these datasets. We describe CH, a framework for learning a Convex Hull representation of multiple instances that is significantly faster than existing MIL algorithms. Our CH framework applies to any standard hyperplane-based learning algorithm, and for some algorithms, is guaranteed to find the global optimal solution. Experimental studies on two different CAD applications further demonstrate that the proposed algorithm significantly improves diagnostic accuracy when compared to both MIL and traditional classifiers. Although not designed for standard MIL problems (which have both positive and negative bags and relatively balanced datasets), comparisons against other MIL methods on benchmark problems also indicate that the proposed method is competitive with the state-of-the-art.

Advances in computer-aided diagnosis for breast cancer

Computer-aided diagnosis (CAD) is a technology used for the detection and characterization of cancer. Although CAD is not limited to a single type of cancer, a large number of CAD systems to date have been designed and used for breast cancer. The aim of this review is to discuss the current state of the CAD systems for breast-cancer diagnosis, their application as a second reader in clinical practice, and studies that have evaluated the effect of CAD on radiologists' performance. A large number of CAD applications are being developed for different imaging modalities. Owing to commercially available Food and Drug Administration (FDA) approved systems, the main clinical use of CAD to date is for screen-film mammography. Many studies have shown that CAD improves radiologists' performance. A large number of academic institutions have devoted a substantial research effort to developing CAD methods. CAD systems will play an increasingly important role in the clinic as a second reader. Clinical trials have shown that CAD can improve the accuracy of breast-cancer detection. Preclinical studies have demonstrated the potential of CAD to improve the classification of malignant and benign lesions. An increased number of CAD systems are being developed for different breast-imaging modalities.

Computer simulation of mammographic imaging for applications in CAD

A method is described for the computer simulation of mammograms. Software has been developed to model the conversion of input X-ray quanta to visible light photons inside a phosphor screen. In this method, we separated the physics processes that convert X-ray quanta to visible light photons inside a fluorescent screen into five mathematical transformations. Standard FFT techniques were used. We extended Swank's one-dimensional fluorescent screen MTF model into a two-dimensional photon diffusion model by solving the 2D Boltzman diffusion equation. The simulation software divides the fluorescent screen into multiple thinner sub-layers and integrates the visible light output of every sub-layer over the whole screen into the final image. The simulation software was tested by two comparisons: modulation transfer function (MTF) and noise power spectrum. Both test results show that the method and the simulation software work correctly. For the modulation transfer function and the shape of the noise power spectra, the difference between the simulation and the experimental values of an actual fluorescent screen was within 8%. Qualitatively, the simulated and actual images of test objects looked similar. Using the developed method, we will be able to produce simulated mammograms that can be used to evaluate computer-aided diagnosis (CAD) methods and techniques.

Lung nodules and beyond: approaches, challenges and opportunities in thoracic CAD

The purpose of this manuscript is to give an overview of current activities in Computer-Aided Diagnosis (CAD) for thoracic applications. Because the large majority of applications of CAD relate to lung cancer, the detection and characterization problems in lung cancer CAD are described. Approaches to each of these problems are described. The current challenges that exist in each of the detection and characterization problems are described. Opportunities that exist in the near future are described including: efforts to create standardized databases by the Lung Image Database Consortium (LIDC) to facilitate research in lung CAD, research into other lung diseases besides lung cancer and the incorporation of other informatics sources into the computer-aided diagnosis approaches.

Application of the mutual information criterion for feature selection in computer-aided diagnosis.

The purpose of this study was to investigate an information theoretic approach to feature selection for computer-aided diagnosis (CAD). The approach is based on the mutual information (MI) concept. MI measures the general dependence of random variables without making any assumptions about the nature of their underlying relationships. Consequently, MI can potentially offer some advantages over feature selection techniques that focus only on the linear relationships of variables. This study was based on a database of statistical texture features extracted from perfusion lung scans. The ultimate goal was to select the optimal subset of features for the computer-aided diagnosis of acute pulmonary embolism (PE). Initially, the study addressed issues regarding the approximation of MI in a limited dataset as it is often the case in CAD applications. The MI selected features were compared to those features selected using stepwise linear discriminant analysis and genetic algorithms for the same PE database. Linear and nonlinear decision models were implemented to merge the selected features into a final diagnosis. Results showed that the MI is an effective feature selection criterion for nonlinear CAD models overcoming some of the well-known limitations and computational complexities of other popular feature selection techniques in the field.

Classifier design for computer-aided diagnosis: effects of finite sample size on the mean performance of classical and neural network classifiers.

Classifier design is one of the key steps in the development of computer-aided diagnosis (CAD) algorithms. A classifier is designed with case samples drawn from the patient population. Generally, the sample size available for classifier design is limited, which introduces variance and bias into the performance of the trained classifier, relative to that obtained with an infinite sample size. For CAD applications, a commonly used performance index for a classifier is the area, Az, under the receiver operating characteristic (ROC) curve. We have conducted a computer simulation study to investigate the dependence of the mean performance, in terms of Az, on design sample size for a linear discriminant and two nonlinear classifiers, the quadratic discriminant and the backpropagation neural network (ANN). The performances of the classifiers were compared for four types of class distributions that have specific properties: multivariate normal distributions with equal covariance matrices and unequal means, unequal covariance matrices and unequal means, and unequal covariance matrices and equal means, and a feature space where the two classes were uniformly distributed in disjoint checkerboard regions. We evaluated the performances of the classifiers in feature spaces of dimensionality ranging from 3 to 15, and design sample sizes from 20 to 800 per class. The dependence of the resubstitution and hold-out performance on design (training) sample size (Nt) was investigated. For multivariate normal class distributions with equal covariance matrices, the linear discriminant is the optimal classifier. It was found that its Az-versus-1/Nt curves can be closely approximated by linear dependences over the range of sample sizes studied. In the feature spaces with unequal covariance matrices where the quadratic discriminant is optimal, the linear discriminant is inferior to the quadratic discriminant or the ANN when the design sample size is large. However, when the design sample is small, a relatively simple classifier, such as the linear discriminant or an ANN with very few hidden nodes, may be preferred because performance bias increases with the complexity of the classifier. In the regime where the classifier performance is dominated by the 1/Nt term, the performance in the limit of infinite sample size can be estimated as the intercept (1/Nt= 0) of a linear regression of Az versus 1/Nt. The understanding of the performance of the classifiers under the constraint of a finite design sample size is expected to facilitate the selection of a proper classifier for a given classification task and the design of an efficient resampling scheme.

The Relative Accuracy of a Variety of Medical Diagnostic Programs

Acute abdominal pain is one of the most widely studied applications of computer-aided diagnosis. The usual approach is to apply Bayes' theorem with the assumption of conditional independence ("independence Bayes"). We compared various approaches to designing diagnostic programs for abdominal pain of suspected gynaecological origin. The methods range from statistical to knowledge-based. All programs were evaluated using a database of 1,270 cases collected retrospectively. Our results suggest that in this application no significant improvement in accuracy can be made by taking interactions into account, either by statistical or by knowledge-based means; independence Bayes is near-optimal. As far as accuracy is concerned, there appears to be little point in pursuing knowledge-based approaches. However, the "nearest neighbours" method using a new metric appears to be at least as accurate as independence Bayes. We argue that the nearest neighbours method is more suitable than independence Bayes for clinical use because of greater accountability.