Tagged: mammography

Training neural networks for computer-aided diagnosis: experience in the intelligence community

Neural networks are often used in computer-aided diagnosis (CAD) systems for detecting clinically significant objects. They have also been applied in the AI community to cue image analysts (IAs) for assisted target recognition and wide-area searching. Given the similarity between the applications in the two communities, there are a number of common issues that must be considered when training these neural networks. Two such issues are: (1) exploiting information at multiple scales (e.g. context and detail structure), and (2) dealing with uncertainty (e.g. errors in truth data). We address these two issues, transferring architectures and training algorithms originally developed for assisting IAs in search applications, to improve CAD for mammography. These include hierarchical pyramid neural net (HPNN) architectures that automatically learn and integrate multi-resolution features for improving microcalcification and mass detection in CAD systems. These networks are trained using an uncertain object position (UOP) error function for the supervised learning of image searching/detection tasks when the position of the objects to be found is uncertain or ill-defined. The results show that the HPNN architecture trained using the UOP error function reduces the false-positive rate of a mammographic CAD system by 30%-50% without any significant loss in sensitivity. We conclude that the transfer of assisted target recognition technology from the AI community to the medical community can significantly impact the clinical utility of CAD systems.

Application of Information Theory to Improve Computer-Aided Diagnosis

Mammographic Computer-Aided Diagnosis (CAD) systems are an approach for low-cost double reading. Though results to date have been promising, current systems often suffer from unacceptably high false positive rates. Improved methods are needed for optimally setting the system parameters, particularly in the case of statistical models that are common elements of most CAD systems. In this research project we developed a framework for building hierarchical pattern recognizers for CAD based on information theoretic criteria, e.g., the minimum description length (MDL). As part of this framework, we developed a hierarchical image probability (HIP) model. HIP models are well-suited to information theoretic methods since they are generative. We developed architecture search algorithms based on information theory, and applied these to mammographic CAD. The resulting mass detection algorithm, for example, reduced the false positive rate of a CAD system by 30% with no loss of sensitivity. We showed that the criteria reliably correlate with performance on new data. The framework allows many other applications not possible with most pattern recognition algorithms, including rejection of novel examples that can’t be reliably classified, synthesis of artificial images to investigate the structure learned by the model, and compression, which is as good as JPEG.

Mammographic mass detection with a hierarchical image probability (HIP) model

We formulate a model for probability distributions on image spaces. We show that any distribution of images can be factored exactly into conditional distributions of feature vectors at one resolution (pyramid level) conditioned on the image information at lower resolutions. We would like to factor this over positions in the pyramid levels to make it tractable, but such factoring may miss long-range dependencies. To fix this, we introduce hidden class labels at each pixel in the pyramid. The result is a hierarchical mixture of conditional probabilities, similar to a hidden Markov model on a tree. The model parameters can be found with maximum likelihood estimation using the EM algorithm. We have obtained encouraging preliminary results on the problems of detecting masses in mammograms.

The role of feature selection in building pattern recognizers for computer-aided diagnosis

In this paper we explore the use of feature selection techniques to improve the generalization performance of pattern recognizers for computer-aided diagnosis. We apply a modified version of the sequential forward floating selection (SFFS) of Pudil et al. to the problem of selecting an optimal feature subset for mass detection in digitized mammograms. The complete feature set consists of multi-scale tangential and radial gradients in the mammogram region of interest. We train a simple multi-layer perceptron (MLP) using the SFFS algorithm and compare its performance, using a jackknife procedure, to an MLP trained on the complete feature set (35 features). Results indicate that a variable number of features is chosen in each of the jackknife sets (12 +/- 4) and the test performance, Az, using the chosen feature subset is no better than the performance using the entire feature set. These results may be attributed to the fact that the feature set is noisy and the data set used for training/testing is small. We next modify the feature selection technique by using the results of the jackknife to compute the frequency at which different features are selected. We construct a classifier by choosing the top N features, selected most frequently, which maximize performance on the training data. We find that by adding this `hand-tuning’ component to the feature selection process, we can reduce the feature set from 35 to 8 features and at the same time have a statistically significant increase in generalization performance (p < 0.015).

Learning contextual relationships in mammograms using a hierarchical pyramid neural network

This paper describes a pattern recognition architecture, which we term hierarchical pyramid/neural network (HPNN), that learns to exploit image structure at multiple resolutions for detecting clinically significant features in digital/digitized mammograms. The HPNN architecture consists of a hierarchy of neural networks, each network receiving feature inputs at a given scale as well as features constructed by networks lower in the hierarchy. Networks are trained using a novel error function for the supervised learning of image search/detection tasks when the position of the objects to be found is uncertain or ill defined. We have evaluated the HPNN’s ability to eliminate false positive (FP) regions of interest generated by the University of Chicago’s (UofC) Computer-aided diagnosis (CAD) systems for microcalcification and mass detection. Results show that the HPNN architecture, trained using the uncertain object position (UOP) error function, reduces the FP rate of a mammographic CAD system by approximately 50% without significant loss in sensitivity. Investigation into the types of FPs that the HPNN eliminates suggests that the pattern recognizer is automatically learning and exploiting contextual information. Clinical utility is demonstrated through the evaluation of an integrated system in a clinical reader study. We conclude that the HPNN architecture learns contextual relationships between features at multiple scales and integrates these features for detecting microcalcifications and breast masses.