2006

Using Computational Intelligence for Computer-Aided Diagnosis of Screen Film Mammograms. (book chapter)

SPIE MI 2006. An adaptive image segmentation process for the classification of lung biopsy images.

2005

CIT 2005. Emerging Technology in Teaching Computational Intelligence for Engineers and Computer Scientists via Distance Learning.

ASEE St. Lawrence Section Conference. Teaching Computational Intelligence for Engineers in the 21st Century via Distance Learning.

2004

SPIE MI 2004. Breast cancer classification improvements using a new kernel function with evolutionary-programming-configured Support Vector Machines.

SPIE MI 2004. New results in Computer Aided Diagnosis (CAD) of breast cancer using a recently developed SVM/GRNN Oracle hybrid.

2003

IEEE CSMC 2003. Breast Cancer Computer Aided Diagnosis (CAD) Using a Recently Developed SVM/GRNN Oracle Hybrid.

SPIE MI 2003. Improving the predictive value of mammography using a specialized evolutionary programming hybrid and fitness functions.

SPIE MI 2003. Application of Support Vector Machines to breast cancer screening using mammogram and clinical history data.

2002

ANNIE '02. Improving mammogram screening using a bank of support vector machines (SVMs).

IEEE CEC 2002. Performance tradeoff between evolutionary computation (EC)/adaptive boosting (AB) hybrid and support vector machine breast cancer classification paradigms.

2001

ISAS 2001. Performance analysis of evolutionary computation (EC)/adaptive boosting (AB) hybrids for breast cancer classification.

Master's Thesis. Boosting evolved artificial neural networks to improve breast cancer classification accuracy.

IEEE SMCia 2001. New results in breast cancer classification obtained from an evolutionary computation/adaptive boosting hybrid using mammogram and history data. Runner up for Best Paper Award.

IEEE CEC 2001. Application of evolutionary computation and neural network hybrids for breast cancer classification using mammogram and history data.

SPIE MI 2001. Application of adaptive boosting to EP-derived multi-layer feedforward neural networks (MLFNs) to improve benign/malignant breast cancer classification.

2000

SCI2000. Using evolutionary computation to develop neural network breast cancer benign/malignant classification models.

Paper Details

• Using Computational Intelligence for Computer-Aided Diagnosis of Screen Film Mammograms. (book chapter)

  • Land, Walker H., Jr., McKee, Daniel W., Anderson, Frances R., Masters, Timothy, Lo, Joseph Y., Embrechts, Mark and Heine John. “Using Computational Intelligence for Computer-Aided Diagnosis of Screen Film Mammograms.” (book chapter) In Suri, Jasjit S. and Rangayyan, Rangaraj M., eds. Recent Advances in Mammography, Breast Imaging, and Computer-aided Diagnosis of Breast Cancer. SPIE Press, 2006.

  • This chapter describes paradigms and hybrids applied to the Computer Aided Diagnosis (CAD) of breast cancer, using mammogram screen film data. The chapter is organized into the following sections:

    • Evolutionary Programming (EP) / Evolutionary Strategies (ES) family of neural networks (NNs), derived and evaluated using statistical cross validation techniques with results measured by ROC analysis.

    • EP/ Adaptive Boosting (AB) hybrid derived family of NNs, with CAD results.

    • EP approach which evolves not only NN parameters, but also the architecture, with CAD results.

    • EP derived Support Vector Machines (SVMs), with CAD results.

    • SVM/GRNN oracle hybrid formulation, including the development of the Probabilistic Neural network (PNN) and the Generalized Regression Neural Network (GRNN), with CAD performance results.

    • Partial Least Squares (PLS) and Kernel-PLS (K-PLS), with CAD performance results.

    • Modular Artificial Intelligence (AI) system design and the use of Knowledge Engineering (KE) and Knowledge Representation (KR), with CAD performance results.
       

• SPIE MI 2006. An adaptive image segmentation process for the classification of lung biopsy images.

  • McKee, D. W., Land, W. H., Jr., Zhukov, T., Song, D., and Qian, W. “An adaptive image segmentation process for the classification of lung biopsy images.” SPIE Medical Imaging 2006, San Diego, CA, 11-16 February, 2006.

  • Abstract - The purpose of this study was to develop a computer-based second opinion diagnostic tool that could read microscope images of lung tissue and classify the tissue sample as normal or cancerous. This problem can be broken down into three areas: segmentation, feature extraction and measurement, and classification. We introduce a kernel-based extension of fuzzy c-means to provide a coarse initial segmentation, with heuristically-based mechanisms to improve the accuracy of the segmentation. The segmented image is then processed to extract and quantify features. Finally, the measured features are used by a Support Vector Machine (SVM) to classify the tissue sample. The performance of this approach was tested using a database of 85 images collected at the Moffitt Cancer Center and Research Institute. These images represent a wide variety of normal lung tissue samples, as well as multiple types of lung cancer. When used with a subset of the data containing images from the normal and adenocarcinoma classes, we were able to correctly classify 78% of the images, with a ROC AZ of 0.758.
     

• CIT 2005. Emerging Technology in Teaching Computational Intelligence for Engineers and Computer Scientists via Distance Learning.

  • Land, W. H., Jr., and McKee, D. W. “Emerging Technology in Teaching Computational Intelligence for Engineers and Computer Scientists via Distance Learning.” SUNY Conference on Instructional Technology, CIT 2005, Binghamton, NY, 23-26 May, 2005.

  • Abstract - The presenter shows the practical benefits of teaching computational intelligence technology by distance learning. Specifically, engineers, computer scientists, medical personnel as well as some of the Liberal Arts disciplines will be using computers driven by intelligent software as a matter of course in the 21st century. This assertion is currently supported by the fact that computer aided diagnostic (CAD) intelligent software is currently being developed as a second opinion diagnostic aid for breast and lung cancer, in addition to being used for diagnosing heart abnormalities and prostate cancer. In addition, intelligent computer detection and classification software has been and is being developed for the detection of deadly bio terrorism nerve agents, such as organophosphates. The primary objective of this paper is to describe how distance learning is currently being employed to teach Support Vector Machine (SVM) concepts as a mechanism to develop these intelligent software packages. In addition a 25 minute movie titled 'Topics in Computational Intelligence' will be shown. It demonstrates the proficiency with which graduate and upper level undergraduate students' learn this CAD technology in the distance learning environment, by reporting on research projects in which they developed a CAD tool to diagnose breast cancer from a mammogram screen film data set using SVM technology. This film was previously shown twice: first as part of an invited paper on teaching Evolutionary Computational Concepts at the 2002 International World Congress for Computational Intelligence, and secondly as a paper for the 2005 St. Lawrence conference. No additional equipment beyond a screen and a connection to a laptop computer will be required to view the film. In this film, the students demonstrate their understanding of SVM theory, its application, development and testing as applied to the development of CAD diagnostic software packages.
     

• ASEE St. Lawrence Section Conference. Teaching Computational Intelligence for Engineers in the 21st Century via Distance Learning.

  • Land, W. H., Jr., and McKee, D. W. “Teaching Computational Intelligence for Engineers in the 21st Century via Distance Learning.” American Society for Engineering Education (ASEE) St. Lawrence Section Conference, Binghamton. NY, 8-9 April, 2005.

  • Abstract - Engineers, computer scientists, medical personnel as well as some of the Liberal Arts disciplines will be using computers driven by intelligent software as a matter of course in 2020. This assertion is currently supported by the fact that computer aided diagnostic (CAD) intelligent software is currently being developed as a second opinion diagnostic aids for breast and lung cancer in addition to being used for diagnosing heart abnormalities and prostate cancer. In addition, intelligent computer detection and classification software has been and is being developed for the detection of deadly bio terrorism nerve agents, such as organophosphates.

    The primary objective of this paper is to describe how distance learning is currently being employed to teach Support Vector Machine (SVM) concepts as a mechanism to develop these intelligent software packages. This presentation also contains a 25min movie titled "Topics in Computational Intelligence" that demonstrates the proficiency with which graduate and upper level undergraduate students' learn this CAD technology in the distance learning environment, by reporting on research projects in which they developed a CAD tool to diagnose breast cancer from a mammogram screen film data set using SVM technology. This film was first shown as part of an invited paper on teaching Evolutionary Computational Concepts at the 2002 International World Congress for Computational Intelligence. No additional equipment beyond a screen and a connection to a laptop computer will be required to view the film. In this film, the students demonstrate their understanding of SVM theory, its application, development and testing as applied to the development of CAD diagnostic software packages.
     

• SPIE MI 2004. Breast cancer classification improvements using a new kernel function with evolutionary-programming-configured Support Vector Machines.

  • Land, W. H., Jr., McKee, D. W., Anderson, F. R., and Lo, J. Y. "Breast cancer classification improvements using a new kernel function with evolutionary-programming-configured Support Vector Machines." SPIE Medical Imaging 2004, San Diego, CA, 14-19 February, 2004.

  • Abstract - Mammography is an effective tool for the early detection of breast cancer; however, most women referred for biopsy based on mammographic findings do not, in fact, have cancer. This study is part of an ongoing effort to reduce the number of benign cases referred for biopsy by developing tools to aid physicians in classifying suspicious lesions. Specifically, this study examines the use of an Evolutionary Programming (EP)-derived Support Vector Machine (SVM) with a modified radial basis function (RBF) kernel, and compares this with results using a normal Gaussian radial basis function kernel. Results demonstrate that the modified kernel can provide moderate performance improvements; however, due to its ability to create a more complex decision surface, this kernel can easily begin to memorize the training data resulting in a loss of generalization ability. Nonetheless, these methods could reduce the number of benign cases referred for biopsy by over half, while missing less than 5% of malignancies. Future work will focus on methods to improve the EP process to preserve SVMs which generalize well.

   

• SPIE MI 2004. New results in Computer Aided Diagnosis (CAD) of breast cancer using a recently developed SVM/GRNN Oracle hybrid.

  • Land, W. H., Jr., Wong, L, McKee, D., Masters, T., Anderson, F., Raturi, A. and Lo, J. Y. "New results in Computer Aided Diagnosis (CAD) of breast cancer using a recently developed SVM/GRNN Oracle hybrid." SPIE Medical Imaging 2004, San Diego, CA, 14-19 February, 2004.

  • Abstract - Breast cancer is second only to lung cancer as a tumor-related cause of death in women. Currently, the method of choice for the early detection of breast cancer is mammography. While sensitive to the detection of non palpable breast lesions, its positive predictive value (PPV) is low, resulting in biopsies that are only 15%-34% likely to reveal malignancy. This paper explores the use of a recently designed Support Vector Machine (SVM)/Generalized Regression Neural Network (GRNN) Oracle hybrid to classify breast lesions and evaluate the software’s performance as an interpretive aid to radiologists. The main objective of the research was to perform an independent analysis, using a new, integrated film screen mammogram data base of approximately 2500 cases from five separate institutions, to verify results obtained previously.  This study demonstrated the following:

    • The DE crossover constant has little, if any,  effect on measures of performance (MOP).

    • A specificity of approximately 5.6% is achieved at 100% sensitivity, which increases to approximately 36% at 95% sensitivity.

    • PPV increases from 51% to 56% as sensitivity is decreased from 100 to 95%, respectively.

• Breast Cancer Computer Aided Diagnosis (CAD) Using a Recently Developed SVM/GRNN Oracle Hybrid.

  • Land, W. H., Jr., Wong, L., McKee, D. W., Masters, T., and Anderson, F. R. "Breast Cancer Computer Aided Diagnosis (CAD) Using a Recently Developed SVM/GRNN Oracle Hybrid." 2003 IEEE International Conference on Systems, Man & Cybernetics, Washington, DC, 5-8 October, 2003.

  • Abstract - Carcinoma of the breast is second only to lung cancer as a tumor-related cause of death in women. For 2003, it has been reported that 211,300 new cases and 39,800 deaths will occur just in the US. It has been proposed, however, that the mortality from breast cancer could be decreased by up to 25% if all women in appropriate age groups were screened regularly. Currently, the method of choice for the early detection of breast cancer is mammography, due to its general widespread availability, low cost, speed, and non-invasiveness. At the same time, while mammography is sensitive to the detection of breast cancer, its positive predictive value (PPV) is low, resulting in costly, invasive biopsies that are only 15%-34% likely to reveal malignancy at histological examination. This paper explores the use of a newly designed Support Vector Machine (SVM) / Generalized Regression Neural Network (GRNN) Oracle hybrid and evaluates its performance as an interpretive aid to radiologists. The authors demonstrate that this hybrid has the potential to (1) improve both the specificity and PPV of screen film mammography at 95-100% sensitivity, and (2) consistently produce partial AZ values (defined as average specificity over the top 10% of the ROC curve) of greater than 50%, using a data set of ~2000 lesions from four different hospitals. As expected, initial experiments demonstrated that combining age and mass margin (AgeMM) provided the most accurate diagnostic performance. Secondly, the value of the crossover constant, CR = 0.6, provided the best AZ, while CR =0.8 resulted in the most accurate partial AZ, Specificity, and PPV at the lower sensitivities. Finally, the results of the GRNN oracle output were essentially the same those of the SVM suggesting that the SVMs, as anticipated, had optimized the diagnostic performance. Practically, this means that at 100% sensitivity (which means no cancers lesions are misdiagnosed) and using a crossover constant of 0.8, approximately 454 biopsies would be avoided using this SVM/GRNN oracle diagnostic aid when compared to the circumstance where all 1979 samples were biopsied. (Note that the distribution of benign/malignant samples were about 50/50 for this data set.)

   

• Improving the predictive value of mammography using a specialized evolutionary programming hybrid and fitness functions.

  • Land, W. H., McKee, D. W., Lo, J. Y., and Anderson, F. R. "Improving the predictive value of mammography using a specialized evolutionary programming hybrid and fitness functions." SPIE Medical Imaging 2003, San Diego, CA, 15-20 February, 2003.

  • Abstract - Mammography is an effective tool for the early detection of breast cancer; however, most women referred for biopsy based on mammographic findings do not, have cancer. This study is part of an ongoing effort to reduce the number of benign cases referred for biopsy by developing tools to aid physicians in classifying suspicious lesions. Specifically, this study examines the use of an Evolutionary Programming (EP)/Adaptive Boosting (AB) hybrid, specifically modified to focus on improving the performance of computer-assisted diagnostic (CAD) tools at high specificity levels (missing few or no cancers). An EP/AB hybrid developed by the authors and used in previous studies was modified with two new fitness functions: 1) a function which favored networks with high PPV values at thresholds corresponding to high sensitivities, and 2) a function which favored networks with the highest partial ROC Az (normalized area above 90% sensitivity). The modified hybrid with specialized fitness functions was evaluated using k-fold cross-validation against two real-word mammogram data sets. Results indicate that the number of benign cases referred for biopsy might be reduced by over a third, while missing no cancers. If sensitivity is allowed to decrease to 97% (missing 3% of the cancers), the number of spared biopsies could be raised to over half.
     

• Application of Support Vector Machines to breast cancer screening using mammogram and clinical history data.

  • Land, W. H., McKee, D. W., Velazquez, R., Wong, L., Lo, J. Y., and Anderson, F. R. "Application of Support Vector Machines to breast cancer screening using mammogram and clinical history data." SPIE Medical Imaging 2003, San Diego, CA, 15-20 February, 2003.

  • Abstract - The objectives of this paper are to discuss: (1) the development and testing of a new Evolutionary Programming (EP) method to optimally configure Support Vector Machine (SVM) parameters for facilitating the diagnosis of breast cancer; (2) evaluation of EP derived learning machines when the number of BI-RADS(TM) and clinical history discriminators are reduced from 16 to 7; (3) establishing system performance for several SVM kernels in addition to the EP/Adaptive Boosting (EP/AB) hybrid using the Digital Database for Screening Mammography, University of South Florida (DDSM USF) and Duke data sets; and (4) obtaining a preliminary evaluation of the measurement of SVM learning machine inter-institutional generalization capability using BI-RADS^TM data.  Measuring performance of the SVM designs and EP/AB hybrid against these objectives will provide quantitative evidence that the software packages described can generalize to larger patient data sets from different institutions.  Most iterative methods currently in use to optimize learning machine parameters are time consuming processes, which sometimes yield sub-optimal values resulting in performance degradation.  SVMs are new machine intelligence paradigms, which use the Structural Risk Minimization (SRM) concept to develop learning machines.  These learning machines can always be trained to provide global minima, given that the machine parameters are optimally computed.  In addition, several system performance studies are described which include EP derived SVM performance as a function of: (a) population and generation size as well as a method for generating initial populations and (b) iteratively derived versus EP derived learning machine parameters.  Finally, the authors describe a set of experiments providing preliminary evidence that both the EP/AB hybrid and SVM Computer Aided Diagnostic C++ software packages will work across a large population of patients, based on a data set of approximately 2,500 samples from five different institutions.
     

• Improving mammogram screening using a bank of support vector machines (SVMs).

  • Land, W. H., Jr., McKee, D. W., Lo, J. Y., & Anderson, F. R. "Improving mammogram screening using a bank of support vector machines (SVMs)." Artificial Neural Networks in Engineering (ANNIE ’02).  St. Louis, MO, 10-13 November, 2002.

  • Abstract - The focus of this study was to build and evaluate a new bank of SVM designs to address the problem of high false positives that currently results from mammogram screening,. The basis of the design is to partition the BIRADS™ variables into three separate categories based on our understanding of the discriminating information contained in the mammogram BIRADS^TM findings. That is, after ascertaining the presence of a suspicious finding on a mammogram that would be recommended for biopsy, the radiologist documents the BIRADS™ lesion descriptor values. This information, along with the clinical history, would be used as input to this new bank of SVMs an aid to the physician for improving the specificity and positive predictive value PPV of the benign/malignant diagnosis task. Comparing the new SVM mass classifier with the previously configured single SVM that used all data base inputs provided significant classification accuracy improvements for all performance measures. That is, overall Az improved by 11.6%, specificity and PPV improved by 110.6% and 31.6%, respectively, at 100% sensitivity (missing no cancers), while specificity and PPV improved by 54% and 35.9%,respectively, at 95% sensitivity (missing 5% of the cancers).
     

• Performance tradeoff between evolutionary computation (EC)/adaptive boosting (AB) hybrid and support vector machine breast cancer classification paradigms.

  • Land, W. H., Jr., Bryden, M., Lo, J. Y., McKee, D. W., & Anderson, F. R. "Performance tradeoff between evolutionary computation (EC)/adaptive boosting (AB) hybrid and support vector machine breast cancer classification paradigms."  IEEE Congress on Evolutionary Computation, Honolulu, HI, 12-17 May, 2002.

  • Abstract - This paper describes a breast cancer classification performance trade-off analysis using two computational intelligence paradigms. The first, an Evolutionary Programming (EP) / Adaptive Boosting (AB) based hybrid, intelligently combines the outputs from an iteratively “called” weak learning algorithm (one which performs at least slightly better than random guessing) in order to “boost” the performance of an EP-derived weak learner. The second paradigm is Support Vector Machines (SVMs). SVMs are new and radically different types of classifiers and learning machines that use a hypothesis space of linear functions in a high dimensional feature space. The most important advantage of a SVM, unlike neural networks, is that SVM training always finds a global minimum. Furthermore, the SVM has inherent ability to solve pattern classification without incorporating any problem-domain knowledge. In this study, the both the EP/AB hybrid and SVM were employed as pattern classifiers, operating on mammography data used for breast cancer detection. The main focus of the study was to construct and seek the best EP/AB hybrid and SVM configurations for optimum specificity and positive predictive value at very high sensitivities. Using a mammogram database of 500 biopsy proven samples, the best performing SVM, on average, was able to achieve (under statistical 5-fold cross-validation) a specificity of 45.0% and a positive predictive value (PPV) of 50.1% at 100% sensitivity. At 97% sensitivity, a specificity of 55.8% and a PPV of 55.2% were obtained. The best performing EP/AB hybrid obtained slightly lower, but comparable, results.
     

• Performance analysis of evolutionary computation (EC)/adaptive boosting (AB) hybrids for breast cancer classification.

  • Land, W. H., Jr., Masters, T., Lo, J. Y., McKee, D. W., & Anderson, F. R. “Performance analysis of evolutionary computation (EC)/adaptive boosting (AB) hybrids for breast cancer classification.”  6th International conference on Information Systems Analysis and Synthesis (ISAS2001), Orlando, FL, 22-25 July, 2001.

  • Abstract - A new neural network technology was developed to improve the diagnosis of breast cancer using mammogram findings. The paradigm, Adaptive Boosting (AB), focuses on finding weak learning algorithm(s) that initially need to provide slightly better than “random” performance (i.e., approximately 55%) when processing a mammogram training set. By successive development of additional architectures (using the mammogram training set), the adaptive boosting process improves performance of the basic Evolutionary Programming derived neural network architectures. The results of these several EP-derived hybrid architectures are then intelligently combined and tested using a similar validation mammogram data set. Because we were particularly interested in maximizing positive predictive value (PPV) and specificity at high sensitivity levels, the error-minimization selection method used by the EP component was replaced with a selection method favoring those solutions with the best predictive value. Preliminary results using the new fitness function were then compared with optimized results using the original fitness function. Preliminary results using the Duke University mammogram database of 500 biopsy proven samples show that this PPV-based hybrid was able to achieve (under statistical 5-fold cross validation), on average, PPV and specificity results comparable to the best results obtained using the error-minimization hybrid.
     

• Boosting evolved artificial neural networks to improve breast cancer classification accuracy.

  • McKee, D. W. Boosting evolved artificial neural networks to improve breast cancer classification accuracy. Master's Thesis, State University of New York, Binghamton, NY, 2001.

  • Abstract - Two parallel models were developed to predict malignancy based upon mammogram data. Both used evolutionary computation to develop multi-layer feedforward networks, which were later used as the basis for a new adaptive boosting algorithm. Because good performance at very high sensitivity levels (missing few cancers) is more clinically relevant than overall performance, a model was developed specifically to enhance positive predictive value at high sensitivities. The performance of this model, with and without boosting, was compared to the performance of a more traditional error-minimization model. These results of these models were also compared to the results of previous studies that used the same data.
    
    This study found that boosted evolved networks can improve the predictive value of mammography by as much as 49%, and that the evolutionary and boosting processes can be modified to specifically enhance that predictive value.
     

• New results in breast cancer classification obtained from an evolutionary computation/adaptive boosting hybrid using mammogram and history data.

  • Land, W. H., Masters, T., Lo, J. Y., McKee, D. W., & Anderson, F. R. “New results in breast cancer classification obtained from an evolutionary computation/adaptive boosting hybrid using mammogram and history data.” IEEE Mountain Workshop on Soft Computing in Industrial Applications, Blackburg, VA, 25-27 June, 2001.

  • Abstract - A new neural network technology was developed to improve the diagnosis of breast cancer using mammogram findings. The paradigm, Adaptive Boosting (AB), uses a markedly different theory in solving Computational Intelligence (CI) problems. AB, a new machine learning paradigm, focuses on finding weak learning algorithm(s) that initially need to provide slightly better than “random” performance (i.e., approximately 55%) when processing a mammogram training set. By successive development of additional architectures (using the mammogram training set), the adaptive boosting process improves performance of the basic Evolutionary Programming derived neural network architectures. The results of these several EP-derived hybrid architectures are then intelligently combined and tested using a similar validation mammogram data set. Optimization, focused on improving specificity and positive predictive value at very high sensitivities, with an analysis of the performance of the hybrid would be most meaningful. Using the DUKE mammogram database of 500 biopsy proven samples, this hybrid, on average, was able to achieve (under statistical 5-fold cross-validation) a specificity of 48.3% and a positive predictive value (PPV) of 51.8% while maintaining 100% sensitivity. At 97% sensitivity, a specificity of 56.6% and a PPV of 55.8% were obtained.
     

• Application of evolutionary computation and neural network hybrids for breast cancer classification using mammogram and history data.

  • Land, W. H., Jr., Masters, T., Lo, J. Y., and McKee, D. W.  “Application of evolutionary computation and neural network hybrids for breast cancer classification using mammogram and history data.”  IEEE Congress on Evolutionary Computation, Seoul, Korea, 27-29 May 2001.

  • Abstract - Mammography is the modality of choice for early detection of breast cancer, primarily because of its sensitivity. However, mammography has low predictive power, resulting in the biopsy of a large number of benign lesions. This paper explores the use and evaluation of two neural network hybrids as an aid to clinicians. These hybrids have the potential for improving both the sensitivity and specificity of breast cancer screening. The first hybrid, the Generalized Regression Neural Network (GRNN) Oracle, improves the performance output of a set of learning algorithms that operate accurately over the entire (defined) learning space. The second hybrid, an Evolutionary Programming/Adaptive Boosting (EP/AB) based hybrid, intelligently combines the outputs from an iteratively “weak” learning algorithm to “boost” the performance of the weak learner. The second part of this paper discusses modifications to improve the EP/AB hybrid’s performance to show how the use of the EP/AB hybrid is superior to use of an EP only classification system for breast cancer screening.
     

• Application of adaptive boosting to EP-derived multi-layer feedforward neural networks (MLFNs) to improve benign/malignant breast cancer classification.

  • Land, W. H., Jr., Masters, T., Lo, J. Y., and McKee, D. W.  “Application of adaptive boosting to EP-derived multi-layer feedforward neural networks (MLFNs) to improve benign/malignant breast cancer classification.”  SPIE Medical Imaging Conference, San Diego, CA, 17-23 February 2001.

  • Abstract - A new neural network technology was developed for improving the benign/malignant diagnosis of breast cancer using mammogram findings. A new paradigm, Adaptive Boosting (AB), uses a markedly different theory in solutioning Computational Intelligence (CI) problems. AB, a new machine learning paradigm, focuses on finding weak learning algorithm(s) that initially need to provide slightly better than “random” performance (i.e., approximately 55%) when processing a mammogram training set. Then, by successive development of additional architectures (using the mammogram training set), the adaptive boosting process improves the performance of the basic Evolutionary Programming derived neural network architectures. The results of these several EP-derived hybrid architectures are then intelligently combined and tested using a similar validation mammogram data set. Optimization focused on improving specificity and positive predictive value at very high sensitivities, where an analysis of the performance of the hybrid would be most meaningful. Using the DUKE mammogram database of 500 biopsy proven samples, on average this hybrid was able to achieve (under statistical 5-fold cross-validation) a specificity of 48.3% and a positive predictive value (PPV) of 51.8% while maintaining 100% sensitivity. At 97% sensitivity, a specificity of 56.6% and a PPV of 55.8% were obtained.
     

• Using evolutionary computation to develop neural network breast cancer benign/malignant classification models.

  • Land, W. H., Jr., Masters, T., Lo, J. Y., and McKee, D.  “Using evolutionary computation to develop neural network breast cancer benign/malignant classification models.” 4^th World Conference on Systemics, Cybernetics and Informatics (SCI2000), 10:343-347.

  • Abstract - An Evolutionary Programming (EP) benign/malignant breast cancer neural network classification model was developed and investigated which predicts the outcome of mammography-induced breast biopsy. By combining input variables consisting of mammography lesion descriptors and patient history data, the EP derived neural network predicted whether the lesion was benign or malignant. Computer Aided Diagnostic (CAD) tools such as this model may aid in reducing the number of unnecessary biopsies and thus the cost of mammography screening for breast cancer. The EP process was optimized using a data set of 500 biopsy-proven cases from Duke University Medical Center. Results showed that the best EP derived neural network classifier provided an ROC Az index of 0.843 ± 0.053 when averaging performance over 5-fold cross-validation statistical experiments.