Lut Wong

Application of support vector machines to breast cancer screening using mammogram and clinical history data

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 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 data. Measuring performance of the SVM designs and EP/AB hybrid against these objectives will provide quantative 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.

Breast cancer computer aided diagnosis (CAD) using a recently developed SVM/GRNN Oracle hybrid

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 occurred 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 improve both the specificity and PPV of screen film mammography at 95-100% sensitivity, and consistently produce partial A/sub Z/ values (defined as average specificity over the top 10% of the ROC curve) of grater than 50% using a data set of /spl sim/2000 lesions from four different hospitals. As expected, initial experiments demonstrated that combining age and mass margin (AgeMM) that provided the most accurate diagnostic performance. Secondly, the value of the crossover constant, CR = 0.6, provided the best A/sub Z/, while CR = 0.8 resulted in the most accurate partial A/sub Z/ specificity, and PPV at the lower sensitivities. Finally, the results of the GRNN oracle output were essentially the same as 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 cancerous lesions are misdiagnosed) and using a crossover constant of 0.8, approximately 454 biopsies is avoided using this SVM/GRNN oracle diagnostic aid when compared to the circumstance where all 1979 samples were biopsied.

Applying support vector machines to breast cancer diagnosis using screen film mammogram data

This paper explores the use of different support vector machines (SVM) kernels, and combinations of kernels, to ascertain the diagnostic accuracy of a screen film mammogram data set containing /spl cong/ 2500 samples from five different institutions. This research has demonstrated that: (1) specificity improves, on the average, of about 4% at 100% sensitivity and about 18%, on the average, at 98% sensitivity. This means that approximately 52 and 134 women would not have to undergo biopsy, at 100% and 98% sensitivity, when compared to the case of every women being biopsied, which would be necessary to identify all cancers in the absence of a computer aided diagnostic (CAD) process, (2) positive predictive value (PPV) at these same values of sensitivity are much better, ranging from 48% to 51 % as sensitivity is decreased from 100 to 98%. Finally, the average specificity over the top 10% or the ROC curve (which is the average specificity between 90-100% sensitivity) is about 30%. This means that, on the average, 440 women would not have to undergo biopsy, when compared to the case of all women being biopsied.

Data fusion of several support-vector-machine breast-cancer diagnostic paradigms using a GRNN oracle

Breast cancer is second to lung cancer as a tumor-related cause of death in women. For 2003, it was reported that 211,300 new cases and 39,800 deaths would occur in the US. It has been proposed that breast cancer mortality could be decreased by 25% if women in appropriate age groups were screened regularly. Currently, the preferred method for breast cancer screening is mammography, due to its 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 histologic examination. This paper explores the use of a newly designed Support Vector Machine (SVM)/Generalized Regression Neural Network (GRNN) Oracle hybrid and evaluates the hybrid’s performance as an interpretive aid to radiologists. The authors demonstrate that this hybrid has the potential to (1) improve both 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 30%, using a data set of ~2500 lesions from five different hospitals and/or institutions.

New results using multi array sensors and support vector machines for the detection and classification of organophosphate nerve agents

The threat of chemical and biological agents (CBAs) becoming the preferred-weapon of mass destruction (WMD) for international terrorist organizations has produced a significant effort to develop tools that can detect CBAs and effectively combat biochemical warfare. CBAs can attack large populations while leaving infrastructures intact. Despite the availability of numerous sensing devices, intelligent hybrid sensors that can detect and degrade CBAs are virtually nonexistent. This paper reports the integration of multi-array sensors with support vector machines (SVMs) for the detection of organophosphate nerve agents using parathion and dichlorvos as model stimulants compounds. SVMs were used for the design and evaluation of new and more accurate data extraction, preprocessing and classification. Experimental results using structural risk minimization show a significant increase in classification accuracy when compared to the existing AromaScan baseline system. For the most difficult classification task, the Parathion vs Paraoxon pair, the following results were achieved (using the three SVM kernels), as described in this paper: (1) ROC Az indices from approximately 93% to greater than 99%, (2) partial Az values from /spl ap/79% to 93%, (3) specificities from 76% to /spl ap/84% at 100 and 98% sensitivity, and (4) PPVs from 73% to /spl ap/84% at 100% and 98% sensitivities. These are excellent results, considering only one atom differentiates these nerve agents.

Applying knowledge engineering and representation methods to improve support vector machine and multivariate probabilistic neural network CAD performance

Achieving consistent and correct database cases is crucial to the correct evaluation of any computer-assisted diagnostic (CAD) paradigm. This paper describes the application of artificial intelligence (AI), knowledge engineering (KE) and knowledge representation (KR) to a data set of ≈2500 cases from six separate hospitals, with the objective of removing/reducing inconsistent outlier data. Several support vector machine (SVM) kernels were used to measure diagnostic performance of the original and a “cleaned” data set. Specifically, KE and ER principles were applied to the two data sets which were re-examined with respect to the environment and agents. One data set was found to contain 25 non-characterizable sets. The other data set contained 180 non-characterizable sets. CAD system performance was measured with both the original and “cleaned” data sets using two SVM kernels as well as a multivariate probabilistic neural network (PNN). Results demonstrated: (i) a 10% average improvement in overall Az and (ii) approximately a 50% average improvement in partial Az.

Integrating knowledge representation/engineering, the multivariant PNN, and machine learning to improve breast cancer diagnosis

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 breast cancer, its positive predictive value (PPV) is low. One of the main deterrents to achieving high computer aided diagnostic (CAD) accuracy is carelessly developed databases. These “noisy” data sets have always appeared to disrupt learning agents from learning correctly. A new statistical method for cleaning data sets was developed that improves the performance of CAD systems. Initial research efforts showed the following: PLS Az value improved by 8.79% and partial Az improved by 49.71%. The K-PLS Az value at Sigma 4.1 improved by 9.18% and the partial Az by 43.47%. The K-PLS at Sigma 3.6 (best fit sigma with this data set) Az value improved by 9.24% and the partial Az by 44.29%. With larger data sets, the ROC curves potentially could look much better than they do now. The Az value for K-PLS (0.892565) is better than PLS, PNN, and most SVMs. The SVM-rbf kernel was the only agent that out performed the K-PLS with an Az value of 0.895362. However, K-PLS runs much faster and appears to be just as accurate as the SVM-rbf kernel.

New results in computer-aided diagnosis (CAD) of breast cancer using a recently developed SVM/GRNN Oracle hybrid

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 softwares 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[14]. This study demonstrated the following: (1) The DE crossover constant has little, if any, effect on measures of performance (MOP). (2) A specificity of approximately 5.6% is achieved at 100% sensitivity, which increases to approximately 36% at 95% sensitivity. (3) PPV increases from 51% to 56% as sensitivity is decreased from 100 to 95%, respectively.

Using an object-based grid system to evaluate a newly developed EP approach to formulate SVMs as applied to the classification of organophosphate nerve agents

This paper extends the classification approaches described in reference [1] in the following way: (1.) developing and evaluating a new method for evolving organophosphate nerve agent Support Vector Machine (SVM) classifiers using Evolutionary Programming, (2.) conducting research experiments using a larger database of organophosphate nerve agents, and (3.) upgrading the architecture to an object-based grid system for evaluating the classification of EP derived SVMs. Due to the increased threats of chemical and biological weapons of mass destruction (WMD) by international terrorist organizations, a significant effort is underway to develop tools that can be used to detect and effectively combat biochemical warfare. This paper reports the integration of multi-array sensors with Support Vector Machines (SVMs) for the detection of organophosphates nerve agents using a grid computing system called Legion. Grid computing is the use of large collections of heterogeneous, distributed resources (including machines, databases, devices, and users) to support large-scale computations and wide-area data access. Finally, preliminary results using EP derived support vector machines designed to operate on distributed systems have provided accurate classification results. In addition, distributed training time architectures are 50 times faster when compared to standard iterative training time methods.

Integration of multi-array sensors and support vector machines for the detection and classification of organophosphate nerve agents

Due to the increased threats of chemical and biological weapons of mass destruction (WMD) by international terrorist organizations, a significant effort is underway to develop tools that can be used to detect and effectively combat biochemical warfare. Furthermore, recent events have highlighted awareness that chemical and biological agents (CBAs) may become the preferred, cheap alternative WMD, because these agents can effectively attack large populations while leaving infrastructures intact. Despite the availability of numerous sensing devices, intelligent hybrid sensors that can detect and degrade CBAs are virtually nonexistent. This paper reports the integration of multi-array sensors with Support Vector Machines (SVMs) for the detection of organophosphates nerve agents using parathion and dichlorvos as model stimulants compounds. SVMs were used for the design and evaluation of new and more accurate data extraction, preprocessing and classification. Experimental results for the paradigms developed using Structural Risk Minimization, show a significant increase in classification accuracy when compared to the existing AromaScan baseline system. Specifically, the results of this research has demonstrated that, for the Parathion versus Dichlorvos pair, when compared to the AromaScan baseline system: (1) a 23% improvement in the overall ROC Az index using the S2000 kernel, with similar improvements with the Gaussian and polynomial (of degree 2) kernels, (2) a significant 173% improvement in specificity with the S2000 kernel. This means that the number of false negative errors were reduced by 173%, while making no false positive errors, when compared to the AromaScan base line performance. (3) The Gaussian and polynomial kernels demonstrated similar specificity at 100% sensitivity. All SVM classifiers provided essentially perfect classification performance for the Dichlorvos versus Trichlorfon pair. For the most difficult classification task, the Parathion versus Paraoxon pair, the following results were achieved (using the three SVM kernels: (1) ROC Az indices from approximately 93% to greater than 99%, (2) partial Az values from ≈79% to 93%, (3) specificities from 76% to ≈84% at 100 and 98% sensitivity, and (4) PPVs from 73% to ≈84% at 100% and 98% sensitivities. These are excellent results, considering only one atom differentiates these nerve agents.