Evaluating a Neural Network Decision-Support Tool for the Diagnosis of Breast Cancer

This paper describes the evaluation of an application of the ARTMAP neural network model to the diagnosis of cancer from fine-needle aspirates of the breast. The network has previously demonstrated very high performance when used with high-quality data provided by an expert pathologist. New performance results are provided for its use with “noisy” data provided by an inexperienced pathologist. Additionally, ARTMAP supports the extraction of symbolic rules from a trained network and the validity of these autonomously-acquired rules is discussed. It is concluded that the symbolic rules provide an appropriate mapping of input features to category classes in the domain. However, the network in its present form is only suitable for use as a decision-support tool by a senior pathologist, since its performance deteriorated greatly with poor-quality data provided by a junior pathologist. The implications of the findings are discussed.

[1]  Chee Peng Lim,et al.  Modified Fuzzy ARTMAP Approaches Bayes Optimal Classification Rates: An Empirical Demonstration , 1997, Neural Networks.

[2]  George Cybenko,et al.  Approximation by superpositions of a sigmoidal function , 1992, Math. Control. Signals Syst..

[3]  P. Silcocks,et al.  Measuring repeatability and validity of histological diagnosis--a brief review with some practical examples. , 1983, Journal of clinical pathology.

[4]  I. Ellis,et al.  Pathology and breast screening , 1990, Histopathology.

[5]  N Kirkham,et al.  Computer assisted diagnosis of fine needle aspirate of the breast. , 1990, Journal of clinical pathology.

[6]  Stephen Grossberg,et al.  A massively parallel architecture for a self-organizing neural pattern recognition machine , 1988, Comput. Vis. Graph. Image Process..

[7]  Jooyoung Park,et al.  Universal Approximation Using Radial-Basis-Function Networks , 1991, Neural Computation.

[8]  I. Ellis,et al.  Guidelines for cytology procedures and reporting on fine needle aspirates of the breast , 1994, Cytopathology : official journal of the British Society for Clinical Cytology.

[9]  Chee Peng Lim,et al.  Autonomously Learning Neural Networks for Clinical Decision Support. , 1994 .

[10]  Stephen Grossberg,et al.  Fuzzy ARTMAP: A neural network architecture for incremental supervised learning of analog multidimensional maps , 1992, IEEE Trans. Neural Networks.

[11]  P. Bartels,et al.  Expert system support using Bayesian belief networks in the diagnosis of fine needle aspiration biopsy specimens of the breast. , 1994, Journal of clinical pathology.

[12]  Robert F. Harrison,et al.  A Prototype Neural Network Decision-Support Tool for the Early Diagnosis of Acute Myocardial Infarction , 1995, AIME.

[13]  A. Shorthouse,et al.  Fine needle aspiration of palpable breast lesions. Results obtained with cytocentrifuge preparation of aspirates. , 1988, Acta cytologica.

[14]  Geoffrey E. Hinton,et al.  Learning representations by back-propagating errors , 1986, Nature.

[15]  O L Mangasarian,et al.  Computer-designed expert systems for breast cytology diagnosis. , 1993, Analytical and quantitative cytology and histology.

[16]  Stephen Grossberg,et al.  ARTMAP: supervised real-time learning and classification of nonstationary data by a self-organizing neural network , 1991, [1991 Proceedings] IEEE Conference on Neural Networks for Ocean Engineering.

[17]  Ah-Hwee Tan,et al.  Rule Extraction, Fuzzy ARTMAP, and Medical Databases , 1993 .

[18]  John Moody,et al.  Fast Learning in Networks of Locally-Tuned Processing Units , 1989, Neural Computation.

[19]  Frederick Hayes-Roth,et al.  Building expert systems , 1983, Advanced book program.