APPLICATION OF ARTIFICIAL NEURAL NETWORKS TO MEDICAL IMAGE PATTERN-RECOGNITION - DETECTION OF CLUSTERED MICROCALCIFICATIONS ON MAMMOGRAMS AND LUNG-CANCER ON CHEST RADIOGRAPHS

Three neural network models were employed to evaluate their performanc es in the recognition of medical image patterns associated with lung c ancer and breast cancer in radiography. The first method was a pattern match neural network. The second was a conventional backpropagation n eural network. The third method was a backpropagation trained neocogni tron in which the signal propagation is operated with the convolution calculation from one layer to the next. In the convolution neural netw ork (CNN) experiment, several output association methods and trainer i mposed driving functions in conjunction with the convolution neural ne twork are proposed for general medical image pattern recognition. An u nconventional method of applying rotation and shift invariance is also used to enhance the performance of the neural nets. We have tested th ese methods for the detection of microcalcifications on mammograms and lung nodules on chest radiographs. Pre-scan methods were previously d escribed in our early publications. The artificial neural networks act as final detection classifiers to determine ifa disease pattern is pr esented on the suspected image area. We found that the convolution neu ral network, which internally performs feature extraction and classifi cation, achieves the best performance among the three neural network m odels. These results show that some processing associated with disease feature extraction is a necessary step before a classifier can make a n accurate determination.