Neural Networks For Pattern Recognition In Medical Diagnosis
John H. Frenster · 2005
Neural network simulation systems have been adapted for two types of pattern recognition analyses, utilizing up to 400 input quantitative diagnostic findings in each of 850 output medical diagnostic categories, leading to the correct selection of one in up to 400 competing output diagnoses per trained neural network, in a total of 6 trained d iagnostic neural networks. The input data was either entered quantitatively within i ndividual pixels, or semi-quant ita tively within larger data fields. The output medical diagnoses were displayed either in text or in 3-5 digit ICD-9-CM diagnostic codes.