Computer Diagnosis of Breast Thermograms

Abstract

A computer-based technique was developed for the automated diagnosis of breast thermogram. Eighty-five thermograms were digitized and analyzed, and 23 parameters computed for each thermogram. A statistical decision program, based on a linear discriminant analysis technique, classified the thermograms into normal and abnormal categories. The accuracy of classification was evaluated by several techniques and found to be comparable to that of trained thermographers.