Tumor Disease Diagnosis Model Based on BP Neural Network

摘要

The study is based on the principles of mathematical morphology and the features of tumor cell image about unclearness and uncertainty. It adopts fast median filtering in spatial domain and classification method of statistical pattern recognition to process the image segmentation of cell images. To describe the objects to be recognized, it extracts the features and adopts neural network theory for disease diagnosis. GA is used to optimized BP algorithm to fast its convergence speed, to acquire the effect of rapid and accurate diagnosis. The system tests adopts 210 case samples to train the system. After 18 reverse error adjustment the total error is less than 0.0001 and it indicates our model can make a correct diagnosis according to the training samples.