Medical Image Analysis by Machine Learning Source code and Document

Project Discription

One of important part of human life is health and some of information storage provide such valuable data. In order to increase the trust on such type of stored data different technique was applied by various researchers. There have been dramatic increase in the number of medical images which are taken for treatment planning, diagnosis, and other clinical purposes. In the current clinical standards, these measurements (annotations) from medical scans are done by expert physicians. This work proposed HFCMID  (Histogram and Fuzzy C Mean Clustering Medical Image Diagnosis). Input images for training undergoes pre-processing and extracted features were used for the training of neural network. Histogram visual feature was used for the training with fuzz C means cluster center pixel values. Combination of these two feature has improved the learning of neural network. Explanation of whole model was done with an running example of where input image is pre-process and feature extract for prediction. Experiment was done on real dataset image of medical diagnosis. It was found that proposed model has increase the prediction accuracy by 44.55% as compared to CNN based model [30]. Similarly the precision value of prediction by 41.36% as compared to previous method. Use of histogram feature and FCM cluster center pixel values increased the learning of the neural network and predict more accurate image class. HFCMID model has achieved precision value 100% in all case of testing images.

Project Sample Image