Image feature extraction method with SIFT to diagnose prostate cancer

Number of prostate cancer patients are increasing but diagnosing pathologists are not enough. Then supporting system to diagnose prostate cancer on biopsy image is needed by pathologists. However staining conditions of prostate biopsy images are uneven, so extracting their form by automatic method is difficult. Because of this problem, previous study aiming to diagnose each glandular needed to except imperfect extraction glandular. Then we propose new feature extraction method (Area Rates on Components and SIFT + BoK) aiming to improve accuracy on imperfect extracted glands. In our experiment, diagnosing accuracy is improved 6.3% ~ 13.3%.