AUTOMATED PROSTATE CANCER DETECTION AND LOCALIZATION IN THE PERIPHERAL ZONE OF THE PROSTATE IN MULTI-PARAMETRIC MR IMAGES

摘要

In a multi-parameter magnetic resonance imaging (MRI) image according to an exemplary embodiment of the present invention, a method of automatically detecting and localizing a prostate cancer in a region of the prostate gland may include: (A) performing a training multi-parameter magnetic resonance imaging Normalizing the signal intensity histogram between training identical parameters MR images; (B) Normalized training. The feature vectors are extracted from the MR image, and the optimal feature vectors for classifying the normal tissue and the prostate cancer are selected from the extracted feature vectors. The machine learning classification model ; And (C) normalizing the signal intensity histogram between test identical parameter MR images in a test multi-parameter magnetic resonance image for the region of the prostate gland, and comparing the selected feature (s) for each parameter MR image with the normalized test multiple- Extracting the extracted feature vectors, and sequentially applying the machine learning classification model for each parameter MR image according to the priority order of each parameter MR image to the extracted feature vectors to detect the prostate cancer in the region of the prostate circumference.