چكيده به لاتين
Prostate lesions are one of the most common diseases in men, So the prostate cancer is the second deadliest cancer among men after lung cancer. Non-cancerous prostate lesions include prostate cysts, calcification or prostate stones, benign prostatic hyperplasia or prostate enlargement, and prostate inflammation or prostatitis.
Since CT images have been taken in multiple sections in parallel from one patient member, reviewing all images manually is time consuming and error prone. Therefore, the use of image processing techniques can systematically examine all images in less time and prevent errors. Automated detection of lesions through medical imaging can also help in the future in the development of surgeon assistive robots. In fact, the first step in designing these systems and robots to help physicians is to make computer vision possible.
Due to the tissue similarity of the prostate to the other organs around it, one of the challenges in abdominal and pelvic of CT-Scan images is diagnosis of prostate gland. In this thesis we aim to propose a novel approach for detection, classification and segmentation of CT-Scan images. Then, by introduction of a three-step innovative approach, we investigate the presence of three common non-cancerous lesions in the prostate gland. Finally, by designing a system, the steps mentioned in this thesis are presented in the form of a medical assistant system.
Keywords:
Image processing, Prostate lesions, Neural networks, Deep learning
