3Accurate detection of cancer region from the MRI/ultrasound images and segmentation of the prostate contour is critical for the diagnosis of prostate cancer. However, the process is very demanding on the urologist’s expertise. An autonomous image processing technique is desired to promote the diagnosis and thus the treatment outcomes.
A solution based on a neural network is proposed to achieve autonomous prostate segmentation and cancer detection. A Deep-Convolution Neural Network was developed as the backbone of the solution for improved accuracy and reduced inference time. Experiments were conducted in which medical images were fed to the proposed processor and the outcomes were compared to the ground truth labelled by experienced urologists. The functioning of the proposed solution was successfully verified.
Personnel - Mr Hanwen Kang (PhD student)