Medical image segmentation using random forest and convolutional neural networks

• Main Author: Duan, You
• Other Authors: Ponnuthurai N. Suganthan
• Format: Theses and Dissertations
• Language: English
• Published: 2018
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/76333
• Institution: Nanyang Technological University

Medical image analysis has received attention in recent years. It is an important part of disease analysis, diagnosis and treatment. Due to the limited number of doctors, some patients cannot receive timely treatment. In recent years, supervised learning algorithms have developed rapidly and have made major breakthroughs in the field of medical image analysis. Among them, Convolutional Neural Network (CNN) and Random Forest (RF) are proved to be powerful tools for computer vision tasks, including target recognition, segmentation and localization. In this dissertation, I used the CNN and RF methods to process the Nuclei Segmentation image dataset, which include big number of targets and boundaries among these targets are not obvious, and then analyze and compare with the previous experimental phenomena, then use CNN and RF methods to process the digital retinal image (DRIVE) database, whose target is not obvious and light is unevenly distributed, and compare with state-of-the-art experiment. This dissertation found that Convolutional Neural Network (CNN) works well in the field of image segmentation. Random Forest (RF) can also obtain good results when processing image segmentation, but it will be affected by the image noise.