Master Thesis Defense: Qingbo Kang
Speaker: Qingbo Kang
Supervisor: Dr. T. Fevens
Drs. T. D. Bui, A. Krzyzak, D. Pankratov (Chair)
Title: Medical Image Segmentation by Deep Convolutional Neural Networks
Date: Thursday, August 1, 2019
Place: EV 1.162
Medical image segmentation is a fundamental and critical step for medical image analysis. Due to the complexity and diversity of medical images, the segmentation of medical images continues to be a challenging problem. Recently, deep learning techniques, especially Convolution Neural Networks (CNNs) have received extensive research and achieve great success in many vision tasks. Specifically, with the advent of Fully Convolutional Networks (FCNs), automatic medical image segmentation based on FCNs is a promising research field. This thesis focuses on two medical image segmentation tasks: lung segmentation in chest X-ray images and nuclei segmentation in histopathological images.
For the lung segmentation task, we investigate several FCNs that have been successful in semantic and medical image segmentation. We evaluate the performance of these different FCNs on three publicly available chest X-ray image datasets.
For the nuclei segmentation task, since the challenges of this task are difficulty in segmenting the small, overlapping and touching nuclei, and limited ability of generalization to nuclei in different organs and tissue types, we propose a novel nuclei segmentation approach based on a two-stage learning framework and Deep Layer Aggregation (DLA). We convert the original binary segmentation task into a two-step task by adding nuclei-boundary prediction (3-classes) as an intermediate step. To solve our two-step task, we design a two-stage learning framework by stacking two U-Nets. The first stage estimates nuclei and their coarse boundaries while the second stage outputs the final fine-grained segmentation map.
Furthermore, we also extend the U-Nets with DLA by iteratively merging features across different levels. We evaluate our proposed method on two public diverse nuclei datasets. The experimental results show that our proposed approach outperforms many standard segmentation architectures and
recently proposed nuclei segmentation methods, and can be easily generalized across different cell types in various organs.