Addressing Variations and Uncertainties in Medical Image Data for Deep Learning

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering


PhD Thesis Defence


Title: "Addressing Variations and Uncertainties in Medical Image Data for Deep 
Learning"

By

Mr. Zengqiang YAN


Abstract

One reason for the success of deep learning in natural image data is the 
availability of large-scale labeled data. However, labeled medical image data 
often is limited, as annotating medical image data requires extensive human 
efforts and expertise. Consequently, the capacity of deep learning usually 
cannot be fully explored. In this thesis, we propose to improve deep learning 
performance by addressing variations and uncertainties in medical image data.

First, we study the inter-observer problem in retinal vessel segmentation, 
where human observers can generate different pixel-wise annotations given the 
same medical image. To address the problem, we design a vessel thickness 
similarity measure and construct a segment-level loss function based on the 
measure. Then, we integrate the loss function with a two-branch deep learning 
framework. Comprehensive experiments on publicly available datasets demonstrate 
the effectiveness of our approach.

Second, we investigate the boundary uncertainty problem in gland instance 
segmentation. Due to limited image resolution, annotated boundaries usually are 
not absolutely correct, which makes it challenging to preserve shape 
information in gland instance segmentation. To address the problem, we propose 
a new shape-preserving loss function, together with pseudo domain adaptation, 
to enable one single deep learning model for accurate gland instance 
segmentation. Our evaluations confirm that our method can obtain better 
performance compared with other state-of-the-art methods.

Lastly, we discuss the cross-client variation problem, where image data from 
different sources can vary significantly. It will be the bottleneck when 
applying federated learning to train deep learning models from multi-source 
decentralized medical image data. We, for the first time, propose a 
variation-aware federated learning framework to address the problem. 
Experimental results on the classification of clinically significant prostate 
cancer from multi-source decentralized ADC image data show that our framework 
outperforms the current federated learning framework, especially in dealing 
with small datasets.


Date:			Monday, 24 August 2020

Time:			2:30pm - 4:30pm

Zoom Meeting:		https://hkust.zoom.com.cn/j/94378050514

Chairperson:		Prof. Jiheng ZHANG (IEDA)

Committee Members:	Prof. Tim CHENG (Supervisor)
 			Prof. Albert CHUNG
 			Prof. Pedro SANDER
 			Prof. Weichuan YU (ECE)
 			Prof. Pheng Ann HENG (CUHK)


**** ALL are Welcome ****