03/28/2023
By Hengyong Yu

The Francis College of Engineering, Department of Electrical & Computer Engineering (ECE), invites you to attend a Ph.D. dissertation proposal defense by Yongshun Xu on “Medical Image Data Synthesis and Its Deep Learning based Applications.”

Ph.D. Candidate: Yongshun Xu
Defense Date: Tuesday, April 11, 2023
Time: 9 to 10:30 a.m.
Location: This will be a virtual defense via Zoom. Those interested in attending should contact the student (Yongshun_xu@student.uml.edu) or committee advisor (Hengyong_Yu@uml.edu) at least 24 hours prior to the defense to request access to the meeting.

Committee Chair: Advisor Hengyong Yu, Professor, Electrical and Computer Engineering, University of Massachusetts Lowell

Committee Members:

  • Tricia Chigan, Ph.D., Professor, Electrical & Computer Engineering, University of Massachusetts Lowell
  • Yan Luo, Ph.D., Professor, Electrical & Computer Engineering, University of Massachusetts Lowell
  • Yu Cao, Ph.D., Professor, Computer Science, University of Massachusetts Lowell

Abstract: Medical imaging has many different modalities, including computerized tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET), ultrasound, etc. They are widely used in hospitals and clinics for early detection, diagnosis, and treatment of diseases. The interpretation and analysis of the corresponding images need professional and experienced radiologists. This is affected by subjectivity, variation perception, and fatigue. Using computer aided diagnosis technology to analyze and process images can assist clinical physicians to improve the accuracy and reliability. In recent years, great success in computer vision has promoted the application of deep learning in medical image processing tasks, such as image classification, image segmentation, and image recognition. Deep learning has demonstrated the ability to perform clinical decision making exceeds that of expert human operators. Currently, the major challenge is to collect large amounts of well-labeled training data to train a good neural network. This be expert annotation is expensive and there are many regularizations on the collection of real patient’s data. As an alternative strategy, synthetic image data could provide high-quality large datasets. We validate that the synthetic image-augmented cardiac CT image could improve the network classification accuracy on the real dataset. For the evaluation of cardiovascular diseases using cardiac CT image, high-quality image could provide more sufficient and accurate information for diagnosis, and image quality is highly relevant to radiation dose. Using the synthesized data, we also evaluate a deep learning technique for image denoising and image inpainting on limit-angle CT image reconstruction, which can help to improve the temporal resolution and reduce the radiation dose. Finally, we evaluate the deep learning method for classification of non-alcoholic fatty liver disease from ultrasound images.

All interested students and faculty members are invited to attend the online defense via remote access.