Deep Generative Models in Biomedical Imaging: A Survey

PhD Qualifying Examination


Title: "Deep Generative Models in Biomedical Imaging: A Survey"

by

Mr. Juyoung BAE


Abstract:

The past decade has witnessed remarkable progress in deep generative models, 
particularly in their capacity to produce increasingly realistic visual data. 
However, translating these capabilities into clinical or scientific practice 
faces challenges distinct from natural image domains, such as limited and 
imbalanced datasets, compromises in acquisition quality, multi-site domain 
shifts, and the need for clinically meaningful validation. Existing surveys 
have largely organized the literature by model architecture or individual 
task, providing limited guidance on where generative methods deliver 
real-world value. This survey addresses this gap through an 
implementation-agnostic application taxonomy grounded in translational impact, 
organizing the field by the clinical or scientific problem being solved. The 
survey first traces methodological progression and systematically reviews 
representative methods within each translational category to highlight 
underlying technical choices, clinical or biological contexts, practical 
contributions, and limitations. It further examines evaluation and quality 
assurance practices, concluding with an outline of essential future directions 
imperative for realizing the field's translational promise.


Date:                   Monday, 2 March 2026

Time:                   3:00pm - 5:00pm

Venue:                  Room 2132C
                        Lift 19

Committee Members:      Dr. Hao Chen (Supervisor)
                        Prof. Chi-Keung Tang (Chairperson)
                        Dr. Terence Tsz Wai Wong (CBE)