More about HKUST
Foundation Models in Multi-sequence MRI Analysis
PhD Qualifying Examination
Title: "Foundation Models in Multi-sequence MRI Analysis"
by
Mr. Zelin QIU
Abstract:
Magnetic Resonance Imaging (MRI) is an essential diagnostic tool in modern
radiology, valued for its exceptional soft-tissue contrast and lack of
ionizing radiation. A single MRI examination typically yields multiple
sequences, each sensitive to a distinct set of tissue properties. Therefore,
accurate interpretation requires integrating complementary information from
these sequences, and their heterogeneity prolongs reading time and contributes
to the increasing workload of radiologists. Deep learning has demonstrated
superior performance in MRI analysis across various clinical applications,
facilitating more accurate diagnostics, enabling personalized treatment
regimens, uncovering biological insights, and optimizing healthcare delivery
efficiency. Nevertheless, the acquisition of specialized training data still
faces significant obstacles due to the high cost of annotation. Foundation
models have recently emerged as a promising approach to mitigate these
challenges and are adaptable to numerous clinical tasks, as they learn
generalizable representations by pre-training on extensive, diverse MRI
datasets. In this survey, we present a comprehensive review of recent advances
in foundation models for multi-sequence MRI analysis. We review the benchmarks
and evaluation metrics employed to assess these models. Then we organize
existing methods via a taxonomy of self-supervised pre-training paradigms and
discuss the advanced strategies tailored to the multi-sequence setting.
Finally, we identify the core challenges in the field and outline promising
future research directions towards a more generalizable, interpretable, and
collaborative framework.
Date: Thursday, 10 September 2026
Time: 10:00am - 12:00noon
Venue: Room 3494
Lift 25/26
Committee Members: Dr. Hao Chen (Supervisor)
Dr. Long Chen (Chairperson)
Dr. Sehi L'Yi