More about HKUST
Deep Learning for Computational Cytopathology
PhD Thesis Proposal Defence
Title: "Deep Learning for Computational Cytopathology"
by
Mr. Hao JIANG
Abstract:
Computational cytopathology holds promise for precancerous screening through
morphometric assessment of cellular specimens, yet clinical translation is
constrained by morphological ambiguity, spatial crowding, and domain shifts
across multi-center cohorts. This proposal establishes a systematic framework
progressing from instance-level analysis to whole-slide screening. The work
surveys prevailing paradigms for cytological perception and weakly-supervised
classification, then addresses categorical ambiguity via contrastive
feature-space consistency for robust instance characterization. For
segmentation under spatial occlusion, a decomposition-recombination strategy
disentangles crowded clusters, extended by structure-aware refinement with
anomaly-guided supervision for precise morphological delineation. At the
patient level, these insights inform a generalizable screening architecture
integrating self-supervised pretraining, multi-instance aggregation, and
test-time domain adaptation-translating cellular evidence into actionable
decisions across heterogeneous settings. Collectively, this work charts an
evolutionary path from cellular morphometrics to holistic diagnostic
reasoning, advancing AI-assisted cytopathology for precancerous screening and
patient triage.
Date: Monday, 6 July 2026
Time: 3:00pm - 5:00pm
Venue: Room 3494
Lifts 25/26
Committee Members: Dr. Hao Chen (Supervisor)
Prof. Pedro Sander (Chairperson)
Dr. Terence Tsz Wai Wong (CBE)