MEMP PhD Thesis Defense: Luca Weishaupt

Augmenting Pathology Workflows via Interpretable and Multimodal AI Systems

In clinical pathology, diagnostic consensus relies on the manual morphological assessment of tissue. While computational methods can streamline workflows and reduce inter-observer variability, the opaque nature of most deep learning systems precludes rigorous auditability by pathologists. This lack of transparency remains a primary barrier to clinical translation. This thesis presents a framework of computational pathology systems at three increasing levels of automation. Across these systems, we demonstrate that computational autonomy can scale without compromising interpretability, provided that model predictions remain explicitly tied to the morphological evidence a pathologist would use for verification. At the first tier, we introduce REMORA, an interactive segmentation framework for measuring glomerular basement membrane thickness. Evaluated in a clinical reader study, REMORA accelerates diagnostic workflows and reduces inter-expert discordance while maintaining standard-of-care accuracy. At the second tier, we present ToxScribe, a visual question-answering system for veterinary toxicologic pathology, a domain where human-pathology models fail to transfer. To evaluate this system, we curated VIPER, the first expert-annotated benchmark in this field. On this benchmark, ToxScribe outperforms both general-purpose and pathology-specific vision-language models. At the third tier, we detail SlideSeek, an autonomous whole-slide image diagnostic system. SlideSeek establishes complete diagnoses and explicitly cites the supporting spatial regions for each claim. On a benchmark of rare diseases, SlideSeek substantially outperforms current leading models. Across all three systems, the degree of manual intervention changes, but the pathologist's diagnostic authority is preserved through evidence-grounded architectures.

Thesis Supervisor:
Faisal Mahmood, PhD
Associate Professor of Pathology, Harvard Medical School and Brigham & Women's Hospital

Thesis Committee Chair:
Thomas Heldt, PhD
Professor of Electrical Engineering and Computer Science, MIT

Thesis Reader:
Long Phi Le, MD, PhD
Director of Computational Pathology, Massachusetts General Hospital and Harvard Medical School
Vice Chair for Pathology Informatics at Massachusetts General Hospital
Assistant Professor of Pathology at Harvard Medical School
________________________________________________________________________________________

Zoom Invitation
Luca Weishaupt is inviting you to a scheduled Zoom meeting

Topic: Luca Weishaupt MEMP PhD Thesis Defense
Time: Friday, August 7, 2026, 4:00 PM Eastern Time (US and Canada)

Your participation is important to us: please notify hst [at] mit.edu (hst[at]mit[dot]edu), at least 3 business days in advance, if you require accommodations in order to access this event.

Join Zoom Meeting
https://mit.zoom.us/j/91361754761

One tap mobile
+16465588656,,91361754761# US (New York)
+16699006833,,91361754761# US (San Jose)

Meeting ID: 913 6175 4761

US: +1 646 558 8656 or +1 669 900 6833

International Numbers: https://mit.zoom.us/u/aCPXY35V4

Join by SIP
91361754761 [at] zoomcrc.com (91361754761[at]zoomcrc[dot]com)

Join by Skype for Business
https://mit.zoom.us/skype/91361754761