From Thermograms to Concepts: Using Concept Bottleneck Models to Balance Accuracy and Interpretability in Clinical AI Thermography
This is a hybrid event. Please find the Teams link in the abstract.
This presentation, “From Thermograms to Concepts: Using Post hoc Concept Bottleneck Models (PCBMs) to Balance Accuracy and Interpretability in Clinical AI Thermography,” will explore the critical role of interpretable AI in clinical settings, particularly within the evolving field of thermography.
The discussion will introduce the core idea of concept-based models (CBMs), highlighting their inherent interpretability while acknowledging the practical limitations they face in real-world deployments. Building on this, we present Post hoc Concept Bottleneck Models (PCBMs), demonstrating how they overcome these challenges. We’ll discuss PCBMs’ two primary approaches to reveal decision pathways and biases
The discussion then centers on the specific complexities of clinical thermography. Given its evolving potential and the multifactorial risk factors that necessitate longitudinal clinical trials, we will illustrate how PCBMs can provide a crucial tool for understanding which predefined, guideline-driven concepts genuinely inform stratified risk assessments.
Finally, Practical implementation is addressed by highlighting how PCBMs can be applied to thermographic data across longitudinal outcome studies. And how this approach not only supports outcome validation but also strengthens clinician engagement. By enabling healthcare professionals to interrogate, validate, and even influence the model’s conceptual focus.
Teams link: teams.microsoft.com/l/meetup-join/19%3ameeting_NGY2OWZmMWMtN2I1My00ODJkLWI1MmEtNDdiMzYwOTIzNjlj%40thread.v2/0?context=%7b%22Tid%22%3a%22cc95de1b-97f5-4f93-b4ba-fe68b852cf91%22%2c%22Oid%22%3a%22ecdf56c5-98aa-4971-b0e3-75dcdb719f12%22%7d
Bio:
Mustafa Alghali is the Co-founder and Chief Technology Officer of Amplifai Health, a Saudi-based healthcare technology company established in 2023. With a background in computer science and a focus on applied artificial intelligence, Mustafa has contributed to cutting-edge AI research at Mila – Quebec AI Institute and developed large-scale AI systems at Unity Technologies. As a technical visionary, he now leads the R&D team at Amplifai Health in building TFScan an AI-powered thermal imaging system designed for the early detection of diabetic foot ulcers.
Date:
10 June 2025, 13:00
Venue:
Wolfson College, Linton Road OX2 6UD
Venue Details:
Seminar Room 3
Speaker:
Mustafa Alghali (Amplifai Health)
Organising department:
Wolfson College
Organiser:
Dr. Yi Yin (Wolfson College, University of Oxford)
Organiser contact email address:
yi.yin@wrh.ox.ac.uk
Part of:
Oxford Cross-Disciplinary Machine Learning (OxfordXML) Research Cluster Seminar Series
Booking required?:
Not required
Cost:
Free (Coffee, tea and cake will be provided)
Audience:
Public
Editor:
Yi Yin
