Projects
Selected research projects, each summarising its motivation, main achievements, key publications, software, datasets, and the students I supervised on it.
Ophthalmology — Uveitis
AI-assisted grading of intraocular inflammation in uveitis from retinal fluorescein angiography, developed with the Jules-Gonin Eye Hospital.
Neurophysiology — Sleep Medicine
Generalisable, reproducible automatic sleep-stage classification from polysomnography, working toward foundation models for sleep medicine.
Trustworthy and Fair Medical AI
Making medical AI fair across patient subgroups and useful in the clinic, with evaluation frameworks, foundation-model bias analysis, and utility-fairness trade-offs.
Ophthalmology — Retinal Image Analysis
Segmentation and analysis of retinal fundus images (vessels, optic disc, blood flow) with minimalistic, reproducible, and clinically deployable models.
Radiology — Robust Image Analysis
Interpretable, generalisable computer-aided analysis of radiological images — detecting tuberculosis on chest X-rays and extracting stable quantitative biomarkers from CT — built to hold up across scanners and populations.
Epidemiology — Anticipating Risk
Machine learning on routine clinical records, chest radiographs, and patient-contact networks to flag who is most at risk — from diabetic-foot complications to drug-resistant infection to tuberculosis.
Angiology — Fibromuscular Dysplasia
Machine learning for the characterisation and diagnosis of fibromuscular dysplasia, in collaboration with CHUV's angiology department.
Reproducible Research and Open Science
Software suites and open platforms that make computational research repeatable, shareable, extensible, and stable, including the BEAT open-science platform.
Earlier career
Remote Photoplethysmography and Vital Signs
2016–2017
Reproducible, camera-based estimation of vital signs, putting remote photoplethysmography on a shared, openly benchmarked footing.
Biometrics and Presentation-Attack Detection
2010–2018
Earlier work with the Idiap Biometrics group on robust recognition and presentation-attack detection that generalises to unseen attacks.
High-Energy Physics — ATLAS Trigger (CERN)
2001–2010
Earlier work at CERN's ATLAS experiment on real-time trigger and data-acquisition systems, and neural-network methods for particle recognition.