Reading skin disease
with computation, biology,
and the clinic.

Scalable, self-supervised vision models pretrained on large, heterogeneous skin-image collections — a shared backbone for diagnosis, segmentation, and retrieval.

An open dataset of pigment-rich paediatric skin disease from Sub-Saharan Africa — 4,901 images from 1,653 patients — closing the diversity gap in dermatology AI. (MICCAI 2024)

Datasets and methods that measure and correct how skin-tone imbalance degrades model performance — so tools work for everyone.

Clinician-plus-AI melanoma screening, validated at planet scale — including independent validation in Korea. (NPJ Digital Medicine 2025)

Generative models that show the future cost of sun exposure — a novel, image-driven route to skin-cancer prevention.

3D cultures of healthy and diseased epidermis — a model of Darier disease and other monogenic keratinocyte disorders, built for high-throughput drug screening.

Why some medications trigger life-threatening skin reactions (SJS/TEN) — tracing IL-1 family cytokines and CD8⁺ T-cell–driven keratinocyte death.

The molecular drivers of a chronic inflammatory disease affecting millions — the antimicrobial peptide LL-37 and JAK/STAT signalling.

3D total-body photography (VECTRA) for early melanoma detection in high-risk patients — prospective evaluation of AI-powered whole-body screening.

Deep clinical phenotyping of generalised pustular psoriasis across 500+ patients in Europe and Egypt — flare patterns, comorbidity, and outcomes.

Contributing to the European Baseline Series — tracking emerging allergens across large multicentre patch-test cohorts.