Publications

Counting monkeypox lesions in patient photographs: limits of agreement of manual counts and artificial intelligence

J Invest Dermatol, 2023, Feb

In support of a randomized controlled trial (NCT05559099; PI: Muyembe-Tamfum), this study demonstrated the ability of our AI algorithms to count cutaneous mpox (formerly monkeypox) lesions from photos with human-level accuracy. This is important because lesion count is the primary factor in World Health Organization mpox severity scoring and must be done frequently in the clinical trial, but is very laborious and error-prone when done manually.

Artificial intelligence recognition of cutaneous chronic graft-versus-host disease by a deep learning neural network

British Journal of Haematology, 2022 Jun

This work demonstrated the ability of our AI computer vision algorithm to recognize and mark out (segment) the surface area of skin affected by chronic graft-vs-host disease (GVHD) from photos with superior accuracy to standard clinical estimates of body surface area.

Prognostic Value of Cutaneous Disease Severity Estimates on Survival Outcomes in Patients With Chronic Graft-vs-Host Disease

JAMA Dermatol, 2023 Apr.

This analysis of multicenter data revealed that erythema body surface area at baseline and follow-up is an important longitudinal biomarker that predicts survival more accurately than the reference standard (NIH Skin Score) in patients requiring immunosuppression for chronic graft-vs-host disease (GVHD).

Association of leukocyte adhesion and rolling in skin with patient outcomes after hematopoietic cell transplantation using noninvasive reflectance confocal videomicroscopy

JAMA Dermatol, 2022 Jun

In this project, we discovered that live videos of white blood cells in the skin, taken noninvasively at patient bedside with confocal microscopy, more strongly predict relapse and survival outcomes after transplant than the best available clinical risk assessment.