The SKIN DEEP Study: Advancing AI Diagnosis of Skin Diseases through Multi-Sensory Data Enhancement and Processing - This project addresses persistent gaps in access to expert diagnosis and management of skin diseases by integrating thermography and artificial intelligence (AI) into scalable, field-adapted solutions. Skin diseases affect more than 1.8 billion people globally, and in the United States, 1 in 5 individuals seek treatment annually. Despite this burden, access to dermatologic expertise remains limited in many rural, medically underserved, and underserved urban communities, where specialist shortages, prolonged wait times, and long travel distances delay care and contribute to preventable disease progression and chronic morbidity. To address these challenges, this project will develop and implement a multi-sensor remote diagnostic platform designed to improve early triage, diagnostic accuracy, and assessment of disease progression, with direct relevance to expanding timely dermatologic care in underserved U.S. communities. This study will be conducted in rural Côte d’Ivoire due to the high endemicity and co-endemicity of clinically distinct skin diseases not available at comparable scale in the United States. This epidemiologic setting provides a scientifically and ethically compelling environment to develop, train, and validate AI-enabled diagnostic tools across diverse skin disease presentations directly relevant to U.S. conditions, including inflammatory diseases, infections, and chronic wounds. The concentration and diversity of cases will enable robust algorithm development and validation within a shorter timeframe and with greater clinical variability than feasible in a single U.S. population. In addition, conducting this work in Côte d’Ivoire allows rigorous testing of diagnostic performance and implementation strategies under challenging conditions—including limited connectivity, workforce shortages, and resource constraints—that increasingly characterize healthcare delivery in medically underserved regions of the United States. During the R21 phase, to be conducted at Tulane University, we will collect multi-sensor data, develop and validate algorithms drawing upon existing collaborations and datasets, and employ a two-level masking approach to generate AI algorithms capable of improved diagnostic performance. In the subsequent R33 phase, we will develop a user-friendly, smartphone-based multi-sensor data platform and field-test our approach in more challenging settings in Côte d'Ivoire. We will gather large-scale image datasets, evaluate performance and user acceptability, and iteratively refine on-device AI algorithms. By integrating thermography and AI technology into an accessible mobile platform, this project aims to bridge diagnostic access gaps in low-resource environments while informing scalable solutions with direct relevance to improving dermatologic care in rural and underserved regions of the United States and broader global application. Rigorous evaluation and validation will position this technology for wide adoption and sustained impact.
