Physics-informed hyperspectral imaging of diabetic foot ulcers in Karshi: reconstruction and clinical validation
Abstract
Early identification of non-healing diabetic foot ulcers is crucial for preventing infection, hospitalisation and lower-limb amputation, yet bedside assessment remains largely subjective. This study investigates physics-informed hyperspectral imaging for quantitative evaluation of diabetic foot ulcers in patients treated at outpatient and inpatient facilities in Karshi, Uzbekistan. A custom hyperspectral camera spanning visible and near-infrared wavelengths was combined with a reconstruction framework that embeds light transport and tissue optics models into a learning-based algorithm. Ulcer regions were imaged serially during routine visits, yielding spectral cubes from which oxygenation, perfusion and tissue composition indices were derived. Healing trajectories over twelve weeks were compared with clinical outcomes adjudicated by a multidisciplinary team. The physics-informed method was benchmarked against purely data-driven reconstruction and standard colour photography. Physics-informed hyperspectral imaging improved prediction of delayed healing and infection, particularly in visually indeterminate wounds, and demonstrated feasibility for integration into multidisciplinary diabetic foot care in a resource-constrained Central Asian setting.