Raspberry Pi-Based System for Tissue Classification and Measurement of Chronic Wounds Using Attention U-Net and TOF Sensor

Chronic wounds affect a significant portion of the global population, with healing often complicated by underlying health conditions and inconsistencies in clinical assessment. This challenge is especially evident in resource-limited environments where standardized procedures may be difficult to implement. In response, the study developed an assistive device utilizing RGB images processed by Attention U-Net segmentation and a Time-of-Flight (ToF) sensor for quantitative wound monitoring. The system analyzes captured images to determine dimensions (length, width, depth) and classify tissue composition percentages (epithelial, granulating, sloughy, necrotic). Experimental evaluation demonstrated optimal planar measurement performance, with average length and width errors of 17.31% and 17.79%, respectively, and an average depth measurement correlation of $R^{2}=0.8527$ across various distances. Tissue classification exhibited high reliability, achieving an F1-score of 1.00 for epithelial tissue and F1-scores of 0.79, 0.76, and 0.86 for granulating, sloughy, and necrotic tissues, respectively. Despite the identified distance trade-off for measurements, the system demonstrates significant potential to provide clinicians with standardized, data-driven insights, promoting more consistent and efficient wound assessments.