A Region-Specific Dataset of Pressure Injury Photographs for Automated Staging and Clinical AI Applications

Background Accurate staging of pressure injuries is critical for clinical management but remains challenging because of inter-observer variability. Existing datasets often lack diversity in skin tones, limiting the generalizability of artificial intelligence (AI) models. Methods This Data Note describes a curated dataset of 267 clinically labeled pressure injury images collected from hospitalized patients in southeastern Iran between May 2023 and April 2024. Images were captured using standard mobile phone cameras under variable lighting, angle, and distance to reflect real-world clinical settings. Each image was labeled by a panel of three wound care specialists according to the National Pressure Injury Advisory Panel (NPIAP) staging system, covering Stages 1 to 4. Labeling was determined by majority consensus, and 38 images were excluded because of persistent disagreement among experts. The final dataset contains anonymized JPEG images with stage labels. Conclusions This dataset is intended to support the development and validation of deep learning models for automated pressure injury staging, particularly for populations with olive skin tones that are underrepresented in current literature. The dataset is available from Figshare under a Creative Commons Attribution 4.0 International license.