Automated Pressure Injury Assessment and Documentation Generation Using Vision-Language Model.
Journal:
Studies in health technology and informatics
Published Date:
Aug 7, 2025
Abstract
Pressure injury assessment and documentation are crucial but time-consuming tasks in healthcare settings, with current inter-rater reliability among assessors only reaching 60-70%. This study presents an automated approach using the Florence-2 vision-language model for pressure injury assessment and clinical description generation. The model was trained on 946 pressure injury images, augmented to 275 images per grade through various transformations. Results demonstrate robust performance across pressure injury grades with F1 scores ranging from 73.17% to 95.00%, and strong capability in generating standardized clinical descriptions (BERTScore: 85.58%). Despite challenges in distinguishing between Stage 3 and 4 injuries, the integrated approach shows potential for improving assessment consistency and documentation efficiency in clinical settings. This solution addresses the critical needs for standardization and efficiency in pressure injury documentation while maintaining clinical accuracy.
