Pressure Ulcer Stage Evaluation Algorithm Based on Multi-Scale Attention Residual Network

Pressure ulcer, alternatively referred to as a pressure injury, mostly occurs in long-term bedridden patients. This issue has emerged as a significant global health concern. Accurate diagnosis is the key to the treatment of pressure sores in clinical application scenarios. Conventional contact-based artificial pressure ulcer detection methods can cause discomfort to patients and potentially elevate the risk of infection. In response to these problems, a pressure ulcer stage evaluation algorithm based on multi-scale attention residual network was proposed. The residual network was used as the basic framework, the stage 0 was improved to enhance the network’s nonlinear expressive capacity, and the residual structure was improved to enhanced the model’s generalization capability, the FPN feature pyramid was added to fuse the high-level features with the underlying features to achieve feature pyramid feature extraction. The incorporation of the attention mechanism enhanced the model’s overall performance and recognition accuracy. The pressure damage stage dataset was used in an experiment to validate the model’s performance. The experimental outcomes demonstrate that the proposed method could achieve the precision, recall, F1-score and accuracy of the algorithm of $82.64 \%, 83.26 \%, 0.83$ and 82.37% respectively. Compared with other deep learning models, this proposed algorithm can improve the accuracy and identify the pressure ulcer effectively.