JMIR Publications

ABSTRACT

Background:

Pressure ulcers (PU) and incontinence-associated dermatitis (IAD) are prevalent conditions in clinical settings, posing significant challenges due to their similar presentations but differing treatment needs. Accurate differentiation between PU and IAD is essential for appropriate patient care, yet it remains a burden for nursing staff and wound care experts.

Objective:

This study aims to develop and introduce a robust multimodal deep learning framework for the classification of pressure ulcers and incontinence-associated dermatitis, along with the fine-grained categorization of their respective wound severities, to enhance diagnostic accuracy and support clinical decision-making.

Methods:

We collected and annotated a dataset of 1,555 wound images, achieving consensus among 4 wound experts. Our framework integrates wound images with categorical patient data to improve classification performance. We evaluated four models—two convolutional neural networks and two transformer-based architectures—each with approximately 25 million parameters. Various data preprocessing strategies, augmentation techniques, training methods (including multimodal data integration, wound cropping, and sampling), and post-processing approaches (including ensembling and test-time augmentation) were systematically tested to optimize model performance.

Results:

The transformer-based TinyViT model achieved the highest performance in binary classification of PU and IAD, with an F1-score of 92.62%, outperforming wound care experts and nursing staff on the test dataset. In fine-grained classification of wound categories, the TinyViT model also performed best for PU categories with an F1-score of 68.07%, while ConvNeXtV2 showed superior performance in IAD category classification with an F1-score of 50.76%. Incorporating multimodal data improved performance in binary classification but had less impact on fine-grained categorization. Augmentation strategies and training techniques significantly influenced model performance, with ensembling enhancing accuracy across all tasks.

Conclusions:

Our multimodal deep learning framework effectively differentiates between pressure ulcers and incontinence-associated dermatitis, achieving high accuracy. By integrating wound images with patient data, the framework enhances diagnostic precision, demonstrating significant potential to support healthcare providers in clinical decision- making. However, the classification of precise PU and IAD categories remains a challenging task due to subtle visual differences and class imbalance. This advancement can aid clinicians and nursing staff in accurately identifying wound types and severities, ultimately improving patient care and outcomes in wound management.