Unimodal and Multimodal Deep Learning for Pressure Injury Identification: Scoping Review.

BACKGROUND

Pressure injury (PI) can cause severe infections or death, and has a high global prevalence. Accurate staging is vital for effective intervention. Deep learning streamlines PI assessment; enhances efficiency; and yields practical, accurate results.

OBJECTIVE

The primary objective of this scoping review was to map the landscape of deep learning applications for PI identification and categorize the existing body of evidence into distinct technological trajectories: unimodal imaging and multimodal integration.

METHODS

We searched the following databases and sources: PubMed, the Cochrane Library, IEEE Xplore, and Web of Science. This scoping review was conducted in accordance with the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines.

RESULTS

A total of 15 articles were included: 40% (n=6) studies applied multimodal integration, whereas 60% (n=9) applied unimodal imaging. In total, 26 models were involved. Different models exhibited varying accuracy rates in staging PI, with overall accuracy fluctuating between 54.8% and 93.7%. The same model demonstrated significant variations in recognition accuracy across different studies. Multimodal integration appeared to be a distinguishing characteristic of the studies that will progress toward clinical validation.

CONCLUSIONS

The current landscape of deep learning in PI identification is heavily skewed toward unimodal imaging, creating a modality gap that hinders clinical translation. While unimodal models demonstrate high accuracy in controlled settings, future research must prioritize bridging the gap from unimodal imaging to true multimodal integration to achieve clinical-grade reliability.