EPI-Sense: an integrated hemodynamic sensing and analysis system for early pressure injury warning

Pressure injury poses a critical challenge in geriatric care. Current computer-aided diagnosis is limited by an exclusive reliance on surface texture, neglecting subcutaneous hemodynamics essential for distinguishing Reactive Hyperemia from Stage 1 PI and detecting Deep Tissue Pressure Injuries. To address this, we constructed a customized cross-polarized acquisition system and a dedicated clinical dataset.

We present the Physio-Guided Network (PGNet), a multi-modal framework leveraging remote Photoplethysmography (rPPG) to provide complementary non-contact hemodynamic cues related to local microcirculatory dynamics. Specifically, we engineered a Frequency-Domain SNR-Weighted strategy to retrieve attenuated physiological signals from necrotic tissue, and a Physio-Visual Gating Unit for dynamic modality arbitration.

Patient-level leave-one-out validation supports the feasibility of the proposed framework for differentiating morphologically similar early pressure injury states. The results indicate that rPPG-driven hemodynamic guidance provides complementary information beyond visual-only analysis.

By bridging superficial morphology with non-contact hemodynamic sensing, this study demonstrates the clinical relevance and technical feasibility of rPPG-guided multimodal analysis for early pressure injury warning, providing patient-level evidence that supports its potential value for differentiating clinically ambiguous RH, S1, and DTPI cases and laying the groundwork for future multicenter, device-independent validation.