A Deep Multi-Modal Method for Patient Wound Healing Assessment

TL;DR

The paper tackles high wound-care costs by predicting hospitalization risk using a deep multi-modal approach that leverages wound images and clinician data. It employs transfer learning with a pre-trained Xception network to predict five wound attributes from images and then fuses these outputs with 16 clinician features in a LightGBM classifier to predict healing outcomes. A wound image dataset annotated for all wound variables across 20 ulcer types is introduced, with a two-step training regime and 70/10/20 splits plus 5-fold cross-validation. Results demonstrate meaningful attribute prediction and strong risk discrimination, including hospitalization precision 0.68, recall 0.91, and treatment-complete precision 0.99, recall 0.79, complemented by attention heatmaps highlighting relevant wound regions. Overall, the method offers a data-driven path to early complication detection and potential reductions in clinician time spent on wound assessment.

Abstract

Hospitalization of patients is one of the major factors for high wound care costs. Most patients do not acquire a wound which needs immediate hospitalization. However, due to factors such as delay in treatment, patient's non-compliance or existing co-morbid conditions, an injury can deteriorate and ultimately lead to patient hospitalization. In this paper, we propose a deep multi-modal method to predict the patient's risk of hospitalization. Our goal is to predict the risk confidently by collectively using the wound variables and wound images of the patient. Existing works in this domain have mainly focused on healing trajectories based on distinct wound types. We developed a transfer learning-based wound assessment solution, which can predict both wound variables from wound images and their healing trajectories, which is our primary contribution. We argue that the development of a novel model can help in early detection of the complexities in the wound, which might affect the healing process and also reduce the time spent by a clinician to diagnose the wound.

Figures (2)

• Figure 1: Manually labeled wound ulcer type images
• Figure 2: Attention Heatmaps generated by CNN