Using Textural Analysis of Thermal Imaging to Predict Healing Status of Diabetes-Related Neuropathic Foot Ulcers: Protocol for a Co-Design and Longitudinal Study.

Background: Diabetes-related foot ulcers (DFUs) are common in people with diabetes and a major cause of poor quality of life and disability. If not treated in a timely and appropriate way, DFUs may result in prolonged hospitalization and amputation. Currently, methods to predict the healing trajectory of DFUs lack accuracy. Thermal imaging has been proposed to overcome these limitations but has been unable to accurately predict delayed healing of DFUs in the early stages of ulcer management. This project aims to ascertain whether textural analysis of a thermal image can predict the healing trajectory of DFUs. Objective: The study aims (1) to co-design an accurate, fast, easy-to-use, computer-aided, nontouch test to predict DFU healing trajectory using texture analysis of thermal imaging that is fit for purpose and acceptable to both, those being tested and users of the device and (2) to validate whether textural analysis of thermal images can accurately predict healing of DFUs at week 12 from an image taken at week 1. Methods: This project will be undertaken in 2 phases: Phase 1, co-design and development of the software prototype; and Phase 2, technology validation. Phase 1 requires a participatory action, co-design approach, engaging clinicians and biomedical engineers in a facilitated focus group and email communication. Interviews with adults living with diabetes and a DFU or who have a history of a DFU will be undertaken to understand their information needs about the device and its findings. Phase 2 will be a longitudinal observational study of 120 adults living with a DFU over a 12-week period. Demographic and other data that have been shown to impact wound healing will be collected at baseline, including participant age, gender, wound size, wound duration, and biomedical markers. Thermal and standard red, green, and blue images will be taken at weeks 1, 2, and 12. Wound textural features will be entered into a Bayesian neural network to identify the healing trajectory. Results: Phase 1 has been completed with biomedical engineers and clinicians from High Risk Foot Services at 2 hospitals in Melbourne, Australia. Nine clinicians participated in a co-design focus group, and 6 clinicians communicated via email. Four themes were identified: insights for sector use, how to document and use the device, legal implications, and workflow requirements. Interviews with 4 people living with diabetes were undertaken, with data to be analyzed thematically. Phase 2 data collection will be completed by April 2026. Conclusions: The study aims to co-design, test, and validate an accurate, feasible, and acceptable device to predict the healing of DFUs at week 12 from an image taken at week 1. This has the potential to assist clinicians in making informed and timely decisions for instigating adjuvant therapies, thereby improving healing and preventing lower extremity amputations.