Researchers in Australia have developed a smartphone-based system that combines thermal imaging with artificial intelligence to estimate whether a diabetes-related foot ulcer is likely to heal within twelve weeks. The aim is to identify wounds requiring additional treatment at the first assessment, rather than waiting several weeks to determine whether healing is progressing.
The tool, called THERMUL, was developed by researchers at the University of Melbourne, RMIT University and aged-care provider Bolton Clarke. According to the University of Melbourne, the system has been trained using images and clinical outcomes from 110 patients treated at Austin Health and St Vincent’s Hospital Melbourne.
Diabetes-related foot ulcers can lead to infection, hospitalisation and lower-limb amputation. More than 500 million people worldwide live with diabetes, and up to 34 per cent may develop a foot ulcer during their lifetime. Earlier recognition of wounds that are unlikely to heal could allow clinicians to intensify treatment or refer patients to specialist services sooner.
Heat patterns indicate healing trajectory
THERMUL uses a lightweight thermal camera connected to a smartphone. Researchers captured images when patients first attended the clinic and again during follow-up. After twelve weeks, clinicians assessed whether each wound had healed. These outcomes were used to train an algorithm to recognise thermal patterns associated with healing and delayed healing.
Rather than relying only on the average temperature of a wound, the system analyses the spatial distribution and texture of temperature variations. These patterns may reflect differences in inflammation, blood flow and tissue condition that are not readily visible during a conventional examination.
The published research protocol in JMIR Research Protocols describes how thermal features are extracted and assessed using different machine-learning models, including logistic regression and Bayesian neural networks. The prototype combines a thermal camera with a standard Android smartphone and was designed to require relatively little training from clinical users. Current wound assessment commonly involves monitoring changes in ulcer size during the first four weeks of treatment. THERMUL is intended to provide an earlier indication, potentially during the initial consultation. Its non-contact design could also make it suitable for outpatient clinics, community care and home visits.
Clinical value has not yet been established
The findings announced by the researchers do not yet show how accurately THERMUL distinguishes wounds that will heal from those that will not. No sensitivity, specificity or other performance measures from the 110-patient study have been made public, and a peer-reviewed paper reporting the clinical results has not yet been published.
The study population was also relatively narrow. According to the protocol, participants had neuropathic foot ulcers without infection or peripheral arterial disease. People with infected, ischaemic or neuro-ischaemic ulcers were excluded, even though these conditions are common among patients with more complex diabetes-related foot disease. The original protocol proposed recruiting 120 participants, while the subsequent university announcement reports that the study involved 110 patients. These factors mean the current findings cannot yet establish how reliably the system will perform across wider patient populations and clinical settings.
THERMUL should therefore be regarded as a technology under development rather than a clinically validated diagnostic tool. Demonstrating predictive performance is only one requirement. Researchers will also need to determine whether using the result changes treatment decisions, accelerates appropriate referrals or improves patient outcomes.
Wider validation in community care
The project has received AUD 490,539 through the Australian Economic Accelerator programme, in addition to AUD 50,000 from the University of Melbourne’s Proof of Concept Fund. The new funding will support validation in metropolitan, regional and remote healthcare services and the collection of data from more diverse population groups.
The Royal Flying Doctor Service Victoria and medical software company Software Medical Devices are among the organisations involved in the next phase. That wider testing is particularly relevant because patients in rural and remote communities may have limited access to multidisciplinary diabetes foot services.
A portable decision-support tool could help bring specialist assessment closer to patients, but its value will depend on the forthcoming validation results. The next stage must show not only whether THERMUL can classify wound-healing trajectories accurately, but also whether clinicians can use that information safely and effectively in routine care.
