Advertisement

PREDICTING HEALING OF DFUS WITH MULTISPECTRAL IMAGING

Guidelines recommend adjunctive therapies for diabetic foot ulcers (DFUs) without >50% wound area reduction after 4 weeks of standard care. This delays advanced care in nearly 50% of DFUs. We sought to evaluate the ability of a multispectral imaging to predict DFU healing potential at the initial assessment.

Methods: Patients with a new DFU were enrolled in this multicenter, prospective trial after providing informed consent. Multispectral images of the DFU were obtained at initial evaluation using a novel wound imaging system. Clinicians were blinded to imaging data. A standardized DFU healing assessment was performed by physical exam after 30 days of standard care to determine whether ≥50% wound area reduction was achieved. A machine learning algorithm analyzing multispectral imaging features was trained and tested using cross-validation to measure the imaging system’s accuracy of predicting DFU healing potential at the initial evaluation.

Results: Thirty-six patients (mean age 64±11 years; 84% male) with 48 DFUs were enrolled. Mean HbA1c was 7.6% ± 1.6%; 34 (66%) patients were prescribed insulin, 17 (47%) had prior lower extremity amputation, 7 (19%) had prior lower extremity revascularization, and 26 (72%) were current/former tobacco users. Standardized wound healing assessment identified 32/48 (67%) wounds that failed to achieve ≥50% wound area reduction. Using cross-validation, the overall accuracy for prediction of wound healing potential (likelihood wound would demonstrate ≥50% wound area reduction after 30 days of standard wound care) of the multispectral wound imaging system was 91.7% (95%CI: 83.3– 100%), with 96.9% (95%CI: 89.7–100%) sensitivity and 81.3% (95%CI: 61.5–100%) specificity; area under the curve (AUC) was 0.94, suggesting excellent performance of the classifier.

Conclusions: This study suggests that this novel wound imaging system has the potential to predict DFU healing potential upon initial assessment with a very high degree of accuracy. This wound imaging system would in turn allow clinicians to initiate additional diagnostics, specialist referrals, and advanced wound therapies in appropriately selected patients more quickly, with the potential to save costs and return patients to a higher quality of life with greater expediency. Independent validation of the wound imaging system is warranted.

Source: Bastawros D, Squiers J, Thatcher J, et al. Machine Learning Analysis of Multispectral Imaging Data to Predict Healing Potential of Diabetic Foot Ulcers with Standard Therapy. Available at https://www.dfcon.com/posters/. Publication is pending in the Journal of Vascular Surgery.

Advertisement