Noncontact Size Estimation of Pressure Ulcers Using IR Thermal Imaging

Pressure injuries cause discomfort and potential fatality, underscoring the importance of wound assessment. In the post-COVID era, remote monitoring of wounds, particularly through noncontact methods like infrared (IR) thermal imaging and deep learning, is imperative. This letter proposes a deep learning approach for dimension detection from thermal images, trained on data from 18 subjects. Instance segmentation achieved a maximum accuracy of 0.9542, with classification accuracy reaching 0.9922. The model exhibited a root mean square error (RMSE) of 0.1609 cm for measured dimensions, with superior accuracy in detecting wound length (RMSE: 0.1114 cm) compared to width (RMSE: 0.1506 cm).