• Citer

A Deep Learning Image-to-Image Translation Approach for a More Accessible Estimator of the Healing Time of Burns

par Rozo, Andrea ;Miskovic, Vanja ;Rose, Thomas;Keersebilck, Elkana;Iorio, Carlo Saverio ;Varon, Carolina
Référence IEEE transactions on biomedical engineering, 70, 10, page (2886-2894)
Publication Publié, 2023-04-17

Article révisé par les pairs

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

Objective: An accurate and timely diagnosis of burnseverity is critical to ensure a positive outcome. Laser Dopplerimaging (LDI) has become a very useful tool for this task. Itmeasures the perfusion of the burn and estimates its potentialhealing time. LDIs generate a 6-color palette image, with each colorrepresenting a healing time. This technique has very high costsassociated. In resource-limited areas, such as low- and middle-incomecountries or remote locations like space, where access tospecialized burn care is inadequate, more affordable and portabletools are required. This study proposes a novel image-to-imagetranslation approach to estimate burn healing times, using a digitalimage to approximate the LDI. Methods: This approach consistsof a U-net architecture with a VGG-based encoder and applies theconcept of ordinal classification. Paired digital and LDI images ofburns were collected. The performance was evaluated with 10-foldcross-validation, mean absolute error (MAE), and color distributiondifferences between the ground truth and the estimated LDI.Results: Results showed a satisfactory performance in terms oflow MAE (0.2370 ± 0.0086). However, the unbalanced distributionof colors in the data affects this performance. Significance: Thisnovel and unique approach serves as a basis for developing moreaccessible support tools in the burn care environment in resource-limitedareas.