Automated Wound Image Segmentation: Transfer Learning from Human to Pet via Active Semi-Supervised Learning

SIMPLE SUMMARY: Appropriate wound management shortens healing times and reduces management costs, benefiting the patient in physical terms and potentially reducing the healthcare system economic burden. Artificial intelligence techniques could be used to automate the process of wound healing assessment, easing the effort required by clinicians and removing the inherent subjectivity of the evaluation. However, the training of artificial intelligence models relies on the availability of large datasets of carefully annotated data. The annotation of medical data is a time consuming and expensive process which requires the supervision of high-expertise professionals. In this work, we introduced a novel pipeline for the segmentation of pet wound images, using an advanced training strategy able to minimize human intervention for both the image annotation and wound segmentation. We implemented our solution in a novel mobile app, providing a valuable tool for pet wound treatment and a methodological approach for the generation of large image-segmentation datasets. ABSTRACT: Wound management is a fundamental task in standard clinical practice. Automated solutions already exist for humans, but there is a lack of applications regarding wound management for pets. Precise and efficient wound assessment is helpful to improve diagnosis and to increase the effectiveness of treatment plans for chronic wounds. In this work, we introduced a novel pipeline for the segmentation of pet wound images. Starting from a model pre-trained on human-based wound images, we applied a combination of transfer learning (TL) and active semi-supervised learning (ASSL) to automatically label a large dataset. Additionally, we provided a guideline for future applications of TL+ASSL training strategy on image datasets. We compared the effectiveness of the proposed training strategy, monitoring the performance of an EfficientNet-b3 U-Net model against the lighter solution provided by a MobileNet-v2 U-Net model. We obtained 80% of correctly segmented images after five rounds of ASSL training. The EfficientNet-b3 U-Net model significantly outperformed the MobileNet-v2 one. We proved that the number of available samples is a key factor for the correct usage of ASSL training. The proposed approach is a viable solution to reduce the time required for the generation of a segmentation dataset.