Wound Semantic Segmentation Framework with Encoder-Decoder Architecture

Medical Image Segmentation Framework (MISEG) is proposed on this study to efficiently streamline the image segmentation challenge in the medical fields. While many existing study related to the medical image segmentation are focusing on the algorithms of accurate segmentation, we expanded our scope to integrate the life cycle of machine learning into medical image segmentation. Under the proposed MISEG framework, we leveraged the Encoder-Decoder model to participate the recent DFUC2022 foot ulcer image segmentation challenge. In this challenge, we have recorded an outstanding accuracy of 0.7501 in IoU score, despite only utilizing 80% of the training data.