Hierarchical Meta-Heuristic Encoder-Decoder Architecture With Next-Generation Imaging for Consumer-Centric Segmentation of Diabetic Foot Ulcers

Diabetic foot ulcers represent a pressing global health concern, demanding precise and efficient segmentation for timely intervention and improved patient outcomes. This work proposed an approach named Hierarchical Meta-Heuristic Encoder-Decoder Architecture (HM-EDA) integrated with Next-Generating Imaging (NGI) for consumer-centric segmentation of Diabetic Foot Ulcers (DFUs). The proposed model is designed to address the challenges posed by the intricate nature of foot ulcer characteristics, leveraging the strengths of both hierarchical feature extraction and evolutionary optimization within the DenseNet architecture. HM-EDA is structured in a hierarchical manner, allowing for the extraction of multi-scale features from the input data. Additionally, an evolutionary optimization mechanism is embedded within the DenseNet framework to adaptively fine-tune model parameters, enhancing the overall performance and generalization capabilities. The internal architecture of the model is built up by using the cascading of the plain and reduced cell blocks with skip connections. We validate the proposed HM-EDA-NGI model using an extensive dataset comprising varied DFU cases, demonstrating its superior segmentation accuracy of 97.25%, Dice score as 0.95, and IoU as 0.94, compared to existing methodologies.