Dense-ShuffleGCANet: An Attention-Driven Deep Learning Approach for Diabetic Foot Ulcer Classification Using Refined Spatio-Dimensional Features

Diabetic foot ulcers (DFU) are a common and serious complication of diabetes, often leading to severe health implications like limb amputation if left untreated. Timely intervention and treatment are crucial in mitigating the impact of DFU on patients' well-being. However, manual identification of DFUs remains a challenge due to their heterogeneous visual characteristics, leading to several undiagnosed cases and avoidable complications. Deep learning provides an efficient solution to the challenge by automating the process of DFU detection, thereby alleviating the burden on the healthcare industry. This research introduces a novel model, Dense-ShuffleGCANet, for DFU detection by leveraging DenseNet-169, Channel-Centric Depth-wise Group Shuffle (CCDGS) block, and triplet attention. The architecture of DenseNet-169 is modified to integrate the CCDGS block and triplet attention, enabling the model to efficiently capture long-range dependencies and utilise cross-channel features. These improvements enable the effective extraction of both channel-wise and spatial features. The proposed Dense-ShuffleGCANet model achieved an accuracy of 86.09% and an F1-score of 85.77% on the DFUC2021 dataset, outperforming state-of-the-art architectures.