Yessica Nataliani, Hindriyanto Dwi Purnomo, Ivanna K. Timotius, Purwono Purwono
Diabetic Foot Ulcer (DFU) represents a severe complication of diabetes, often resulting in lower-limb amputation if not diagnosed and treated promptly. Accurate and efficient segmentation of DFU is essential for early detection and timely intervention, particularly in resource-limited settings with limited access to medical professionals. However, conventional U-Net-based segmentation models are computationally intensive, limiting their deployment in mobile and real-time clinical applications.This study introduces DFU-MambaLiteUNet, an efficient and lightweight segmentation model that integrates the Mamba State Space Model (SSM) Lite into a Mobile U-Net framework. The model employs Attention Gate modules to enhance feature selectivity, GeLU activation to improve training stability, and Tversky Loss to effectively handle class imbalance in segmentation tasks. To the best of our knowledge, this is the first work to explore the application of Mamba SSM Lite for DFU segmentation, offering a balanced trade-off between segmentation accuracy and computational efficiency. Comprehensive evaluations were conducted on three datasets: FUSC 2021, DFUC 2022, and a newly constructed primary clinical dataset. The proposed model achieved a Dice Coefficient of 92.23% on FUSC 2021, representing a 5.13% improvement over the baseline, along with a 17.02% reduction in GFLOPs and a 33.9% faster inference time. These findings demonstrate the strong potential of DFU-MambaLiteUNet as a lightweight and efficient model for diabetic foot ulcer segmentation, capable of delivering accurate and real-time performance in clinical and mobile healthcare environments, especially in resource-constrained settings. © 2025 Elsevier Ltd
Faculty of Information Technology, Satya Wacana Christian University, Salatiga, 50715, Indonesia; Department of Electronics Engineering, Satya Wacana Christian University, Salatiga, 50711, Indonesia; Faculty of Science and Technology, Universitas Harapan Bangsa, Purwokerto, 53144, Indonesia
Research at a Glance
Register to unlock
Topics & SDG Alignment
Register to unlock
Collaboration
Register to unlock
Author Profile (Selected)
Register to unlock
References Overview
Register to unlock
Journal & Source
Register to unlock
Metadata & Integrity
Register to unlock