ADTNet: Attention-Guided U-Net with Dynamic CNN and Transformers for Skin Cancer Detection
Precise segmentation of skin cancer lesions plays a vital role in computer-aided diagnostic systems. This work introduces ADTNet, an architecture for skin cancer segmentation that integrates dynamic CNNs with transformers using a U-Net framework, adding attention-driven skip connections and transformer modules in the bottleneck. Evaluated on the ISIC 2018 dataset, the 4-stage version of ADTNet achieved a Dice score of 92.4% and IoU of 87.4%, balancing segmentation accuracy with computational efficiency and performing on par with state-of-the-art techniques.
Recommended citation: L. Sarker, et al. (2024). “ADTNet: Attention-Guided U-Net with Dynamic CNN and Transformers for Skin Cancer Detection.” IEEE ICECE 2024. https://doi.org/10.1109/ICECE64886.2024.11024876