DAM-UNET: Dual Attention Module Based UNET for Lung Tumor Segmentation Using CT Scans

The increasing incidence of lung cancer demands urgent attention from medical scientists. Advanced deep learning-based methods have demonstrated promising capabilities in detecting even small lung tumors, potentially improving early diagnosis rates. This study presents a novel approach for lung tu-mor segmentation using CT scans using deep learning techniques. We introduce a Dual Attention Module UNET (DAM-UNET) architecture that incorporates spatial and channel attention mechanisms to improve segmentation accuracy. The performance of our proposed network is compared with well-known segmentation models including UNET, SegResNet, SegNet, and VNET validated using the Medical Segmentation Decathlon (MSD) dataset. Our DAM-UNET architecture consistently outperforms other models across multiple evaluation metrics i.e., IoU (86.32% ), DSC (82.82%), Recall (91.56%), F1-score (92.66%), Sensitivity (91.56%), and Specificity (99.97%). Our findings suggest that DAM-UNET is a reliable tool for accurate and robust lung tumor segmentation in clinical applications.