Investigation of clinical target volume segmentation for whole breast irradiation using three-dimensional convolutional neural networks with gradient-weighted class activation mapping

This study aims to implement three-dimensional convolutional neural networks (3D-CNN) for clinical target volume (CTV) segmentation whole breast irradiation and investigate the focus of 3D-CNNs during decision-making using gradient-weighted class activation mapping (Grad-CAM). A 3D-UNet CNN was adopted conduct automatic CTV cancer. The trained three datasets left-, right-, both left- right-sided cancer patients. Segmentation accuracy evaluated Dice? similarity coefficient (DSC). Grad-CAM applied CNNs. DSCs breasts were on an average 0.88, 0.89, 0.85, respectively. heatmaps showed that used determined region from target-side tissue by referring opposite-side breast. Although size dataset limited, DSC ? 0.85 achieved 3D-UNet. indicates applicable scope limitations a indicating such decision-making.