Explainable Dual-Task Deep Learning Model for Recurrence Prediction and Dose-Sensitive Region Identification in Head
Huali Li1, Jiajun Cai2, Xuanru Zhou3
1School of Biomedical Engineering, Southern Medical University, Guangzhou, Guangdong, China.
Purpose:
Accurate prediction of distant metastasis (DM) and locoregional recurrence (LR) in patients with head and neck cancer (HNC) following radiation therapy is critical for individualized treatment planning. This study aimed to develop and validate an explainable deep learning model integrating anatomical and dosimetric information for HNC DM and LR risk stratification.
Methods And Materials:
Two hundred thirty-seven patients with HNC treated with definitive radiation therapy were collected and divided into training, internal, and external validation cohorts. A dual-task deep learning model based on 3D squeeze-and-excitation residual networks was constructed to predict DM and LR risk using computed tomography images, 3D dose distributions, and gross tumor volume (GTV) masks. Model performance was evaluated by concordance index, time-dependent receiver operating characteristic, and decision curve analysis. Interpretability was enhanced using gradient-weighted class activation mapping and a novel Activation Volume Histogram method to quantify attention patterns across spatial-dosimetric subregions.
Results:
The model achieved concordance indices of 0.92 and 0.74 for DM and LR, respectively, outperforming radiomic models. Gradient-weighted class activation mapping visualizations revealed distinct activation patterns aligned with recurrence sites-lymphatic regions for DM and tumor zones for LR. Activation Volume Histogram analysis identified GTV extended by 3-mm receiving dose ≥65 Gy and GTV receiving dose ≥65 Gy as the most discriminative subregions, showing significant differences in activation distributions between high- and low-risk groups, particularly within the 0.4 to 0.5 intensity range.
Conclusions:
This dual-task model enables accurate DM risk prediction and showed potential for LR risk prediction in HNC, providing spatial-dosimetric information to support risk-adaptive radiation therapy. Further extensive multicenter validation is warranted to confirm its clinical applicability.