Benchmarking Deep Learning for NSCLC PET/CT Segmentation on a Histologically Confirmed Vietnamese Dataset: Validation and Generalization.
Authors
Affiliations (8)
Affiliations (8)
- Institute for Nuclear Science and Technology, Hanoi 100000, Vietnam.
- Faculty of Engineering Physics, Hanoi University of Science and Technology, Hanoi 100000, Vietnam.
- Department of Radiation Oncology and Radiosurgery, 108 Military Central Hospital, Hanoi 100000, Vietnam.
- Department of Medical Imaging Technology, Hanoi Medical College, Hanoi 100000, Vietnam.
- Hong Ngoc-Phuc Truong Minh General Hospital, Hanoi 100000, Vietnam.
- Faculty of English, Thuong Mai University, Hanoi 100000, Vietnam.
- Nuclear Medicine Department, HCM Oncology Hospital, Ho Chi Minh City 700000, Vietnam.
- Viettel Institute for Strategic Technology, Viettel Group, Hanoi 100000, Vietnam.
Abstract
Accurate segmentation of non-small cell lung cancer (NSCLC) on positron emission tomography/computed tomography (PET/CT) is an essential prerequisite for automated metabolic tumor volume (MTV) quantification and staging. Although deep learning models achieve high performance on large-scale datasets, their generalization across different clinical domains is limited by variations in imaging protocols and patient demographics. This study aims to evaluate several deep learning architectures and investigate a transfer learning strategy to mitigate domain shift. Three architectures, including ResNet-backbone 3D U-Net, nnU-Net v2, and Swin UNETR, were benchmarked from scratch and compared with a fine-tuned nnU-Net initialized with AutoPET II weights. Results on the internal dataset showed that the fine-tuned nnU-Net achieved a Dice similarity coefficient (DSC) of 83.4 ± 6.5%, a 95% Hausdorff distance (HD95) of 5.1 ± 3.6 mm, and a precision of 89.6 ± 8.2%. Compared to the nnU-Net v2, the fine-tuned nnU-Net improved the absolute DSC by 6.8% while reducing local training time by 37.5% by bypassing the initial feature-learning phase. The fine-tuned nnU-Net model also demonstrated a high correlation between the MTV and the ground truth (Pearson <i>r</i> = 0.96, <i>p</i> < 0.001), indicating its potential as a reliable automated approach for quantitative MTV extraction and NSCLC prognostic-related analysis.