AI Model for Lung Cancer Detection and Classification Benchmarking Across Multiple CT Scan Datasets Description This repository contains the model weights for the AI-driven lung cancer detection and classification models developed as part of the study "AI in Lung Health: Benchmarking Detection and Diagnostic Models Across Multiple CT Scan Datasets." The models were trained and validated using a diverse set of CT scan datasets, including DLCSD, LUNA16, and NLST, to ensure robustness and generalizability in lung nodule detection and classification. Model Information Detection Models: DLCSD-mD: Trained on the Duke Lung Cancer Screening Dataset (DLCSD), designed to identify lung nodules within 3D CT volumes. LUNA16-mD: Trained on the LUNA16 dataset for lung nodule detection. Classification Models: ResNet50-SWS++: Enhanced ResNet50 using Strategic Warm-Start++ (SWS++) pretraining. Genesis & MedNet3D: Open-access self-supervised models. FMCB-based classifier: https://www.nature.com/articles/s42256-024-00807-9 Performance Summary Detection Task (AUC Scores): DLCSD-mD: 0.93 (DLCSD) | 0.97 (LUNA16) | 0.75 (NLST) LUNA16-mD: 0.96 (LUNA16) | 0.91 (DLCSD) | 0.71 (NLST) Classification Task (AUC Scores): ResNet50-SWS++: 0.71 (DLCSD) | 0.90 (LUNA16) | 0.81 (NLST) Dataset Information DLCSD: 2,000+ CT scans, 1,613 patients, over 3,000 annotated nodules. LUNA16: 601 patients, 1,186 nodules. NLST: 969 patients, 1,192 nodules (annotations adapted from external sources). Model Weights The uploaded model weights correspond to the best-performing checkpoints based on validation loss and AUC performance. The weights are provided in PyTorch (.pth) format and can be directly used with the MONAI framework. Code Repository The full codebase, including training scripts, preprocessing pipelines, and evaluation metrics, is available at Gitlab: https://gitlab.oit.duke.edu/cvit-public/ai_lung_health_benchmarkingGitHub: https://github.com/fitushar/AI-in-Lung-Health-Benchmarking-Detection-and-Diagnostic-Models-Across-Multiple-CT-Scan-Datasets/ Citation If you use this model in your research, please cite:Tushar et al., "AI in Lung Health: Benchmarking Detection and Diagnostic Models Across Multiple CT Scan Datasets," arXiv:2405.04605. @article{tushar2024ai, title={AI in Lung Health: Benchmarking Detection and Diagnostic Models Across Multiple CT Scan Datasets}, author={Tushar, Fakrul Islam and Wang, Avivah and Dahal, Lavsen and Harowicz, Michael R and Lafata, Kyle J and Tailor, Tina D and Lo, Joseph Y}, journal={arXiv preprint arXiv:2405.04605}, year={2024} }