Specifically, this project will: 1) construct dynamic prediction models based on the NLST dataset, integrating multi-timepoint CT image information to improve predictive accuracy; 2) validate the model's generalizability in the Chongqing Cancer Screening Cohort, assessing its applicability to the Chinese population; and 3) compare the effectiveness of single versus multiple LDCT screenings to determine optimal screening intervals for different risk groups. By building upon the foundational NLST RCT, this research will contribute significantly to optimizing lung cancer screening strategies, improving screening efficiency, and reducing healthcare costs.
1）Develop a Dynamic Prediction Model Based on the NLST Landmark RCT Dataset
Objective: Utilize multi-timepoint CT image data from the NLST dataset, a milestone RCT in lung cancer screening, to develop a dynamic prediction model, enhancing the accuracy of lung cancer risk prediction.
Methods:
a) Extract CT image features from consecutive timepoints in the NLST dataset.
b) Design and train deep learning models that incorporate temporal information.
c) Evaluate model performance, including sensitivity, specificity, and AUC.
2）Conduct External Validation in the Chinese Chongqing Cancer Screening Cohort
Objective: Assess the applicability of the NLST-based dynamic prediction model in the Chinese population, bridging the gap between this landmark Western RCT and Asian populations.
Methods:
a) Collect multi-timepoint CT data from the Chongqing Cancer Screening Cohort.
b) Apply the NLST model to predict lung cancer risk in the Chongqing cohort.
c) Compare model performance between U.S. and Chinese populations, analyzing reasons for any differences.
3）Compare Single versus Multiple LDCT Screenings to Optimize Screening Intervals for Different Risk Groups
Objective: Determine the optimal LDCT screening frequency for different risk groups, extending the insights from the NLST's annual screening protocol.
Methods:
a) Use the dynamic prediction model to simulate the impact of different screening frequencies (e.g., annual, biennial) on lung cancer detection rates.
b) Conduct cost-effectiveness analyses to evaluate the economic impact of different screening strategies.
c) Develop personalized screening recommendations for different populations based on risk stratification.

Shenglin Zhao, Office of Chongqing Cancer Prevention and Treatment, Chongqing University Cancer Hospital, Chongqing 400030, China.
Bibo Li, Department of Oncology, Chongqing Academy of Medical Sciences & Chongqing General Hospital. Address: No.118, Xingguang Avenue, Liangjiang New Area, Chongqing, 401147, China.
Mei He,Office of Chongqing Cancer Prevention and Treatment, Chongqing University Cancer Hospital, Chongqing 400030, China.