To Conduct Multi-omics Integrated Studies in Peripheral Blood, Such as Fragment Omics, Metabolomics and Epigenetics, and Establish Non-invasive Dynamic Follow-up Monitoring Programs During Perioperative and Postoperative Periods (Observational Study)

This project is a prospective observational study designed to comprehensively integrate plasma metabolomic, radiomic, and epigenomic data to develop a predictive model for postoperative recurrence risk in lung cancer. The study will retrospectively enroll 200 patients who underwent radical surgery after neoadjuvant therapy, and prospectively enroll 100 additional post-radical-surgery lung cancer patients who received neoadjuvant treatment as a validation cohort. Peripheral blood samples will be collected at multiple timepoints for metabolomic profiling. Unsupervised clustering, random forest algorithms, and Wilcoxon tests will be applied to identify recurrence-related features and construct a recurrence prediction model.Additionally, using preoperative and first postoperative follow-up CT imaging data, a deep learning-based 3D ResNet will be employed to generate radiomic recurrence risk scores for each patient. Plasma cfDNA will undergo low-pass whole-genome sequencing and methylation analysis to extract multi-dimensional recurrence-associated features. Finally, the study will innovatively utilize the DeepProg deep learning framework to integrate radiomic, cfDNA, and plasma metabolomic data into a non-invasive multi-omics model. Combined with survival data, this model will predict recurrence risk, ultimately achieving high-accuracy stratification of patients' postoperative recurrence probability.