Abstract 141P
Background
It is crucial for clinical decision-making to identify non-small cell lung cancer (NSCLC) patients who are likely to achieve major pathological remission (MPR) following neoadjuvant immunotherapy and chemotherapy (NICT).
Methods
Our study involved an analysis of 243 NSCLC patients from three centers, treated with NICT and surgery and categorized into training, validation and test cohorts. We conducted an extensive analysis of the tumor area, examining the intra-tumoral zone and the surrounding peri-tumoral regions at 2 mm, 4 mm and 6 mm, developing an algorithm for delineating tumor habitats. Features were standardized with Z-scores and de-duplicated by retaining one from each highly correlated pair. We finalized the feature set using Least Absolute Shrinkage and Selection Operator (LASSO) regression and 10-fold cross-validation, forming a robust radiomic signature for machine learning models. Clinical features underwent univariable and multivariable analyses, combining with peri-tumoral and habitat signatures in a nomogram, whose diagnostic accuracy and clinical utility were evaluated using receiver operating characteristic (ROC), calibration curves and decision curve analysis (DCA).
Results
The cohort showed a 68% MPR rate, with histology identified as a key predictor. An integrated nomogram including histology, Peri6mm and habitat features outperformed individual models with an AUC of 0.894 in the training cohort, 0.831 in validation and 0.799 in testing. The nomogram demonstrated a clear advantage in predictive probabilities, as evidenced by DCA curve results.
Conclusions
Our study's development of a predictive model using a nomogram integrating clinical and radiomic features significantly improved MPR prediction in NSCLC patients undergoing NICT, enhancing clinical decision-making.
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.