WCLC 2025 - Posters & ePosters-P1.07.42 Multi-Temporal Imaging and CT-To PET Translation Based Predictive Model on Neoadjuvant Immunotherapy for Non-Small Cell Lung Cancer Patients

P1.07.42 Multi-Temporal Imaging and CT-To PET Translation Based Predictive Model on Neoadjuvant Immunotherapy for Non-Small Cell Lung Cancer Patients

Back to course

Pdf Summary

This study developed a multimodal delta radiomics predictive model to assess neoadjuvant immunotherapy (NIT) efficacy in non-small cell lung cancer (NSCLC) patients, addressing the challenge of tumor spatiotemporal heterogeneity (TSH) in patient selection. Conducted as a single-center retrospective study with 149 patients at Guangdong Provincial People's Hospital, the research integrated CT, PET, and metabolic imaging data, focusing on changes (delta radiomics) between pre-treatment and pre-surgical scans. PET imaging usually offers valuable metabolic information but is limited by cost and complexity; thus, a previously validated CT-to-PET translation (CPT) model was used to supplement PET data synthetically.<br /><br />Results showed that combining delta features from CT and PET produced the highest predictive accuracy for therapeutic response, with an area under the curve (AUC) of 0.855, outperforming models that also included metabolic features and all single-modality models. Among single modalities, delta CT features had better predictive power than PET or metabolic features alone. Models integrating baseline and delta metabolic features also showed good performance but were generally inferior to the CT and PET combination models.<br /><br />The study concludes that integrating multimodal imaging delta radiomics, especially combining CT and PET data, offers a promising approach for predicting NIT efficacy in NSCLC. Future research directions include external validation using multicenter prospective samples and enhancing synthetic PET data to strengthen the predictive framework's clinical applicability. This work supports more personalized treatment decisions in NSCLC by non-invasively capturing dynamic tumor changes during immunotherapy.<br /><br />Key referenced studies underpin the utility of synthetic PET from CT data and the value of radiomic models in predicting pathological response to immunotherapy, indicating this approach's growing relevance in lung cancer care.

Asset Subtitle

Minjian Li

Meta Tag

Speaker Minjian Li

Topic Early-Stage Non-small Cell Lung Cancer

Keywords

multimodal delta radiomics

neoadjuvant immunotherapy

non-small cell lung cancer

tumor spatiotemporal heterogeneity

CT imaging

PET imaging

synthetic PET data

predictive model

treatment response prediction

personalized lung cancer therapy