LAUSR

e14607 Background: Medical imaging techniques play a central role in the evaluation of lung cancer and proper images analysis is essential for determining the location and stage of tumors, for diagnosis and prognosis assessment, guiding therapeutic decisions and for monitoring tumor response during and after treatment. It also supports interventional radiology acts (i.e. biopsies, thermo-ablations or embolization). Yet, this source of information, which is critical for the decision-making process, is not being used to its full potential. Besides, NGS technics allow the precise characterization of the genomic profile of a tumor by detecting mutations, evaluating the TMB and providing additional information such as the PD-L1 status. Methods: We have used an AI-based strategy to improve the stratification of cohorts of patients suffering from lung cancer. Using an advanced segmentation technology, we were able to quickly and precisely extract 3D Radiomic characteristics from the images. We then combined extracted heterogeneity and texture indicators with the biology information thanks to a machine learning based methodology. Results: We have applied this strategy to stage 4 NSCLC with EGFR mutation treated with TKI and have shown that stratification of the cohort with respect to the OS can be delivered using the base-line CT-scan and the first two exams used for the evaluation of the early response to the treatment. The same strategy was developed for the monitoring of lung metastases in order to help scheduling thermo-ablation by providing a prediction of the growth of the lesion in oligo-metastatic cases. Conclusions: This study paves the way to a near future when Artificial Intelligence will provide precise evaluation of lung tumors and full integration of all the medical information available for each patient to support the decision making within the tumor board.