Advertisement

THE ROLE OF RADIOGENOMICS IN EGFR AND KRAS MUTATION STATUS PREDICTION AMONG NON-SMALL CELL LUNG CANCER PATIENTS

      TYPE: Abstract Publication
      TOPIC: Biotechnology
      PURPOSE: The identification of cancer driver mutations allowed the development of targeted therapies, more effective with less side effects than chemotherapies. Epidermal growth factor receptor (EGFR) and Kirsten rat sarcoma viral oncogene (KRAS) are frequently mutated in non-small cell lung carcinoma (NSCLC). Although EGFR inhibitors are currently used, drugs for KRAS-mutated NSCLC are still under investigation. Tissue biopsy is the traditional way to assess the tumour molecular information. Radiogenomics could be helpful in this setting. We developed an artificial intelligence-based tool to correlate thoracic computed tomography (CT) features with genotype, to predict EGFR and KRAS mutation status.
      METHODS: A subset of NSCLC-Radiogenomics dataset (Bakr et al, 2017) including clinical, molecular and CT data was used. 158 (mutated: 41) and 157 (mutated: 40) were considered for EGFR and KRAS mutation status prediction, respectively. CT semantic features, annotated by radiologists, comprised the nodule and surrounding structures. Predictive models based on semantic features were developed using the Area Under Curve (AUC) of receiver operating characteristic (ROC) as the metric to assess their performance.
      RESULTS: There is a correlation between semantic features and EGFR mutation status (AUC = 0.75 ± 0.09) but not for KRAS (AUC = 0.50 ± 0.08). Both nodule and surrounding structures were relevant for EGFR prediction.
      CONCLUSIONS: Although relevant for EGFR-mutated NSCLC, no radiogenomic correlation was found for KRAS. In fact, radiogenomics in KRAS-driven NSCLC require further investigation since promising targeted therapies are currently under research.
      CLINICAL IMPLICATIONS: Radiogenomics may be an alternative to tissue biopsies regarding NSCLC molecular characterization.
      DISCLOSURE: No significant relationships.
      KEYWORDS: NON-SMALL CELL LUNG CANCER, artificial intelligence, cancer driver mutations