Predictive PETs.

Radiomics is an exciting concept. Using quantitative features from imaging studies to make predictions could be extremely useful, which is one reason for this retrospective evaluation of quantitative and qualitative features from PET/CT scans to attempt to identify EGFR-mutation status in 80 patients with stage II-III non-small cell lung cancer (NSCLC): 30 EGFR(+) and 50 EGFR(-) confirmed with pathologic verification. After selecting for the 35 best features (both quantitative and qualitative), the authors found that these features were significantly associated with EGFR-mutation status. A predictive model was created using PET/CT data, and its performance was 0.953 using area under the receiver operating characteristic curve. Non-statistician translation = the model was really good. TBL: Radiomics using PET/CT scan features may be highly accurate at determining EGFR mutation status in patients with NSCLC. | Jiang, Nucl Med Commun 2019