ABSTRACT
Lung cancer as the leading cause of cancer related deaths, the diagnosis and prognostic analysis of lung cancer can assist clinical decision making for large amount of radiologists. The radiomic analysis of lung cancer aims at mining tumor information from CT image to provide a non-invasive and pre-treatment prediction of clinical outcomes in lung cancer. The first step of radiomic analysis requires tumor area segmentation in CT image, which can be annotated manually or by semi-automatic algorithms. Afterwards, radiomic features are extracted inside tumor areas reflecting the intensity, shape, texture and high-dimensional tumor information. Finally, important features are selected to build a radiomic model for diagnostic or prognostic prediction. The published studies demonstrated the diagnostic power of radiomic model in identifying malignant tumors and tumor stage classification. Moreover, the radiomic phenotyping also revealed strong association to genetic profiles such as the EGFR and KRAS mutation status.
