Narumi Taguchi, et al. - Through a retrospective study that enrolled 40 subjects with pathologically verified colorectal cancer who underwent Kirsten rat sarcoma viral oncogene homolog (KRAS) mutation testing, contrast-enhancement computed tomography (CT), and 18F-fluorodeoxyglucose positron emission tomography prior to the treatment, the researchers assessed if a machine learning-based CT texture analysis could foretell the mutation status of V-Ki-ras2 KRAS in colorectal cancer. Twenty individuals out of 40 had mutated KRAS genes, whereas 20 had wild-type KRAS genes. The area under the curve (AUC) of each CT texture parameter ranged from 0.4 to 0.7 and the AUC of the standard uptake values (SUV_max) was 0.58, in the univariate analyses. In comparison to the SUV_max, the multivariate support vector machine with comprehensive CT texture parameters generated an AUC of 0.82, symbolizing a better prediction performance. Hence, for foretelling the KRAS mutation status of colorectal cancer, a machine learning-based CT texture analysis was better to the SUV_max.