No abstract

The machine learning algorithms underperformed with respect to the experienced radiologist's readings used to construct them, and the RBF-NN outperformed the other machine learning algorithm models.

No abstract

BackgroundAlthough microRNAs offer great potential as cancer biomarkers, effective clinical dignostics and tumor maker have not been verified to diagnose with colorectal cancer (CRC). The purpose of our study is to systematically assess the expression of miRNAs in matched cancer and normal tissue samples to identify promising diagnostic microRNA (miRNA) biomarkers for CRC.MethodsIn our study, we examined by Real-Time PCR the expression levels of 96 mature miRNA in 32 CRC patients with differently expressed tumors versus normal colon tissues. Using enter and stepwise variable selection methods separately, conditional logistic regression was conducted to identify miRNAs associated with CRC. The classification performance of these indicators was assessed under the Fisher discriminant analysis. Receiver operating characteristic curve analyses were applied to obtain diagnostic utility of the differentially expressed miRNAs.ResultsIn this study, we confirmed 11 overexpressed miRNAs with no less than twofold difference, and 85 downexpressed miRNAs with up to 0.5-fold difference in CRC from 96 aberrantly expressed miRNAs being identified by real-time PCR. Conditional logistic regression results confirmed that miRNA-378 and miRNA-145 expression profile was statistically significant. The error diagnosis rate of these two miRNAs are 0.194 and 0.113, separeately, showing by discriminant analysis.ConclusionsMiRNA-145 and miRNA-378* are potential biomarkers for early detection of CRC, which may help in diagnosing CRC in early period.

No abstract