The Value of MRI Texture Analysis in Identifying Intraductal Carcinoma of the Prostate Gland

  Objective  To explore the value of a radiomics approach based on MRI texture analysis (TA) in identifying intraductal carcinoma of the prostate gland (IDCP).

  Methods  MRI images of 56 patients with pathological proven prostate cancer were analyzed retrospectively, including 31 patients with pure prostate adenocarcinoma and 25 patients with IDCP component in the prostate adenocarcinoma lesions. After imaging features of lesions were analyzed, then enhanced arterial and venous phase images were imported into Omni-Kinetics software for the extraction the TA features of region of interests’ lesion according to the T2-weighted imaging. In order to set up a regression prediction model which based on texture parameters and morphological features. Furthermore, the comparison of diagnostic accuracy of each TA regression models were assessed by operating characteristic curves (ROC).

  Results  Among the imaging features of peripheral lesions, the incidence of surgical capsule invasion, prostatic capsule involvement, and seminal vesicle invasion in patients with IDCP was higher than that in patients with adenocarcinoma alone (P<0.05). Among the imaging features of transitional zone lesions, patients with adenocarcinoma containing IDCP had a higher incience of imaging features as benign prostatic hyperplasia, extracapsular invasion of the prostate, seminal vesicles invasion, and vascular nerve bundles than that in pure adenocarcinoma group (P<0.05). In order to differential diagnosis of simple adenocarcinoma and adenocarcinoma containing IDCP, 4 arterial phase texture features were used to build the regression model in the peripheral zone group, and the area under the curve (AUC) of the TA model and combined model with or without seminal vesicles invasion were 0.890 and 0.938, respectively. In the transitional zone group, 2 arterial phase texture features and 2 venous phase features were used in TA regression model, and the AUC of texture model and the combined model with or without vascular nerve bundles were 0.844 and 0.901, respectively.

  Conclusion  The incidence of adenocarcinoma with IDCP is higher in high-grade invasive prostate cancer. It is wonderful that when compared with T2WI, enhanced sequential texture analysis is more valuable when using the radiomics approach based on MRI texture analysis in identifying IDCP in prostate cancer.