Objective: To explore the value of artificial intelligence combined with multi-parametric MRI (AI-mpMRI) in the early diagnosis of prostate cancer.
Methods: This retrospective study included 64 cases of prostate cancer confirmed by biopsy and treated by radical prostatectomy from May 2017 to February 2018. The mpMRI images of T2 weighted imaging (T2WI), diffusion weighted imaging (DWI) and dynamic-contrast enhanced (DCE) MRI and the pathological sections corresponding to the three sequential MRI images were collected. The benign and malignant regions were labeled on the pathological slice level, the three sequential MRI axial images at the same level were virtually covered with the pathological slice using computer-aided transparent mapping technology, and selected the fixed-sized benign and malignant regions of interest (ROI). The MATLAB software was used to display the features of the images and screen out the characteristic parameters with P < 0.05, so as to derive high-accuracy analytical methods for the diagnosis of prostate cancer.
Results: A total of 31 image characteristics were extracted with the MATLAB software, and 3 high-accuracy analytical methods screened out for the diagnosis of prostate cancer, including the linear discrimination, logistic regression analysis, and support vector machine classification, with the accuracy rates of 75.9%, 75.4% and 74.9% and the areas under the curve (AUC) of 0.83, 0.82 and 0.82, respectively.
Conclusions: AI-mpMRI can achieve a high detection rate in the early diagnosis of prostate cancer and therefore has a high clinical application value.
Keywords: computer-aided diagnosis; multi-parametric magnetic resonance; prostate cancer; artificial intelligence.