To evaluate the efficacy of a new computer-aided diagnosis (CAD) system that is based on intravoxel incoherent motion (IVIM) imaging features for differentiation of prostate cancer (PCa), benign prostatic hyperplasia (BPH) and healthy prostatic tissue.

 Twenty two patients with elevated PSA level above normal range (0-4 ng/ml) were included in this study. Multiparametric MRI (mp-MRI) using 3T machine were done initially for all cases with diffusion weighted imaging (DWI) was acquired with 9 b-values (0 - to 1400 s/mm2). After annotation and definition of the detected lesions by MRI, and correlated with the hisopathologic findings. Based on findings in mp-MRI, two regions of interest (ROIs) were defined on each patient’s diffusion weighted (DW) images one in identified lesion and the other in contra-lateral normal tissues. Two highest combinations of b-values were used to generate IVIM parameters. Here, we proposed a CAD system that utilize different components of the IVIM  model along with different machine learning classifiers including decision trees, linear discriminant analysis, support vector...

...read more

 Our study included 22 patients with ages ranged from 55-82years (mean age = 76 years), their mean PSA was 10 ng/dl. Thirty lesions were identified by mp-MRI in all patients. Histopathology results found 16 benign and 14 malignant lesions. Using leave-one-out-cross validation schema, RF achieved the highest diagnostic performance, providing an accuracy of 93.33%, a sensitivity of 92.86%, a specificity of 93.75%, and an AUC of 0.9333 in distinguishing between benign and malignant cases using the diffusion coefficient and perfusion fraction.

 In this work, a non-invasive CAD system that utilized IVIM imaging features for diagnosing PCa is presented. The proposed CAD system provides highly accurate results to diagnose PCa and in differentiation between it and BPH as well as normal prostatic tissue.