Diffusion weighted imaging of prostate cancer: Prediction of cancer using texture features from parametric maps of the monoexponential and kurtosis functions

Computer aided diagnosis (CADx) systems for magnetic resonance imaging of prostate have shown potential to increase accuracy for detection of cancer. The purpose of this study is to introduce a method for CADx to detect prostate cancer based on texture features extracted from a grid placed on diffusion weighted imaging (DWI) parametric maps. Texture maps of DWI parametric maps (monoexponential: ADCm, kurtosis: ADCk and K) from 67 patients were obtained. Then the texture maps were divided in cubes, and median texture features were calculated for each cube. The features were used to train prediction models. Area under the curve (AUC) value was used to assess the prediction efficiency. In total, 875 texture features were extracted with Gabor filter, GLCM, LBP, Haar transform, and Hu moments. Statistical features were also calculated. The union of texture features from the ADCm ADCk and K parametric maps demonstrated high performance with AUC values of 0.81 to 0.85.