Search

Purpose: To evaluate repeatability of prostate DWI-derived radiomics and machine learning methods for prostate cancer (PCa) characterization., Methods: A total of 112 patients with diagnosed PCa underwent 2 prostate MRI examinations (Scan1 and Scan2) performed on the same day. DWI was performed using 12 b-values (0-2000 s/mm 2 ), post-processed using kurtosis function, and PCa areas were annotated using whole mount prostatectomy sections. A total of 1694 radiomic features including Sobel, Kirch, Gradient, Zernike Moments, Gabor, Haralick, CoLIAGe, Haar wavelet coefficients, 3D analogue to Laws features, 2D contours, and corner detectors were calculated. Radiomics and 4 feature pruning methods (area under the receiver operator characteristic curve, maximum relevance minimum redundancy, Spearman's ρ, Wilcoxon rank-sum) were evaluated in terms of Scan1-Scan2 repeatability using intraclass correlation coefficient (ICC)(3,1). Classification performance for clinically significant and insignificant PCa with Gleason grade groups 1 versus >1 was evaluated by area under the receiver operator characteristic curve in unseen random 30% data split., Results: The ICC(3,1) values for conventional radiomics and feature pruning methods were in the range of 0.28-0.90. The machine learning classifications varied between Scan1 and Scan2 with % of same class labels between Scan1 and Scan2 in the range of 61-81%. Surface-to-volume ratio and corner detector-based features were among the most represented features with high repeatability, ICC(3,1) >0.75, consistently high ranking using all 4 feature pruning methods, and classification performance with area under the receiver operator characteristic curve >0.70., Conclusion: Surface-to-volume ratio and corner detectors for prostate DWI led to good classification of unseen data and performed similarly in Scan1 and Scan2 in contrast to multiple conventional radiomic features., (© 2019 International Society for Magnetic Resonance in Medicine.)

Purpose: To evaluate different fitting methods for intravoxel incoherent motion (IVIM) imaging of prostate cancer in the terms of repeatability and Gleason score prediction., Methods: Eighty-one patients with histologically confirmed prostate cancer underwent two repeated 3 Tesla diffusion-weighted imaging (DWI) examinations performed using 14 b-values in the range of 0-500 s/mm 2 and diffusion time of 19.004 ms. Mean signal intensities of regions-of-interest were fitted using five different fitting methods for IVIM as well as monoexponential, kurtosis, and stretched exponential models. The fitting methods and models were evaluated in the terms of fitting quality [Akaike information criteria (AIC)], repeatability, and Gleason score prediction. Tumors were classified into three groups (3 + 3, 3 + 4, > 3 + 4). Machine learning algorithms were used to evaluate the performance of the combined use of the parameters. Simulation studies were performed to evaluate robustness of the fitting methods against noise., Results: Monoexponential model was preferred over IVIM based on AIC. The "pseudodiffusion" parameters demonstrated low repeatability and clinical value. Median "pseudodiffusion" fraction values were below 8.00%. Combined use of the parameters did not outperform the monoexponential model., Conclusion: Monoexponential model demonstrated the highest repeatability and clinical values in the regions-of-interest based analysis of prostate cancer DWI, b-values in the range of 0-500 s/mm 2 . Magn Reson Med 77:1249-1264, 2017. © 2016 International Society for Magnetic Resonance in Medicine., (© 2016 International Society for Magnetic Resonance in Medicine.)

Purpose: To evaluate the performance of relaxation along a fictitious field (RAFF) relaxation time (TRAFF ), diffusion-weighted imaging (DWI)-derived parameters, and T2 relaxation time values for prostate cancer (PCa) detection and characterization., Methods: Fifty patients underwent 3T MR examination using surface array coils before prostatectomy. DWI was performed using 14 and 12 b values in the ranges of 0-500 s/mm(2) and 0-2000 s/mm(2) , respectively. Repeated MR examination was performed in 16 patients. TRAFF , DWI-derived parameters (monoexponential, kurtosis, biexponential models), and T2 values were measured and averaged over regions of interest placed in PCa and normal tissue. Repeatability of TRAFF and DWI-derived parameters were assessed by coefficient of repeatability and intraclass correlation coefficient ICC(3,1). Areas under the receiver operating characteristic curve (AUCs) for PCa detection and Gleason score classification were estimated. The parameters were correlated with Gleason score groups using Spearman correlation coefficient (ρ)., Results: ICC(3,1) values for TRAFF were in the range of 0.82-0.92. TRAFF values had higher AUC values for Gleason score classification compared with DWI-derived parameters and T2 . The RAFF method demonstrated the highest ρ value (-0.65)., Conclusion: In a quantitative region of interest-based analysis, RAFF outperformed DWI ("low" and "high" b values) and T2 mapping in the characterization of PCa., (© 2015 Wiley Periodicals, Inc.)

Purpose: To investigate relaxation along a fictitious field (RAFF) and continuous wave (cw) T1ρ imaging of prostate cancer (PCa) in the terms of repeatability, PCa detection, and characterization., Methods: Thirty-six patients (PSA 11.6 ± 7.6 ng/mL, mean ± standard deviation) with histologically confirmed PCa underwent two repeated 3T MR examinations using surface array coils before prostatectomy. Relaxation along fictitious field, cw T1ρ, and T2 relaxation times (TRAFF, T1ρcw, T2) were measured and averaged over regions of interest placed in PCa, normal peripheral zone (PZ), and normal central gland (CG) positioned using whole-mount prostatectomy sections and anatomical T2-weighted images. Receiver operating characteristic curve analysis with area under the curve (AUC) was calculated to distinguish PCa from PZ/CG and PCa with Gleason score (GS) of 3+3 from GS of 3+4/≥ 3+4., Results: TRAFF and T1ρcw relaxation times were repeatable with coefficients of repeatability as a percentage of median value in the range of 7.8-23.2%. AUC (mean, 95% confidence interval) in the differentiation of PCa with GS of 3+3 from PCa with CS of ≥ 3+4 were 0.88 (0.72-0.99), 0.69 (0.46-0.90), and 0.68 (0.45-0.88), for TRAFF, T1ρcw, and T2, respectively., Conclusion: In quantitative region of interest based analysis, TRAFF outperformed T1ρcw and T2 in PCa detection and characterization., (© 2015 Wiley Periodicals, Inc.)

Purpose: To evaluate four mathematical models for diffusion weighted imaging (DWI) of prostate cancer (PCa) in terms of PCa detection and characterization., Methods: Fifty patients with histologically confirmed PCa underwent two repeated 3 Tesla DWI examinations using 12 equally distributed b values, the highest b value of 2000 s/mm(2) . Normalized mean signal intensities of regions-of-interest were fitted using monoexponential, kurtosis, stretched exponential, and biexponential models. Tumors were classified into low, intermediate, and high Gleason score groups. Areas under receiver operating characteristic curve (AUCs) were estimated to evaluate performance in PCa detection and Gleason score classifications. The fitted parameters were correlated with Gleason score groups by using the Spearman correlation coefficient (ρ). Coefficient of repeatability and intraclass correlation coefficient [specifically ICC(3,1)], were calculated to evaluate repeatability of the fitted parameters., Results: The AUC and ρ values were similar between parameters of monoexponential, kurtosis, and stretched exponential (with the exception of the α parameter) models. The absolute ρ values for ADCm , ADCk , K, and ADCs were in the range from 0.31 to 0.53 (P < 0.01). Parameters of the biexponential model demonstrated low repeatability., Conclusion: In region-of-interest based analysis, the monoexponential model for DWI of PCa using b values up to 2000 s/mm(2) was sufficient for PCa detection and characterization., (© 2014 Wiley Periodicals, Inc.)

Purpose: To evaluate monoexponential, stretched exponential, kurtosis, and biexponential models for diffusion-weighted imaging (DWI) of normal prostate and prostate cancer (PCa), using b-values up to 2000 s/mm(2) , in terms of fitting quality and repeatability., Methods: Eight healthy volunteers and 16 PCa patients underwent a total of four repeated 3T DWI examinations using 16 and 12 b-values, respectively. The highest b-value was 2000 s/mm(2) . The normalized mean signal intensities of regions of interest, placed in normal tissue and PCa using anatomical images and prostatectomy sections, were fitted using the four models. The fitting quality was evaluated using Akaike information criteria and F-ratio. Repeatability of the fitted parameters was evaluated using intraclass correlation coefficient ICC(3,1)., Results: The biexponential model provided the best fit to normal prostate and PCa DWI data. The parameters of the monoexponential, kurtosis, and stretched exponential (with the exception of the α parameter) models had higher ICC(3,1) values compared with the biexponential model. The kurtosis model provided a better fit to DWI data of normal prostate and PCa than the monoexponential model, whereas these models had comparable reliability and repeatability based on ICC(3,1) values., Conclusion: Considering the model fit and repeatability, the kurtosis model seems to be the preferred model for characterization of normal prostate and PCa DWI using b-values up to 2000 s/mm(2) ., (© 2014 Wiley Periodicals, Inc.)