Premium
Diffusion in prostate cancer detection on a 3T scanner: How many b‐values are needed?
Author(s) -
de Perrot Thomas,
Scheffler Max,
Boto José,
Delattre Bénédicte M. A.,
Combescure Christophe,
Pusztaszeri Marc,
Tille JeanChristophe,
Iselin Christophe,
Vallée JeanPaul
Publication year - 2016
Publication title -
journal of magnetic resonance imaging
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.563
H-Index - 160
eISSN - 1522-2586
pISSN - 1053-1807
DOI - 10.1002/jmri.25206
Subject(s) - voxel , prostate cancer , diffusion mri , nuclear medicine , medicine , effective diffusion coefficient , receiver operating characteristic , perfusion , magnetic resonance imaging , prostate , prostatectomy , fractional anisotropy , radiology , area under the curve , cancer
Purpose To assess the influence of perfusion on apparent coefficient diffusion (ADC) maps, the contribution of b‐value images, and the number of b‐values needed in prostate cancer detection by diffusion‐weighted imaging (DWI). Materials and Methods Patients scheduled for prostatectomy were scanned by 3T magnetic resonance imaging (MRI) with DWI based on b‐values 0‐500‐1000‐1500 s/mm 2 . A monoexponential model was fitted to obtain ADC using multiple b‐values, with or without b0 (perfusion‐sensitive ADC 4b‐b0‐500‐1000‐1500 , perfusion‐insensitive ADC 3b‐b500‐1000‐1500 ), or two b‐values (ADC 2b‐b0‐500 , ADC 2b‐b0‐1000 , ADC 2b‐b0‐1500 ). Prostate and cancer foci were segmented to label voxels as normal or tumoral, according to histology. Areas under receiver operating characteristic curves (AUC) were calculated for each ADC and b‐value, then for multivariate logistic regression models combining them. A threshold of 85 tumoral voxels (=0.5 cm 3 ) was used to stratify AUC analysis. Results In all, 21 patients were selected. Segmentation collected 143,665 prostatic voxels including 10,069 tumoral voxels. In five patients, tumor segmentation provided fewer than 85 voxels, resulting in an ADC with AUC inferior to 0.52. In 16 patients with larger tumors, perfusion‐sensitive ADC 4b‐b0‐500‐1000‐1500 performed better than perfusion‐insensitive ADC 3b‐b500‐1000‐1500 and similar to ADC 2b‐b0‐1500 (AUC of 0.840, 0.809, and 0.838, respectively). In comparison to the ADC alone, models combining ADC 4b‐b0‐500‐1000‐1500 or ADC 2b‐b0‐1500 with b1500 improved performance, leading to similar AUCs of 0.884 and 0.883, respectively. In both models, ADC and b1500 were significant markers ( P < 0.001). Conclusion Including b0 in ADC calculation provided superior ADC maps for prostate cancer detection. b1500 images as a combined parameter with ADC also improved performance. Using more than two b‐values showed no improvement. J. Magn. Reson. Imaging 2016;44:601–609.