Contribution of Dynamic Contrast-enhanced and Diffusion MRI to PI-RADS for Detecting Clinically Significant Prostate Cancer.

Background Prostate Imaging Reporting and Data System (PI-RADS) version 2.0 requires multiparametric MRI of the prostate, including diffusion-weighted imaging (DWI) and dynamic contrast-enhanced (DCE) imaging sequences; however, the contribution of DCE imaging remains unclear. Purpose To assess whether DCE imaging in addition to apparent diffusion coefficient (ADC) and normalized T2 values improves PI-RADS version 2.0 for prediction of clinically significant prostate cancer (csPCa). Materials and Methods In this retrospective study, clinically reported PI-RADS lesions in consecutive men who underwent 3-T multiparametric MRI (T2-weighted, DWI, and DCE MRI) from May 2015 to September 2016 were analyzed quantitatively and compared with systematic and targeted MRI-transrectal US fusion biopsy. The normalized T2 signal (nT2), ADC measurement, mean early-phase DCE signal (mDCE), and heuristic DCE parameters were calculated. Logistic regression analysis indicated the most predictive DCE parameters for csPCa (Gleason grade group ≥2). Receiver operating characteristic parameter models were compared using the Obuchowski test. Recursive partitioning analysis determined ADC and mDCE value ranges for combined use with PI-RADS. Results Overall, 260 men (median age, 64 years [IQR, 58-69 years]) with 432 lesions (csPCa [n = 152] and no csPCa [n = 280]) were included. The mDCE parameter was predictive of csPCa when accounting for the ADC and nT2 parameter in the peripheral zone (odds ratio [OR], 1.76; 95% CI: 1.30, 2.44; P = .001) but not the transition zone (OR, 1.17; 95% CI: 0.81, 1.69; P = .41). Recursive partitioning analysis selected an ADC cutoff of 0.897 × 10-3 mm2/sec (P = .04) as a classifier for peripheral zone lesions with a PI-RADS score assessed on the ADC map (hereafter, ADC PI-RADS) of 3. The mDCE parameter did not differentiate ADC PI-RADS 3 lesions (P = .11), but classified lesions with ADC PI-RADS scores greater than 3 with low ADC values (less than 0.903 × 10-3 mm2/sec, P < .001) into groups with csPCa rates of 70% and 97% (P = .008). A lesion size cutoff of 1.5 cm and qualitative DCE parameters were not defined as classifiers according to recursive partitioning (P > .05). Conclusion Quantitative or qualitative dynamic contrast-enhanced MRI was not relevant for Prostate Imaging Reporting and Data System (PI-RADS) 3 lesion risk stratification, while quantitative apparent diffusion coefficient (ADC) values were helpful in upgrading PI-RADS 3 and PI-RADS 4 lesions. Quantitative ADC measurement may be more important for risk stratification than current methods in future versions of PI-RADS. © RSNA, 2022 Online supplemental material is available for this article See also the editorial by Goh in this issue.

Diffusion pore imaging in the presence of extraporal water.

Diffusion-weighted imaging (DWI) is a powerful, non-invasive tool which is widely used in clinical routine. Mostly, apparent diffusion coefficient maps are acquired, which cannot be related directly to cellular structure. More recently it was shown that DWI is able to reconstruct pore shapes using a specialized magnetic field gradient scheme so that cell size distributions may be obtained. So far, artificial systems have been used for experimental demonstration without extraporal signal components and relatively low gradient amplitudes. The aim of this study was to investigate the feasibility of diffusion pore imaging in the presence of extraporal fluids and to develop correction methods for the effects arising from extraporal signal contributions. Monte Carlo simulations and validation experiments on a 14.1 T NMR spectrometer equipped with a dedicated diffusion probe head were performed. Both by using a filter gradient approach suppressing extraporal signal components as well as by using post-processing methods relying on the Gaussian phase approximation, it was possible to reconstruct pore space functions in the presence of extraporal fluids with little to no deviations from the expectations. These results may be a significant step towards application of diffusion pore imaging to biological samples.

Echo time dependence of biexponential and triexponential intravoxel incoherent motion parameters in the liver.

PURPOSE: Intravoxel incoherent motion (IVIM) studies are performed with different acquisition protocols. Comparing them requires knowledge of echo time (TE) dependencies. The TE-dependence of the biexponential perfusion fraction f is well-documented, unlike that of its triexponential counterparts f1 and f2 and the biexponential and triexponential pseudodiffusion coefficients D* , D1∗ , and D2∗ . The purpose was to investigate the TE-dependence of these parameters and to check whether the triexponential pseudodiffusion compartments are associated with arterial and venous blood. METHODS: Fifteen healthy volunteers (19-58 y; mean: 24.7 y) underwent diffusion-weighted imaging of the abdomen with 24 b-values (0.2-800 s/mm2 ) at TEs of 45, 60, 75, and 90 ms. Regions of interest (ROIs) were manually drawn in the liver. One set of bi- and triexponential IVIM parameters per volunteer and TE was determined. The TE-dependence was assessed with the Kruskal-Wallis test. RESULTS: TE-dependence was observed for f (P < .001), f1 (P = .001), and f2 (P < .001). Their median values at the four measured TEs were: f: 0.198/0.240/0.274/0.359, f1 : 0.113/0.139/0.146/0.205, f2 : 0.115/0.155/0.182/0.194. D, D* , D1∗ , and D2∗ showed no significant TE-dependence. Their values were: diffusion coefficient D (10-4 mm2 /s): 9.45/9.63/9.75/9.41, biexponential D* (10-2 mm2 /s): 5.26/5.52/6.13/5.82, triexponential D1∗ (10-2 mm2 /s): 1.73/2.91/2.25/2.51, triexponential D2∗ (mm2 /s): 0.478/1.385/0.616/0.846. CONCLUSION: f1 and f2 show similar TE-dependence as f, ie, increase with rising TE; an effect that must be accounted for when comparing different studies. The diffusion and pseudodiffusion coefficients might be compared without TE correction. Because of the similar TE-dependence of f1 and f2 , the triexponential pseudodiffusion compartments are most probably not associated to venous and arterial blood.

Pancreatic imaging using diffusivity mapping - Influence of sequence technique on qualitative and quantitative analysis.

PURPOSE: To compare image quality of an optimized diffusion weighted imaging (DWI) sequence with advanced post-processing and motion correction (advanced-EPI) to a standard DWI protocol (standard-EPI) in pancreatic imaging. MATERIALS AND METHODS: 62 consecutive patients underwent abdominal MRI at 1.5 T were included in this retrospective analysis of data collected as part of an IRB approved study. All patients received a standard-EPI and an advanced-EPI DWI with advanced post-processing and motion correction. Two blinded radiologists evaluated the parameters image quality, detail of parenchyma, sharpness of boundaries and discernibility from adjacent structures on b = 900 s/mm2 images using a Likert-like scale. Segmentation of pancreatic head, body and tail were obtained and apparent diffusion coefficient (ADC) was calculated separately for each region. Apparent tissue-to-background ratio (TBR) was calculated at b = 50 s/mm2 and at b = 900 s/mm2. RESULTS: The advanced-EPI yielded significantly higher scores for pancreatic parameters of image quality, detail level of parenchyma, sharpness of boundaries and discernibility from adjacent structures in comparison to standard-EPI (p < 0.001 for all, kappa = [0.46,0.71]) and was preferred in 96% of the cases when directly compared. ADC of the pancreas was 7% lower in advanced-EPI (1.236 ± 0.152 vs. 1.146 ± 0.126 µm2/ms, p < 0.001). ADC in the pancreatic tail was significantly lower for both sequences compared to head and body (all p < 0.001). There was comparable TBR for both sequences at b = 50 s/mm2 (standard-EPI: 19.0 ± 5.9 vs. advanced-EPI: 19.0 ± 6.4, p = 0.96), whereas at b = 900 s/mm2, TBR was 51% higher for advanced-EPI (standard-EPI: 7.1 ± 2.5 vs. advanced-EPI: 10.8 ± 5.1, p < 0.001). CONCLUSION: An advanced DWI sequence might increase image quality for focused imaging of the pancreas and providing improved parenchymal detail levels compared to a standard DWI.

Improvement of PI-RADS-dependent prostate cancer classification by quantitative image assessment using radiomics or mean ADC.

Background Currently, interpretation of prostate MRI is performed qualitatively. Quantitative assessment of the mean apparent diffusion coefficient (mADC) is promising to improve diagnostic accuracy while radiomic machine learning (RML) allows to probe complex parameter spaces to identify the most promising multi-parametric models. We have previously developed quantitative RML and ADC classifiers for prediction of clinically significant prostate cancer (sPC) from prostate MRI, however these have not been combined with radiologist PI-RADS assessment. Purpose To propose and evaluate diagnostic algorithms combining quantitative ADC or RML and qualitative PI-RADS assessment for prediction of sPC. Methods and population The previously published quantitative models (RML and mADC) were utilized to construct four algorithms: 1) Down(ADC) and 2) Down(RML): clinically detected PI-RADS positive prostate lesions (defined as either PI-RADS≥3 or ≥4) were downgraded to MRI negative upon negative quantitative assessment; and 3) Up(ADC) and 4) Up(RML): MRI-negative lesions were upgraded to MRI-positive upon positive assessment of quantitative parameters. Analyses were performed at the individual lesion level and the patient level in 133 consecutive patients with suspicion for clinically significant prostate cancer (sPC, International Society of Urological Pathology (ISUP) grade group≥2), the test set subcohort of a previously published patient population. McNemar test was used to compare differences in sensitivity, specificity and accuracy. Differences between lesions of different prostate zones were assessed using ANOVA. Reduction in false positive assessments was assessed as ratios. Results Compared to clinical assessment at the PI-RADS≥4 cut-off alone, algorithms Down(ADC/RML) improved specificity from 43% to 65% (p = 0.001)/62% (p = 0.003), while sensitivity did not change significantly at 89% compared to 87% (p = 1.0)/89% (unchanged) on the patient level. Reduction of false positive lesions was 50% [26/52] in the PZ and 53% [15/28] in the TZ. Algorithms Up(ADC/RML) led, on a patient basis, to an unfavorable loss of specificity from 43% to 30% (p = 0.039)/32% (p = 0.106), with insignificant increase of sensitivity from 89% to 96%/96% (both p = 1.0). Compared to clinical assessment at the PI-RADS≥3 cut-off alone, similar results were observed for Down(ADC) with significantly increased specificity from 2% to 23% (p < 0.001) and unchanged sensitivity on the lesion level; patient level specificity increased only non-significantly. Conclusion Downgrading PI-RADS≥3 and ≥ 4 lesions based on quantitative mADC measurements or RML classifiers can increase diagnostic accuracy by enhancing specificity and preserving sensitivity for detection of sPC and reduce false positives.

Fully Automatic Deep Learning in Bi-institutional Prostate Magnetic Resonance Imaging: Effects of Cohort Size and Heterogeneity.

BACKGROUND: The potential of deep learning to support radiologist prostate magnetic resonance imaging (MRI) interpretation has been demonstrated. PURPOSE: The aim of this study was to evaluate the effects of increased and diversified training data (TD) on deep learning performance for detection and segmentation of clinically significant prostate cancer-suspicious lesions. MATERIALS AND METHODS: In this retrospective study, biparametric (T2-weighted and diffusion-weighted) prostate MRI acquired with multiple 1.5-T and 3.0-T MRI scanners in consecutive men was used for training and testing of prostate segmentation and lesion detection networks. Ground truth was the combination of targeted and extended systematic MRI-transrectal ultrasound fusion biopsies, with significant prostate cancer defined as International Society of Urological Pathology grade group greater than or equal to 2. U-Nets were internally validated on full, reduced, and PROSTATEx-enhanced training sets and subsequently externally validated on the institutional test set and the PROSTATEx test set. U-Net segmentation was calibrated to clinically desired levels in cross-validation, and test performance was subsequently compared using sensitivities, specificities, predictive values, and Dice coefficient. RESULTS: One thousand four hundred eighty-eight institutional examinations (median age, 64 years; interquartile range, 58-70 years) were temporally split into training (2014-2017, 806 examinations, supplemented by 204 PROSTATEx examinations) and test (2018-2020, 682 examinations) sets. In the test set, Prostate Imaging-Reporting and Data System (PI-RADS) cutoffs greater than or equal to 3 and greater than or equal to 4 on a per-patient basis had sensitivity of 97% (241/249) and 90% (223/249) at specificity of 19% (82/433) and 56% (242/433), respectively. The full U-Net had corresponding sensitivity of 97% (241/249) and 88% (219/249) with specificity of 20% (86/433) and 59% (254/433), not statistically different from PI-RADS (P > 0.3 for all comparisons). U-Net trained using a reduced set of 171 consecutive examinations achieved inferior performance (P < 0.001). PROSTATEx training enhancement did not improve performance. Dice coefficients were 0.90 for prostate and 0.42/0.53 for MRI lesion segmentation at PI-RADS category 3/4 equivalents. CONCLUSIONS: In a large institutional test set, U-Net confirms similar performance to clinical PI-RADS assessment and benefits from more TD, with neither institutional nor PROSTATEx performance improved by adding multiscanner or bi-institutional TD.

Diffusivity mapping of the ovaries: Variability of apparent diffusion and kurtosis variables over the menstrual cycle and influence of oral contraceptives.

PURPOSE: We aimed to investigate whether quantitative diffusivity variables of healthy ovaries vary during the menstrual cycle and to evaluate alterations in women using oral contraceptives (OC). METHODS: This prospective study (S-339/2016) included 30 healthy female volunteers, with (n = 15) and without (n = 15) intake of OC between 07/2017 and 09/2019. Participants underwent 3T diffusion-weighted MRI (b-values 0-2000 s/mm2) three times during a menstrual cycle (T1 = day 1-5; T2 = day 7-12; T3 = day 19-24). Both ovaries were manually three-dimensionally segmented on b = 1500 s/mm2; apparent diffusion coefficient (ADC) calculation and kurtosis fitting (Dapp, Kapp) were performed. Differences in ADC, Dapp and Kapp between time points and groups were compared using repeated measures ANOVA and t-test after Shapiro-Wilk and Brown-Forsythe test for normality and equal variance. RESULTS: In women with a natural menstrual cycle, ADC and kurtosis variables showed significant changes in ovaries with the dominant follicle between T1 vs T2 and T1 vs T3, whilst no differences were observed between T2 vs T3: ADC ± SD for T1 1.524 ± 0.160, T2 1.737 ± 0.160, and T3 1.747 ± 0.241 µm2/ms (p = 0.01 T2 vs T1; p = 1.0 T2 vs T3, p = 0.003 T3 vs T1); Dapp ± SD for T1 2.018 ± 0.140, T2 2.272 ± 0.189, and T3 2.230 ± 0.256 µm2/ms (p = 0.003 T2 vs T1, p = 1.0 T2 vs T3, p = 0.02 T3 vs T1); Kapp ± SD for T1 0.614 ± 0.0339, T2 0.546 ± 0.0637, and T3 0.529 ± 0.0567 (p < 0.001 T2 vs T1, p = 0.86 T2 vs T3, p < 0.001 T3 vs T1). No significant differences were found in the contralateral ovaries or in females taking OC. CONCLUSION: Physiological cycle-dependent changes in quantitative diffusivity variables of ovaries should be considered especially when interpreting radiomics analyses in reproductive women.

Brain tissues have single-voxel signatures in multi-spectral MRI.

Since the seminal works by Brodmann and contemporaries, it is well-known that different brain regions exhibit unique cytoarchitectonic and myeloarchitectonic features. Transferring the approach of classifying brain tissues - and other tissues - based on their intrinsic features to the realm of magnetic resonance (MR) is a longstanding endeavor. In the 1990s, atlas-based segmentation replaced earlier multi-spectral classification approaches because of the large overlap between the class distributions. Here, we explored the feasibility of performing global brain classification based on intrinsic MR features, and used several technological advances: ultra-high field MRI, q-space trajectory diffusion imaging revealing voxel-intrinsic diffusion properties, chemical exchange saturation transfer and semi-solid magnetization transfer imaging as a marker of myelination and neurochemistry, and current neural network architectures to analyze the data. In particular, we used the raw image data as well to increase the number of input features. We found that a global brain classification of roughly 97 brain regions was feasible with gross classification accuracy of 60%; and that mapping from voxel-intrinsic MR data to the brain region to which the data belongs is possible. This indicates the presence of unique MR signals of different brain regions, similar to their cytoarchitectonic and myeloarchitectonic fingerprints.

Apparent exchange rate imaging: On its applicability and the connection to the real exchange rate.

PURPOSE: Water exchange between the intracellular and extracellular space can be measured using apparent exchange rate (AXR) imaging. The aim of this study was to investigate the relationship between the measured AXR and the geometry of diffusion restrictions, membrane permeability, and the real exchange rate, as well as to explore the applicability of AXR for typical human measurement settings. METHODS: The AXR measurements and the underlying exchange rates were simulated using the Monte Carlo method with different geometries, size distributions, packing densities, and a broad range of membrane permeabilities. Furthermore, the influence of SNR and sequence parameters was analyzed. RESULTS: The estimated AXR values correspond to the simulated values and show the expected proportionality to membrane permeability, except for fast exchange (ie, AXR>20-30s-1 ) and small packing densities. Moreover, it was found that the duration of the filter gradient must be shorter than 2·AXR-1 . In cell size and permeability distributions, AXR depends on the average surface-to-volume ratio, permeability, and the packing density. Finally, AXR can be reliably determined in the presence of orientation dispersion in axon-like structures with sufficient gradient sampling (ie, 30 gradient directions). CONCLUSION: Currently used experimental settings for in vivo human measurements are well suited for determining AXR, with the exception of single-voxel analysis, due to limited SNR. The detection of changes in membrane permeability in diseased tissue is nonetheless challenging because of the AXR dependence on further factors, such as packing density and geometry, which cannot be disentangled without further knowledge of the underlying cell structure.

Comparison of single-scanner single-protocol quantitative ADC measurements to ADC ratios to detect clinically significant prostate cancer.

BACKGROUND: Mean ADC has high predictive value for the presence of clinically significant prostate cancer (sPC). Measurement variability is introduced by different scanners, protocols, intra-and inter-patient variation. Internal calibration by ADC ratios can address such fluctuations however can potentially lower the biological value of quantitative ADC determination by being sensitive to deviations in reference tissue signal. PURPOSE: To better understand the predictive value of quantitative ADC measurements in comparison to internal reference ratios when measured in a single scanner, single protocol setup. MATERIALS AND METHODS: 284 consecutive patients who underwent 3 T MRI on a single scanner followed by MRI-transrectal ultrasound fusion biopsy were included. A board-certified radiologist retrospectively reviewed all MRIs blinded to clinical information and placed regions of interest (ROI) on all focal lesions and the following reference regions: normal-appearing peripheral zone (PZNL) and transition zone (TZNL), the urinary bladder (BLA), and right and left internal obturator muscle (RIOM, LIOM). ROI-based mean ADC and ADC ratios to the reference regions were compared regarding their ability to predict the aggressiveness of prostate cancer. Spearman's rank correlation coefficient was used to estimate the correlation between ADC parameters, Gleason score (GS) and ADC ratios. The primary endpoint was presence of sPC, defined as a GS ≥ 3 + 4. Univariable and multivariable logistic regression models were constructed to predict sPC. Receiver operating characteristics curves (ROC) were used for visualization; DeLong test was used to evaluate the differences of the area under the curve (AUC). Bias-corrected AUC values and corresponding 95 %-CI were calculated using bootstrapping with 100 bootstrap samples. RESULTS: After exclusion of patients who received prior treatment, 259 patients were included in the final cohort of which 220 harbored 351 MR lesions. Mean ADC and ADC ratios demonstrated a negative correlation with the GS. Mean ADC had the strongest correlation with ρ of -0.34, followed by ADCratioPZNL (ρ=-0.32). All ADC parameters except ADCratioLIOM (p = 0.07) were associated with sPC p<0.05). Mean ADC and ADCratioPZNL had the highest ROC AUC of all parameters (0.68). Multivariable models with mean ADC improve predictive performance. CONCLUSIONS: A highly standardized single-scanner mean ADC measurement could not be improved upon using any of the single ADC ratio parameters or combinations of these parameters in predicting the aggressiveness of prostate cancer.

Measured Multipoint Ultra-High b-Value Diffusion MRI in the Assessment of MRI-Detected Prostate Lesions.

OBJECTIVES: The aim of this study was to assess quantitative ultra-high b-value (UHB) diffusion magnetic resonance imaging (MRI)-derived parameters in comparison to standard clinical apparent diffusion coefficient (SD-ADC-2b-1000, SD-ADC-2b-1500) for the prediction of clinically significant prostate cancer, defined as Gleason Grade Group greater than or equal to 2. MATERIALS AND METHODS: Seventy-three patients who underwent 3-T prostate MRI with diffusion-weighted imaging acquired at b = 50/500/1000/1500s/mm2 and b = 100/500/1000/1500/2250/3000/4000 s/mm2 were included. Magnetic resonance lesions were segmented manually on individual sequences, then matched to targeted transrectal ultrasonography/MRI fusion biopsies. Monoexponential 2-point and multipoint fits of standard diffusion and of UHB diffusion were calculated with incremental b-values. Furthermore, a kurtosis fit with parameters Dapp and Kapp with incremental b-values was obtained. Each parameter was examined for prediction of clinically significant prostate cancer using bootstrapped receiver operating characteristics and decision curve analysis. Parameter models were compared using Vuong test. RESULTS: Fifty of 73 men (age, 66 years [interquartile range, 61-72]; prostate-specific antigen, 6.6 ng/mL [interquartile range, 5-9.7]) had 64 MRI-detected lesions. The performance of SD-ADC-2b-1000 (area under the curve, 0.82) and SD-ADC-2b-1500 (area under the curve, 0.82) was not statistically different (P = 0.99), with SD-ADC-2b-1500 selected as reference. Compared with the reference model, none of the 19 tested logistic regression parameter models including multipoint and 2-point UHB-ADC, Dapp, and Kapp with incremental b-values of up to 4000 s/mm2 outperformed SD-ADC-2b-1500 (all P's > 0.05). Decision curve analysis confirmed these results indicating no higher net benefit for UHB parameters in comparison to SD-ADC-2b-1500 in the clinically important range from 3% to 20% of cancer threshold probability. Net reduction analysis showed no reduction of MR lesions requiring biopsy. CONCLUSIONS: Despite evaluation of a large b-value range and inclusion of 2-point, multipoint, and kurtosis models, none of the parameters provided better predictive performance than standard 2-point ADC measurements using b-values 50/1000 or 50/1500. Our results suggest that most of the diagnostic benefits available in diffusion MRI are already represented in an ADC composed of one low and one 1000 to 1500 s/mm2 b-value.

Comparison of Prostate MRI Lesion Segmentation Agreement Between Multiple Radiologists and a Fully Automatic Deep Learning System.

PURPOSE: A recently developed deep learning model (U-Net) approximated the clinical performance of radiologists in the prediction of clinically significant prostate cancer (sPC) from prostate MRI. Here, we compare the agreement between lesion segmentations by U-Net with manual lesion segmentations performed by different radiologists. MATERIALS AND METHODS: 165 patients with suspicion for sPC underwent targeted and systematic fusion biopsy following 3 Tesla multiparametric MRI (mpMRI). Five sets of segmentations were generated retrospectively: segmentations of clinical lesions, independent segmentations by three radiologists, and fully automated bi-parametric U-Net segmentations. Per-lesion agreement was calculated for each rater by averaging Dice coefficients with all overlapping lesions from other raters. Agreement was compared using descriptive statistics and linear mixed models. RESULTS: The mean Dice coefficient for manual segmentations showed only moderate agreement at 0.48-0.52, reflecting the difficult visual task of determining the outline of otherwise jointly detected lesions. U-net segmentations were significantly smaller than manual segmentations (p < 0.0001) and exhibited a lower mean Dice coefficient of 0.22, which was significantly lower compared to manual segmentations (all p < 0.0001). These differences remained after correction for lesion size and were unaffected between sPC and non-sPC lesions and between peripheral and transition zone lesions. CONCLUSION: Knowledge of the order of agreement of manual segmentations of different radiologists is important to set the expectation value for artificial intelligence (AI) systems in the task of prostate MRI lesion segmentation. Perfect agreement (Dice coefficient of one) should not be expected for AI. Lower Dice coefficients of U-Net compared to manual segmentations are only partially explained by smaller segmentation sizes and may result from a focus on the lesion core and a small relative lesion center shift. Although it is primarily important that AI detects sPC correctly, the Dice coefficient for overlapping lesions from multiple raters can be used as a secondary measure for segmentation quality in future studies. KEY POINTS: · Intermediate human Dice coefficients reflect the difficulty of outlining jointly detected lesions.. · Lower Dice coefficients of deep learning motivate further research to approximate human perception.. · Comparable predictive performance of deep learning appears independent of Dice agreement.. · Dice agreement independent of significant cancer presence indicates indistinguishability of some benign imaging findings.. · Improving DWI to T2 registration may improve the observed U-Net Dice coefficients.. CITATION FORMAT: · Schelb P, Tavakoli AA, Tubtawee T et al. Comparison of Prostate MRI Lesion Segmentation Agreement Between Multiple Radiologists and a Fully Automatic Deep Learning System. Fortschr Röntgenstr 2021; 193: 559 - 573.

An optimized b-value distribution for triexponential intravoxel incoherent motion (IVIM) in the liver.

PURPOSE: To find an optimized b-value distribution for reproducible triexponential intravoxel incoherent motion (IVIM) exams in the liver. METHODS: A numeric optimization of b-value distributions was performed using the triexponential IVIM equation and 27 different IVIM parameter sets. Starting with an initially optimized distribution of 6 b-values, the number of b-values was increased stepwise. Each new b-value was chosen from a set of 64 predefined b-values based on the computed summed relative mean error of the fitted triexponential IVIM parameters. This process was repeated for up to 100 b-values. In simulations and in vivo measurements, optimized b-value distributions were compared to 4 representative distributions found in literature. RESULTS: The first 16 optimized b-values were 0, 0.3, 0.3, 70, 200, 800, 70, 1, 3.5, 5, 70, 1.2, 6, 45, 1.5, and 60 in units of s/mm2 . Low b-values were much more frequent than high b-values. The optimized b-value distribution resulted in a higher fit stability compared to distributions used in literature in both, simulation and in vivo measurements. Using more than 6 b-values, ideally 16 or more, increased the fit stability considerably. CONCLUSION: Using optimized b-values, the fit uncertainty in triexponential IVIM can be largely reduced. Ideally, 16 or more b-values should be acquired.

Influence of residual fat signal on diffusion kurtosis MRI of suspicious mammography findings.

Recent studies showed the potential of diffusion kurtosis imaging (DKI) as a tool for improved classification of suspicious breast lesions. However, in diffusion-weighted imaging of the female breast, sufficient fat suppression is one of the main factors determining the success. In this study, the data of 198 patients examined in two study centres was analysed using standard diffusion and kurtosis evaluation methods and three DKI fitting approaches accounting phenomenologically for fat-related signal contamination of the lesions. Receiver operating characteristic curve analysis showed the highest area under the curve (AUC) for the method including fat correction terms (AUC = 0.85, p < 0.015) in comparison to the values obtained with the standard diffusion (AUC = 0.77) and kurtosis approach (AUC = 0.79). Comparing the two study centres, the AUC value improved from 0.77 to 0.86 (p = 0.036) using a fat correction term for the first centre, while no significant difference with no adverse effects was observed for the second centre (AUC 0.89 vs. 0.90, p = 0.95). Contamination of the signal in breast lesions with unsuppressed fat causing a reduction of diagnostic performance of diffusion kurtosis imaging may potentially be counteracted by proposed adapted evaluation methods.

Ultra-High-b-Value Kurtosis Imaging for Noninvasive Tissue Characterization of Ovarian Lesions.

Background MRI with contrast material enhancement is the imaging modality of choice to evaluate sonographically indeterminate adnexal masses. The role of diffusion-weighted MRI, however, remains controversial. Purpose To evaluate the diagnostic performance of ultra-high-b-value diffusion kurtosis MRI in discriminating benign and malignant ovarian lesions. Materials and Methods This prospective cohort study evaluated consecutive women with sonographically indeterminate adnexal masses between November 2016 and December 2018. MRI at 3.0 T was performed, including diffusion-weighted MRI (b values of 0-2000 sec/mm2). Lesions were segmented on b of 1500 sec/mm2 by two readers in consensus and an additional independent reader by using full-lesion segmentations on a single transversal slice. Apparent diffusion coefficient (ADC) calculation and kurtosis fitting were performed. Differences in ADC, kurtosis-derived ADC (Dapp), and apparent kurtosis coefficient (Kapp) between malignant and benign lesions were assessed by using a logistic mixed model. Area under the receiver operating characteristic curve (AUC) for ADC, Dapp, and Kapp to discriminate malignant from benign lesions was calculated, as was specificity at a sensitivity level of 100%. Results from two independent reads were compared. Histopathologic analysis served as the reference standard. Results A total of 79 ovarian lesions in 58 women (mean age ± standard deviation, 48 years ± 14) were evaluated. Sixty-two (78%) lesions showed benign and 17 (22%) lesions showed malignant histologic findings. ADC and Dapp were lower and Kapp was higher in malignant lesions: median ADC, Dapp, and Kapp were 0.74 µm2/msec (range, 0.52-1.44 µm2/msec), 0.98 µm2/msec (range, 0.63-2.12 µm2/msec), and 1.01 (range, 0.69-1.30) for malignant lesions, and 1.13 µm2/msec (range, 0.35-2.63 µm2/msec), 1.45 µm2/msec (range, 0.44-3.34 µm2/msec), and 0.65 (range, 0.44-1.43) for benign lesions (P values of .01, .02, < .001, respectively). AUC for Kapp of 0.85 (95% confidence interval: 0.77, 0.94) was higher than was AUC from ADC of 0.78 (95% confidence interval: 0.67, 0.89; P = .047). Conclusion Diffusion-weighted MRI by using quantitative kurtosis variables is superior to apparent diffusion coefficient values in discriminating benign and malignant ovarian lesions and might be of future help in clinical practice, especially in patients with contraindication to contrast media application. © RSNA, 2020 Online supplemental material is available for this article.

Advanced Diffusion-Weighted Abdominal Imaging: Qualitative and Quantitative Comparison of High and Ultra-High b-Values for Lesion Detection and Image Quality.

INTRODUCTION: Magnetic resonance imaging (MRI) of the abdomen increasingly incorporates diffusion-weighted imaging (DWI) sequences. Whereas DWI can substantially aid in detecting and characterizing suspicious findings, it remains unclear to what extent the use of ultra-high b-value DWI might further be of aid for the radiologist especially when using DWI sequences with advanced processing. The target of this study was therefore to compare high and ultra-high b-value DWI in abdominal MRI examinations. METHODS: This institutional review board-approved, prospective study included abdominal MRI examinations of 70 oncologic patients (mean age, 58 years; range, 21-90 years) examined with a clinical 1.5 T MRI scanner (MAGNETOM Aera, Siemens Healthcare, Erlangen, Germany) with an advanced echo planar DWI sequence (b = 0, 50, 900, and 1500 s/mm) after ex vivo phantom and in vivo volunteer investigations. High b900 and ultra-high b1500 DWIs were compared by a qualitative reading for image quality and lesion conspicuity using a 5-point Likert scale with 2 radiologists as readers. The ratios of apparent signal intensities of suspicious lesions/normal tissue of the same organ (LNTRs) were calculated. Appropriate methods were used for statistical analysis, including Wilcoxon signed-rank test and κ statistic for interreader agreement analysis (P < 0.05/0.0125/0.005 after Bonferroni correction). RESULTS: Image quality was significantly increased with b900 as compared with b1500 DWI (P < 0.001) despite using an advanced DWI sequence. A total of 153 suspicious lesions were analyzed. Overall reader confidence for characterization/detection of malignant lesions and, correspondingly, the LNTR (mean, 2.7 ± 1.8 vs 2.4 ± 1.6) were significantly higher with b900 than with b1500 DWI (P < 0.001 and P < 0.001). The increased confidence of lesion recognition and LNTR in the b900 DWI remained significant qualitatively in lymphatic and hepatic lesions and quantitatively in lymphatic, pulmonal, and osseous lesions. CONCLUSIONS: Using high b-value DWI (900 s/mm) provided an improved image quality and also lesion conspicuity as compared with ultra-high b-value DWI (1500 s/mm) in oncologic abdominal examinations despite using advanced processing. Consequently, the value for additional ultra-high b-value DWI in oncologic examinations should be critically evaluated in future studies.

Contrast-to-noise ratio analysis of microscopic diffusion anisotropy indices in q-space trajectory imaging.

Diffusion anisotropy in diffusion tensor imaging (DTI) is commonly quantified with normalized diffusion anisotropy indices (DAIs). Most often, the fractional anisotropy (FA) is used, but several alternative DAIs have been introduced in attempts to maximize the contrast-to-noise ratio (CNR) in diffusion anisotropy maps. Examples include the scaled relative anisotropy (sRA), the gamma variate anisotropy index (GV), the surface anisotropy (UAsurf), and the lattice index (LI). With the advent of multidimensional diffusion encoding it became possible to determine the presence of microscopic diffusion anisotropy in a voxel, which is theoretically independent of orientation coherence. In accordance with DTI, the microscopic anisotropy is typically quantified by the microscopic fractional anisotropy (µFA). In this work, in addition to the µFA, the four microscopic diffusion anisotropy indices (µDAIs) µsRA, µGV, µUAsurf, and µLI are defined in analogy to the respective DAIs by means of the average diffusion tensor and the covariance tensor. Simulations with three representative distributions of microscopic diffusion tensors revealed distinct CNR differences when differentiating between isotropic and microscopically anisotropic diffusion. q-Space trajectory imaging (QTI) was employed to acquire brain in-vivo maps of all indices. For this purpose, a 15min protocol featuring linear, planar, and spherical tensor encoding was used. The resulting maps were of good quality and exhibited different contrasts, e.g. between gray and white matter. This indicates that it may be beneficial to use more than one µDAI in future investigational studies.

Twice-refocused stimulated echo diffusion imaging: Measuring diffusion time dependence at constant T1 weighting.

PURPOSE: Diffusion times longer than 50 ms are typically probed with stimulated-echo sequences. Varying the diffusion time in stimulated-echo sequences affects the T1 weighting of subcompartments, complicating the analysis of diffusion time dependence. Although inversion recovery preparation could be used to change the T1 weighting, it cannot ensure equal T1 weighting at arbitrary mixing times. In this article, a sequence that ensures constant T1 weighting over a wide range of diffusion times is presented. METHODS: The proposed sequence features 2 independent longitudinal storage periods: TM1 and TM2 . Diffusion encoding is performed during TM1 , effectively coupling the diffusion time and TM1 . Equal T1 weighting at arbitrary diffusion times is realized by keeping the total mixing time TM1 + TM2 constant. The sequence was compared with conventional stimulated-echo measurements of diffusion in a 2-compartment phantom consisting of distilled water and paraffinum perliquidum. Additionally, in vivo DTI of the brain was carried out for 8 healthy volunteers with diffusion times ranging from 50 to 500 ms. RESULTS: Diffusion time dependence of the axial and radial diffusivity was detected in the brain. Both sequences resulted in almost identical diffusivities in white matter. In regions containing partial volumes of gray and white matter, a dependency on T1 weighting was observed. CONCLUSION: In accordance with previous studies, little variance of T1 values appeared to be present in healthy white matter. However, this is likely different in diseased tissue. Here, the proposed sequence can be effective in differentiating between diffusion time dependence and T1 weighting effects.

Stimulated echo double diffusion encoded imaging of closed pores: Influence and removal of unbalanced terms.

Nuclear magnetic resonance (NMR) diffusion pore imaging has been proposed to study the shape of arbitrary closed pores filled with an NMR-detectable medium by use of nonclassical diffusion encoding schemes. Potential applications can be found in biomedical imaging and porous media research. When studying non-point-symmetric pores, NMR signals with nonvanishing imaginary parts arise containing the pore shape information, which is lost for classical diffusion encoding schemes. Key limitations are the required high magnetic field gradient amplitudes and T2 relaxation while approaching the diffusion long-time limit. To benefit from the slower T1 decay, we demonstrate the feasibility of diffusion pore imaging with stimulated echoes using Monte Carlo simulations and experiments with hyperpolarized xenon-129 gas in well-defined geometries and show that the necessary complex-valued signals can be acquired. Analytical derivation of the stimulated echo double diffusion encoded signal was performed to investigate the effect of the additionally arising undesired terms on the complex phase information. These terms correspond to signals arising for spin-echo sequences with unbalanced gradients. For most possible applications, the unbalanced terms can be neglected. If non-negligible, selection of the appropriate signal component using a phase cycling scheme was demonstrated experimentally. Using stimulated echoes may be a step towards application of diffusion pore imaging to larger pores with gradient amplitudes available today in preclinical systems.

Classification of Cancer at Prostate MRI: Deep Learning versus Clinical PI-RADS Assessment.

Background Men suspected of having clinically significant prostate cancer (sPC) increasingly undergo prostate MRI. The potential of deep learning to provide diagnostic support for human interpretation requires further evaluation. Purpose To compare the performance of clinical assessment to a deep learning system optimized for segmentation trained with T2-weighted and diffusion MRI in the task of detection and segmentation of lesions suspicious for sPC. Materials and Methods In this retrospective study, T2-weighted and diffusion prostate MRI sequences from consecutive men examined with a single 3.0-T MRI system between 2015 and 2016 were manually segmented. Ground truth was provided by combined targeted and extended systematic MRI-transrectal US fusion biopsy, with sPC defined as International Society of Urological Pathology Gleason grade group greater than or equal to 2. By using split-sample validation, U-Net was internally validated on the training set (80% of the data) through cross validation and subsequently externally validated on the test set (20% of the data). U-Net-derived sPC probability maps were calibrated by matching sextant-based cross-validation performance to clinical performance of Prostate Imaging Reporting and Data System (PI-RADS). Performance of PI-RADS and U-Net were compared by using sensitivities, specificities, predictive values, and Dice coefficient. Results A total of 312 men (median age, 64 years; interquartile range [IQR], 58-71 years) were evaluated. The training set consisted of 250 men (median age, 64 years; IQR, 58-71 years) and the test set of 62 men (median age, 64 years; IQR, 60-69 years). In the test set, PI-RADS cutoffs greater than or equal to 3 versus cutoffs greater than or equal to 4 on a per-patient basis had sensitivity of 96% (25 of 26) versus 88% (23 of 26) at specificity of 22% (eight of 36) versus 50% (18 of 36). U-Net at probability thresholds of greater than or equal to 0.22 versus greater than or equal to 0.33 had sensitivity of 96% (25 of 26) versus 92% (24 of 26) (both P > .99) with specificity of 31% (11 of 36) versus 47% (17 of 36) (both P > .99), not statistically different from PI-RADS. Dice coefficients were 0.89 for prostate and 0.35 for MRI lesion segmentation. In the test set, coincidence of PI-RADS greater than or equal to 4 with U-Net lesions improved the positive predictive value from 48% (28 of 58) to 67% (24 of 36) for U-Net probability thresholds greater than or equal to 0.33 (P = .01), while the negative predictive value remained unchanged (83% [25 of 30] vs 83% [43 of 52]; P > .99). Conclusion U-Net trained with T2-weighted and diffusion MRI achieves similar performance to clinical Prostate Imaging Reporting and Data System assessment. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Padhani and Turkbey in this issue.