Abstract Widespread clinical use of MRI radiomic tumor profiling for prognostication and treatment planning in cancers faces major obstacles due to limitations in standardization of radiomic features. The purpose of the current work was to assess the impact of different MRI scanning- and normalization protocols for the statistical analyses of tumor radiomic data in two patient cohorts with uterine endometrial-(EC) (n = 136) and cervical (CC) (n = 132) cancer. 1.5 T and 3 T, T1-weighted MRI 2 min post-contrast injection, T2-weighted turbo spin echo imaging, and diffusion-weighted imaging were acquired. Radiomic features were extracted from within manually segmented tumors in 3D and normalized either using z-score normalization or a linear regression model (LRM) accounting for linear dependencies with MRI acquisition parameters. Patients were clustered into two groups based on radiomic profile. Impact of MRI scanning parameters on cluster composition and prognostication were analyzed using Kruskal–Wallis tests, Kaplan–Meier plots, log-rank test, random survival forests and LASSO Cox regression with time-dependent area under curve (tdAUC) (α = 0.05). A large proportion of the radiomic features was statistically associated with MRI scanning protocol in both cohorts (EC: 162/385 [42%]; CC: 180/292 [62%]). A substantial number of EC (49/136 [36%]) and CC (50/132 [38%]) patients changed cluster when clustering was performed after z-score-versus LRM normalization. Prognostic modeling based on cluster groups yielded similar outputs for the two normalization methods in the EC/CC cohorts (log-rank test; z-score: p = 0.02/0.33; LRM: p = 0.01/0.45). Mean tdAUC for prognostic modeling of disease-specific survival (DSS) by the radiomic features in EC/CC was similar for the two normalization methods (random survival forests; z-score: mean tdAUC = 0.77/0.78; LRM: mean tdAUC = 0.80/0.75; LASSO Cox; z-score: mean tdAUC = 0.64/0.76; LRM: mean tdAUC = 0.76/0.75). Severe biases in tumor radiomics data due to MRI scanning parameters exist. Z-score normalization does not eliminate these biases, whereas LRM normalization effectively does. Still, radiomic cluster groups after z-score- and LRM normalization were similarly associated with DSS in EC and CC patients.

Segmentation of the spinal tissues on MRI is the basis for quantitative analyses, but time-consuming if done manually. In this work, we construct a pipeline for automatic vertebrae segmentation from T2w MRI scans, assessing performance and generalizability by external validation. Our study used 15 scans from one site (Haukeland University Hospital, HUH) and 10 scans from another (Sahlgrenska University Hospital, SUH). MRI experts manually delineated the vertebral bodies Th12-L5 on all the HUH data and a subset of six scans from SUH. We trained multiple convolutional neural networks, assessing the performance in an experimental design tailored to small-data contexts and also on external data. Our best model achieved a mean Dice score of 0.899. This is comparable to results in the literature, but our system required much less training data. 1 .

BACKGROUND:It is uncertain whether T2-weighted Dixon water images (DixonT2w) and short tau inversion recovery (STIR) are interchangeable when evaluating vertebral bone edema, or if one method is superior or visualizes the edema differently. PURPOSE:To compare image quality and Modic change (MC)-related edema between DixonT2w and STIR and estimate inter-observer reliability for MC edema on DixonT2w. MATERIAL AND METHODS:Consecutive patients (n = 120) considered for the Antibiotics in Modic changes (AIM) trial underwent lumbar 1.5-T magnetic resonance imaging with two-point DixonT2w and STIR. Two radiologists assessed MC-related high-signal lesions on DixonT2w and compared image quality and lesion extent with STIR. Cohen's kappa and mean of differences ± limits of agreement were calculated. RESULTS:Fat suppression and artefacts were similar on DixonT2w and STIR in 116 of 120 (97%) patients. Lesion conspicuity was similar in 88, better on STIR in 10, and better on DixonT2w in 9 of 107 patients with MC-related high-signal lesions. Contrast-to-noise ratio for STIR versus DixonT2w was 19.1 versus 17.1 (mean of differences 2.0 ± 8.2). Of 228 lesions L4-S1, 215 (94%) had similar extent on DixonT2w and STIR, 11 were smaller/undetected on STIR, and two were smaller/undetected on DixonT2w. Lesions missed on STIR (n = 9) or DixonT2w (n = 1) had a weak signal increase on the other sequence (≤17%; 0% = vertebral body, 100% = cerebrospinal fluid). Inter-observer reliability (mean kappa L4-S1) was very good for presence (0.87), moderate for height (0.44), and good for volume (0.63) of lesions on DixonT2w. CONCLUSION:DixonT2w provided similar visualization of MC-related vertebral edema as STIR.

Uterine cervical cancer (CC) is the most common gynecologic malignancy worldwide. Whole-volume radiomic profiling from pelvic MRI may yield prognostic markers for tailoring treatment in CC. However, radiomic profiling relies on manual tumor segmentation which is unfeasible in the clinic. We present a fully automatic method for the 3D segmentation of primary CC lesions using state-of-the-art deep learning (DL) techniques. In 131 CC patients, the primary tumor was manually segmented on T2-weighted MRI by two radiologists (R1, R2). Patients were separated into a train/validation (n = 105) and a test- (n = 26) cohort. The segmentation performance of the DL algorithm compared with R1/R2 was assessed with Dice coefficients (DSCs) and Hausdorff distances (HDs) in the test cohort. The trained DL network retrieved whole-volume tumor segmentations yielding median DSCs of 0.60 and 0.58 for DL compared with R1 (DL-R1) and R2 (DL-R2), respectively, whereas DSC for R1-R2 was 0.78. Agreement for primary tumor volumes was excellent between raters (R1-R2: intraclass correlation coefficient (ICC) = 0.93), but lower for the DL algorithm and the raters (DL-R1: ICC = 0.43; DL-R2: ICC = 0.44). The developed DL algorithm enables the automated estimation of tumor size and primary CC tumor segmentation. However, segmentation agreement between raters is better than that between DL algorithm and raters.

Abstract Background Modic Changes (MCs) in the vertebral bone marrow were related to back pain in some studies but have uncertain clinical relevance. Diffusion weighted MRI with apparent diffusion coefficient (ADC)-measurements can add information on bone marrow lesions. However, few have studied ADC measurements in MCs. Further studies require reproducible and valid measurements. We expect valid ADC values to be higher in MC type 1 (oedema type) vs type 3 (sclerotic type) vs type 2 (fatty type). Accordingly, the purpose of this study was to evaluate ADC values in MCs for interobserver reproducibility and relation to MC type. Methods We used ADC maps (b 50, 400, 800 s/mm2) from 1.5 T lumbar spine MRI of 90 chronic low back pain patients with MCs in the AIM (Antibiotics In Modic changes)-study. Two radiologists independently measured ADC in fixed-sized regions of interests. Variables were MC-ADC (ADC in MC), MC-ADC% (0% = vertebral body, 100% = cerebrospinal fluid) and MC-ADC-ratio (MC-ADC divided by vertebral body ADC). We calculated mean difference between observers ± limits of agreement (LoA) at separate endplates. The relation between ADC variables and MC type was assessed using linear mixed-effects models and by calculating the area under the receiver operating characteristic curve (AUC). Results The 90 patients (mean age 44 years; 54 women) had 224 MCs Th12-S1 comprising type 1 (n = 111), type 2 (n = 91) and type 3 MC groups (n = 22). All ADC variables had higher predicted mean for type 1 vs 3 vs 2 (p < 0.001 to 0.02): MC-ADC (10− 6 mm2/s) 1201/796/576, MC-ADC% 36/21/14, and MC-ADC-ratio 5.9/4.2/3.1. MC-ADC and MC-ADC% had moderate to high ability to discriminate between the MC type groups (AUC 0.73–0.91). MC-ADC-ratio had low to moderate ability (AUC 0.67–0.85). At L4-S1, widest/narrowest LoA were for MC-ADC 20 ± 407/12 ± 254, MC-ADC% 1.6 ± 18.8/1.4 ± 10.4, and MC-ADC-ratio 0.3 ± 4.3/0.2 ± 3.9. Difference between observers > 50% of their mean value was less frequent for MC-ADC (9% of MCs) vs MC-ADC% and MC-ADC-ratio (17–20%). Conclusions The MC-ADC variable (highest mean ADC in the MC) had best interobserver reproducibility, discriminated between MC type groups, and may be used in further research. ADC values differed between MC types as expected from previously reported MC histology.

Imagedata is a python library to read and write medical image data into Series objects.In particular, imagedata will read, sort and write DICOM ® 3D and 4D series based on defined attributes.As far as possible, imagedata will handle geometry information between the medical image data formats like DICOM, NIfTI (Cox et al., 2004) and ITK (Yoo et al., 2002). Imagedata provides a Series class inheriting thenumpy.ndarray class (Harris et al., 2020), adding DICOM data structures.Plugins provide functions to import and export DICOM and other data formats.The DICOM plugin can read complete series, sorting the data as requested into multidimensional arrays.Input and output data can be accessed on various locations, including local files, DICOM servers and XNAT servers (Marcus et al., 2007).The Series class enables NumPy and derived libraries (like SciPy (Virtanen et al., 2020)) to work on medical images, simplifying input and output.A feature is the conversion between different image formats.E.g., a pipeline based on a clinical DICOM series can be converted to NIfTI, processed by some NIfTI-based tool (e.g.FSL (Smith et al., 2004)).Finally, the result can be converted back to DICOM and stored as a new series in PACS (Picture Archive and Communication system).A viewer is included, allowing the display of a stack of images, including modifying window width and centre, and scrolling through 3D and 4D image stacks.A region of interest (ROI) can be drawn, resulting in a mask as a NumPy ndarray, or as an outline.

Purpose: The composite vascular transport function of a brain voxel consists of one convolutional component for the arteries, one for the capillaries and one for the veins in the voxel of interest. Here, the goal is to find each of these three convolutional components and the associated arterial input function. Pharmacokinetic modelling: The single voxel vascular transport functions for arteries, capillaries and veins were all modelled as causal exponential functions. Each observed multipass tissue contrast function was as a first approximation modelled as the resulting parametric composite vascular transport function convolved with a nonparametric and voxel specific multipass arterial input function. Subsequently, the residue function was used in the true perfusion equation to optimize the three parameters of the exponential functions. Deconvolution methods: For each voxel, the parameters of the three exponential functions were estimated by successive iterative blind deconvolutions using versions of the Lucy-Richardson algorithm. The final multipass arterial input function was then computed by nonblind deconvolution using the Lucy-Richardson algorithm and the estimated composite vascular transport function. Results: Simulations showed that the algorithm worked. The estimated mean transit time of arteries, capillaries and veins of the simulated data agreed with the known input values. For real data, the estimated capillary mean transit times agreed with known values for this parameter. The nonparametric multipass arterial input functions were used to derive the associated map of the arrival time. The arrival time map of a healthy volunteer agreed with known arterial anatomy and physiology. Conclusion: Clinically important new voxelwise hemodynamic information for arteries, capillaries and veins separately can be estimated using multipass tissue contrast functions and the iterative blind Lucy-Richardson deconvolution algorithm.

The temporomandibular joint (TMJ) is typically involved in 45-87% of children with Juvenile Idiopathic Arthritis (JIA). Accurate diagnosis of JIA is difficult as various clinical tests, including MRI, disagree. The purpose of this study is to optimize the methodological aspects of Dynamic Contrast Enhanced (DCE) MRI of the TMJ in children. In this cross-sectional study, including data from 73 JIA affected children, aged 6-15 years, effects of motion correction, sampling rate and parametric modelling on DCE-MRI data is investigated. Consensus among three radiologists determined the regions of interest. Quantitative perfusion parameters were estimated using four perfusion models; the Adiabatic Approximation to Tissue Homogeneity (AATH), Distributed Capillary Adiabatic Tissue Homogeneity (DCATH), Gamma Capillary Transit Time (GCTT) and Two Compartment Exchange (2CXM) models. Effects of motion correction were evaluated by a sum of least squares between corrected raw data and the GCTT model. The effect of systematically down sampling the raw data was tested. The sum of least squares was computed across all pharmacokinetic models. Relative difference perfusion parameters between the left and right TMJ were used for an unsupervised k-means based stratification of the data based on a principal component analysis, as well as for a supervised random forest classification. Diagnostic sensitivity and specificity were computed relative to structural image scorings. Paired sample t-tests, as well as ANOVA tests, were used (significant threshold: p < 0.05) with Tukeys post hoc test. High-level elastic motion correction provides the best least square fit to the GCTT model (percental improvement: 72-84%). A 4 s sampling rate captures more of the potentially disease relevant signal variations. The various parametric models all leave comparable residues (relative standard deviation: 3.4%). In further evaluation of DCE-MRI as a potential diagnostic tool for JIA a high-level elastic motion correction scheme should be adopted, with a sampling rate of at least 4 s. Results suggest that DCE-MRI data can be a valuable part in JIA diagnostics in the TMJ.

BACKGROUND:To investigate whether magnetic resonance (MR) radiomic features combined with machine learning may aid in predicting extraprostatic extension (EPE) in high- and non-favorable intermediate-risk patients with prostate cancer.PURPOSE:To investigate the diagnostic performance of radiomics to detect EPE.MATERIAL AND METHODS:MR radiomic features were extracted from 228 patients, of whom 86 were diagnosed with EPE, using prostate and lesion segmentations. Prediction models were built using Random Forest. Further, EPE was also predicted using a clinical nomogram and routine radiological interpretation and diagnostic performance was assessed for individual and combined models.RESULTS:The MR radiomic model with features extracted from the manually delineated lesions performed best among the radiomic models with an area under the curve (AUC) of 0.74. Radiology interpretation yielded an AUC of 0.75 and the clinical nomogram (MSKCC) an AUC of 0.67. A combination of the three prediction models gave the highest AUC of 0.79.CONCLUSION:Radiomic analysis combined with radiology interpretation aid the MSKCC nomogram in predicting EPE in high- and non-favorable intermediate-risk patients.

Objective: Chronic kidney disease (CKD) is a serious medical condition characterized by gradual loss of kidney function. Early detection and diagnosis is mandatory for adequate therapy and prognostic improvement. Hence, in the current pilot study we explore the use of image registration methods for detecting renal morphologic changes in patients with CKD. Methods: Ten healthy volunteers and nine patients with presumed CKD underwent dynamic T1 weighted imaging without contrast agent. From real and simulated dynamic time series, kidney deformation fields were estimated using a poroelastic deformation model. From the deformation fields several quantitative parameters reflecting pressure gradients, and volumetric and shear deformations were computed. Eight of the patients also underwent a kidney biopsy as a gold standard. Results: We found that the absolute deformation, normalized volume changes, as well as pressure gradients correlated significantly with arteriosclerosis from biopsy assessments. Furthermore, our results indicate that current image registration methodologies are lacking sensitivity to recover mild changes in tissue stiffness. Conclusion: Image registration applied to dynamic time series correlated with structural renal changes and should be further explored as a tool for invasive measurements of arteriosclerosis. Significance: Under the assumption that the proposed framework can be further developed in terms of sensitivity and specificity, it can provide clinicians with a non-invasive tool of high spatial coverage available for characterization of arteriosclerosis and potentially other pathological changes observed in chronic kidney disease.

OBJECTIVE:An extension of single- and multi-channel blind deconvolution is presented to improve the estimation of the arterial input function (AIF) in quantitative dynamic contrast enhanced magnetic resonance imaging (DCE-MRI).METHODS:The Lucy-Richardson expectation-maximization algorithm is used to obtain estimates of the AIF and the tissue residue function (TRF). In the first part of the algorithm, nonparametric estimates of the AIF and TRF are obtained. In the second part, the decaying part of the AIF is approximated by three decaying exponential functions with the same delay, giving an almost noise free semi-parametric AIF. Simultaneously, the TRF is approximated using the adiabatic approximation of the Johnson-Wilson (aaJW) pharmacokinetic model.RESULTS:In simulations and tests on real data, use of this AIF gave perfusion values close to those obtained with the corresponding previously published nonparametric AIF, and are more noise robust.CONCLUSION:When used subsequently in voxelwise perfusion analysis, these semi-parametric AIFs should give more correct perfusion analysis maps less affected by recording noise than the corresponding nonparametric AIFs, and AIFs obtained from arteries.SIGNIFICANCE:This paper presents a method to increase the noise robustness in the estimation of the perfusion parameter values in DCE-MRI.

OBJECTIVE:Estimation of renal filtration using dynamic contrast-enhanced imaging (DCE-MRI) requires a series of analysis steps. The possible number of distinct post-processing chains is large and grows rapidly with increasing number of processing steps or options. In this study we introduce a framework for systematic evaluation of the post-processing chains. The framework is later used to highlight the workflow processing chain sensitivity towards accuracy in estimation of glomerular filtration rate (GFR).METHODS:Twenty healthy volunteers underwent DCE-MRI examinations as well as iohexol clearance for reference GFR measurements. In total, 692 different combinations of post-processing steps were explored for analysis, including options for kidney segmentation, B1 inhomogeneity correction, placement of arterial input function, gadolinium concentration estimation as well as handling of motion-corrupted volumes and breathing motion. The evaluation of various processing chains is presented using a classification tree framework and random forest ensemble learning.RESULTS:Among the processing steps subject to testing, methods for calculating the gadolinium concentration as well as B1 inhomogeneity correction had the largest impact on accuracy of GFR estimations. Different segmentation methods did not play an important role in the post-processing of the MR data except from one processing chain where the automated segmentation outperformed the manual segmentation.CONCLUSION:The proposed classification trees were efficiently used as a statistical tool for visualization and communication of results to distinguish between important and less influential processing steps in renal DCE-MRI. We also identified several crucial factors in the processing chain.

Background High repeatability, accuracy, and precision for renal function measurements need to be achieved to establish renal dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) as a clinically useful diagnostic tool. Purpose To investigate the repeatability, accuracy, and precision of DCE-MRI measured renal perfusion and glomerular filtration rate (GFR) using iohexol–GFR as the reference method. Material and Methods Twenty healthy non-smoking volunteers underwent repeated DCE-MRI and an iohexol–GFR within a period of 10 days. Single-kidney (SK) MRI measurements of perfusion (blood flow, Fb) and filtration (GFR) were derived from parenchymal intensity time curves fitted to a two-compartment filtration model. The repeatability of the SK-MRI measurements was assessed using coefficient of variation (CV). Using iohexol–GFR as reference method, the accuracy of total MR–GFR was determined by mean difference (MD) and precision by limits of agreement (LoA). Results SK-Fb (MR1, 345 ± 84; MR2, 371 ± 103 mL/100 mL/min) and SK–GFR (MR1, 52 ± 14; MR2, 54 ± 10 mL/min/1.73 m2) measurements achieved a repeatability (CV) in the range of 15–22%. With reference to iohexol–GFR, MR–GFR was determined with a low mean difference but high LoA (MR1, MD 1.5 mL/min/1.73 m2, LoA [−42, 45]; MR2, MD 6.1 mL/min/1.73 m2, LoA [−26, 38]). Eighty percent and 90% of MR–GFR measurements were determined within ± 30% of the iohexol–GFR for MR1 and MR2, respectively. Conclusion Good repeatability of SK-MRI measurements and good agreement between MR–GFR and iohexol–GFR provide a high clinical potential of DCE-MRI for renal function assessment. A moderate precision in MR-derived estimates indicates that the method cannot yet be used in clinical routine.

OBJECTIVE. The objective of our study was to investigate whether dynamic contrast-enhanced MRI (DCE-MRI) can detect differences and potential adaption in single-kidney parenchymal volume, blood flow, glomerular filtration rate (GFR), and filtration fraction in the remaining kidney of healthy donors compared with nondonors. Further, we evaluated the agreement in donor GFRs measured using DCE-MRI versus serum clearance of iohexol.SUBJECTS AND METHODS. Twenty living kidney donors and 20 healthy control subjects underwent DCE-MRI and iohexol GFR. Renal parenchymal volume was assessed from maximum-signal-intensity maps. Single-kidney MRI measurements of blood flow and GFR were derived from parenchymal signal intensity-time curves fitted to a two-compartment filtration model. The Student t test, Pearson correlation coefficient, mean differences, and limits of agreement were applied to analyze MRI measurements between groups and agreement with iohexol GFR.RESULTS. MRI findings showed significantly higher blood flow (difference in mean values of donors vs control subjects, 54%; p = 0.001), GFR (78%, p < 0.0001), and renal parenchymal volume (65%, p < 0.0001) in the single kidney of donors compared with the single kidney of healthy control subjects. In the donors, a proportional increase in blood flow and GFR resulted in a comparable filtration fraction, as was observed in the control subjects. Significant correlations were found between MRI-derived GFR and parenchymal volume (p < 0.0016) as well as with iohexol GFR (p < 0.0001). The mean difference between MRIderived GFR and iohexol GFR was 14.0 mL/min, and the limits of agreement between MRIderived GFR and iohexol GFR were -24.1 and 52.1 mL/min.CONCLUSION. DCE-MRI-derived values for single-kidney function and volume in kidney donors were significantly higher than those in control subjects and suggest a future potential benefit of DCE-MRI for diagnostic and prognostic structural and functional assessments in living kidney donors.

OBJECTIVE. The purpose of this article is to compare two 3D dynamic contrast-enhanced (DCE) MRI measurement techniques for MR renography, a radial k-space weighted image contrast (KWIC) sequence and a cartesian FLASH sequence, in terms of intrasubject differences in estimates of renal functional parameters and image quality characteristics. SUBJECTS AND METHODS. Ten healthy volunteers underwent repeated breath-hold KWIC and FLASH sequence examinations with temporal resolutions of 2.5 and 2.8 seconds, respectively. A two-compartment model was used to estimate MRI-derived perfusion parameters and glomerular filtration rate (GFR). The latter was compared with the iohexol GFR and the estimated GFR. Image quality was assessed using a visual grading characteristic analysis of relevant image quality criteria and signal-to-noise ratio calculations. RESULTS. Perfusion estimates from FLASH were closer to literature reference values than were the KWIC sequences. In relation to the iohexol GFR (mean [± SD], 103 ± 11 mL/min/1.73 m(2)), KWIC produced significant underestimations and larger bias in GFR values (mean, 70 ± 30 mL/min/1.73 m(2); bias = -33.2 mL/min/1.73 m(2)) compared with the FLASH GFR (110 ± 29 mL/min/1.73 m(2); bias = 6.4 mL/min/1.73 m(2)). KWIC was statistically significantly (p < 0.005) more impaired by artifacts than was FLASH (AUC = 0.18). The average signal-enhancement ratio (delta ratio) in the cortex was significantly lower for KWIC (delta ratio = 0.99) than for FLASH (delta ratio = 1.40). Other visually graded image quality characteristics and signal-to-noise ratio measurements were not statistically significantly different. CONCLUSION. Using the same postprocessing scheme and pharmacokinetic model, FLASH produced more accurate perfusion and filtration parameters than did KWIC compared with clinical reference methods. Our data suggest an apparent relationship between image quality characteristics and the degree of stability in the numeric model-based renal function estimates.

In dynamic contrast-enhanced magnetic resonance imaging, there has been no consensus in the choice of the pharmacokinetic model. In this paper, a new approach for assessment of the most realistic model for a given tissue is presented. Non-blind and single-channel blind deconvolution algorithms were used in quantitative magnetic resonance dynamic contrast-enhanced imaging of the mouse masseter muscle to compare the realism of two different pharmacokinetic models for the tissue residue function. The first was the adiabatic approximation tissue homogeneity model (aaJW) and the second, the two-compartment exchange model (2CXM). Normals and mice treated with the substance C48/80 were studied. C48/80 increases both blood flow and contrast leakage in muscle substantially. The obtained approximation accuracy was evaluated for both pharmacokinetic models. In addition, the arterial input functions (aifs) estimated using blind deconvolution were compared to the corresponding observed aifs. The hypothesis is that the most realistic model of the tissue residue function leads to the best fits. The non-blind deconvolution did not result in any clear answer. For blind deconvolution, the aifs of the aaJW model were very similar to the corresponding observed aifs, and clearly more so than the aifs of the 2CXM model. Also, the approximation of the observed tracer time sequences was more accurate for the aaJW than the 2CXM model. The realism of different pharmacokinetic models in describing the passage of a tracer through a microvascular bed of a single tissue could be assessed using single-channel blind deconvolution.

Background The use of multiparametric magnetic resonance imaging (mpMRI) to detect and localize prostate cancer has increased in recent years. In 2010, the European Society of Urogenital Radiology (ESUR) published guidelines for mpMRI and introduced the Prostate Imaging Reporting and Data System (PI-RADS) for scoring the different parameters. Purpose To evaluate the reliability and diagnostic performance of endorectal 1.5-T mpMRI using the PI-RADS to localize the index tumor of prostate cancer in patients undergoing prostatectomy. Material and Methods This institutional review board IRB-approved, retrospective study included 63 patients (mean age, 60.7 years, median PSA, 8.0). Three observers read mpMRI parameters (T2W, DWI, and DCE) using the PI-RADS, which were compared with the results from whole-mount histopathology that analyzed 27 regions of interest. Inter-observer agreement was calculated as well as sensitivity, specificity, positive predictive value (PPV), and negative predicted value (NPV) by dichotomizing the PI-RADS criteria scores ≥3. A receiver-operating curve (ROC) analysis was performed for the different MR parameters and overall score. Results Inter-observer agreement on the overall score was 0.41. The overall score in the peripheral zone achieved sensitivities of 0.41, 0.60, and 0.55 with an NPV of 0.80, 0.84, and 0.83, and in the transitional zone, sensitivities of 0.26, 0.15, and 0.19 with an NPV of 0.92, 0.91, and 0.92 for Observers 1, 2, and 3, respectively. The ROC analysis showed a significantly increased area under the curve (AUC) for the overall score when compared to T2W alone for two of the three observers. Conclusion 1.5 T mpMRI using the PI-RADS to localize the index tumor achieved moderate reliability and diagnostic performance.

Multipass dynamic MRI and pharmacokinetic modeling are used to estimate perfusion parameters of leaky capillaries. Curve fitting and nonblind deconvolution are the established methods to derive the perfusion estimates from the observed arterial input function (AIF) and tissue tracer concentration function. These nonblind methods are sensitive to errors in the AIF, measured in some nearby artery or estimated by multichannel blind deconvolution. Here, a single-channel blind deconvolution algorithm is presented, which only uses a single tissue tracer concentration function to estimate the corresponding AIF and tissue impulse response function. That way, many errors affecting these functions are reduced. The validity of the algorithm is supported by simulations and tests on real data from mouse. The corresponding nonblind and multichannel methods are also presented.