Ana E. Rodriguez Soto

Purpose: To develop and evaluate a recurrent neural network that operates directly on undersampled golden-angle radial k-space, enabling low-latency cardiac cine reconstruction for interventional cardiac MRI (iCMR). Methods: KOSTAL (K-space Optimized Spatial Temporal Architecture for Low latency) combines a per-coil k-space convolutional gated recurrent unit (convGRU), which learns a mask-conditioned view-sharing rule over the accumulated k-space, with an image-space recurrent module. It was trained and tested on the open-access multi-coil OCMR dataset using 10 simulated golden-angle spokes per frame against a 1000-spoke reference. Ablations varied recurrent module architecture and placement; temporal behavior was assessed using a imaging-plane transition test. Generalization was evaluated without retraining in 23 pediatric patients prospectively scanned during free-breathing using a radial golden-angle bSSFP sequence with binned XD-GRASP and clinical breath-held cine serving as references. Results: At 10 spokes per frame, KOSTAL reached a median SSIM of 0.907 (gridding: 0.622) in under 34 ms of processing time per frame. Image quality (IQ) matched gridded results formed from 7-11 times more spokes. Recurrence in k-space rather than image space accounted for most of this gain (0.891 versus ≤0.650). IQ recovered from an abrupt plane transition in 75 ms, versus 303 ms for 200-spoke gridding. KOSTAL had the best agreement with the binned reference and the highest long-axis CNR. KOSTAL accurately tracked left-ventricular blood-pool volume (end-diastolic volume bias-0.7 mL) with a shorter lag (71 ms) than iterative reconstruction (239-287 ms). Conclusion: Learned, mask-conditioned recurrence in k-space yields high image quality at a latency compatible with interventional guidance.

Purpose: To quantify the contributions of signal-to-noise ratio (SNR) and velocity-to-encoding ratio (v/VENC) to velocity uncertainty in phase-contrast (PC) MRI and to develop a framework for in vivo voxel-wise uncertainty estimation. Methods: Through-plane 2D PC-MRI of the ascending aorta was acquired using multiple velocity encodings (150, 200, 300 cm/s) and flip angles (0°, 5°, 15°, 20°) to vary v/VENC and SNR. Voxel-wise SNR and velocity uncertainty maps were generated using empirically calibrated phase-noise modeling. Phase-resolved subject-level analyses were performed to quantify the relative contributions of SNR and |v|/VENC to percent velocity uncertainty (%unc). Uncertainty was propagated to flow, stroke volume (SV), and cardiac output (CO). Results: Velocity uncertainty varied substantially across the cardiac cycle and depended on both SNR and |v|/VENC. Across cardiac phases, |v|/VENC accounted for most explained variance in %unc (partial R ² =0.666), while SNR provided a smaller but meaningful contribution (partial R ² =0.287; full R ² =0.909). Near peak systole, SNR contributed more strongly while overall uncertainty remained low. In contrast, diastolic %unc became unstable as velocity approached zero. These effects were most pronounced at low |v|/VENC, where higher VENC settings increased uncertainty despite similar SNR. SV uncertainty ranged from 0.27% to 1.07% across VENC×FA protocols. Conclusion: Velocity uncertainty in PC-MRI depends on both SNR and VENC adequacy in a physiologically phase-dependent manner. Relative uncertainty may become inadequate for precise quantification in low-flow applications, such as diastolic regurgitant jets, despite adequate SNR. Spatiotemporal uncertainty mapping provides a framework for uncertainty-aware PC-MRI acquisition and interpretation.

The aim of this study is to examine cervix‐adapted versions of steady‐state multi‐parameter MRS (SMP MRS) and flip‐angle‐corrected multi‐parameter MRS (CMP MRS), comparing estimated cervix T1 _w and T2 _w for the two sequences. CMP MRS and SMP MRS were adapted from liver versions of the sequences, adding long TR acquisitions to better estimate cervix T1 _w . CMP MRS differs from SMP MRS by correcting for inaccurate B1 calibration. Both CMP MRS and SMP MRS were acquired at 3 T in 13 adult female subjects (10 healthy, 3 with cancer). Values of T1 _w and T2 _w were estimated from both sequences, and the relationship between the values was examined. While there was no significant difference in T1 _w given by the two sequences (CMP T1 _w = 1568 ms, SMP T1 _w = 1571 ms, p = 0.95; SMP T1 _w = 0.657 CMP T1 _w + 541 ms, r = 0.36), there was a single case where SMP MRS underestimated T1 _w by over 400 ms. A significant difference was observed in T2 _w (CMP T2 _w = 39.9 ms, SMP T2 _w = 45.6 ms, p = 0.001; SMP T2 _w = 0.812 CMP T2 _w + 13.3 ms, r = 0.87). Cervix adapted CMP MRS and SMP MRS both successfully estimated values of T1 _w and T2 _w , though the single case where SMP MRS gave a non‐physical T1 _w suggests CMP MRS may be better suited for cervix T1 _w estimation.

Background: Breast cancer screening with dynamic contrast‐enhanced MRI (DCE‐MRI) is recommended for high‐risk women but has limitations, including variable specificity and difficulty in distinguishing cancerous (CL) and high‐risk benign lesions (HRBL) from average‐risk benign lesions (ARBL). Complementary non‐invasive imaging techniques would be useful to improve specificity.Purpose: To evaluate the performance of a previously‐developed breast‐specific diffusion‐weighted MRI (DW‐MRI) model (BS‐RSI3C) to improve discrimination between CL, HRBL, and ARBL in an enriched screening population.Study Type: Prospective.Subjects: Exactly 187 women, either with mammography screening recommending additional imaging (N = 49) or high‐risk individuals undergoing routine breast MRI (N = 138), before the biopsy.Field Strength/Sequence: Multishell DW‐MRI echo planar imaging sequence with a reduced field of view at 3.0 T.Assessment: A total of 72 women had at least one biopsied lesion, with 89 lesions categorized into ARBL, HRBL, CL, and combined CLs and HRBLs (CHRLs). DW‐MRI data were processed to produce apparent diffusion coefficient (ADC) maps, and estimate signal contributions (C₁, C₂, and C₃—restricted, hindered, and free diffusion, respectively) from the BS‐RSI3C model. Lesion regions of interest (ROIs) were delineated on DW images based on suspicious DCE‐MRI findings by two radiologists; control ROIs were drawn in the contralateral breast.Statistical Tests: One‐way ANOVA and two‐sided t‐tests were used to assess differences in signal contributions and ADC values among groups. P‐values were adjusted using the Bonferroni method for multiple testing, P = 0.05 was used for the significance level. Receiver operating characteristics (ROC) curves and intra‐class correlations (ICC) were also evaluated.Results: C₁, √C₁C₂, and were significantly different in HRBLs compared with ARBLs (P‐values < 0.05). The had the highest AUC (0.821) in differentiating CHRLs from ARBLs, performing better than ADC (0.696), especially in non‐mass enhancement (0.776 vs. 0.517).Data Conclusion: This study demonstrated the BS‐RSI3C could differentiate HRBLs from ARBLs in a screening population, and separate CHRLs from ARBLs better than ADC.Level of Evidence: 1.Technical Efficacy Stage: 2.

Background: In breast diffusion‐weighted imaging (DWI), distortion and physiologic artifacts affect clinical interpretation. Image quality can be optimized by addressing the effect of phase encoding (PE) direction on these artifacts.Purpose: To compare distortion artifacts in breast DWI acquired with different PE directions and polarities, and to discuss their clinical implications.Study Type: Prospective.Population: Eleven healthy volunteers (median age: 47 years old; range: 22–74 years old) and a breast phantom.Field Strength/Sequence: Single‐shot echo planar DWI and three‐dimensional fast gradient echo sequences at 3 T.Assessment: All DWI data were acquired with left–right, right–left, posterior–anterior, and anterior–posterior PE directions. In phantom data, displacement magnitude was evaluated by comparing the location of landmarks in anatomical and DWI images. Three breast radiologists (5, 17, and 23 years of experience) assessed the presence or absence of physiologic artifacts in volunteers' DWI datasets and indicated their PE‐direction preference.Statistical Tests: Analysis of variance with post‐hoc tests were used to assess differences in displacement magnitude across DWI datasets and observers. A binomial test and a chi‐squared test were used to evaluate if each in vivo DWI dataset had an equal probability (25%) of being preferred by radiologists. Inter‐reader agreement was evaluated using Gwet's AC1 agreement coefficient. A P‐value P = 0.545). In the in vivo data, there were no physiological artifacts observed in any dataset, regardless of PE direction. In the reader study, there was a significant preference for the posterior–anterior datasets which were selected 94% of the time. There was good agreement between readers (0.936).Data Conclusion: This study showed the impact of PE direction on distortion artifacts in breast DWI. In healthy volunteers, the posterior‐to‐anterior PE direction was preferred by readers.Level of Evidence: 2Technical Efficacy: Stage 1

Background: To date, the accuracy and variability of diffusion‐weighted MRI (DW‐MRI) metrics have been reported in a limited number of scanner/protocol/coil combinations.Purpose: To evaluate the reproducibility of DW‐MRI estimates across multiple scanners and DW‐MRI protocols and to assess the effects of using an 8‐channel vs. 16‐channel breast coil in a breast phantom.Study Type: Prospective.Phantom: Breast phantom containing tubes of water and differing polyvinylpyrrolidone (PVP) concentrations with apparent diffusion coefficients (ADCs) matching breast tissue.Field Strength/Sequence: 3 T (three standard and one wide bore), three DW‐MRI single‐shot echo planar imaging protocols of varying acquired spatial resolution.Assessment: Accuracy of estimated ADCs was assessed using percent differences (PD) relative to nuclear magnetic resonance spectroscopy‐derived reference values. Coefficients of variation (CV) were calculated to determine variation across scanners. CVs based on the median standard deviation (CV_M) were used to evaluate tube‐specific dispersion using 8‐ or 16‐channel coils at a single scanner. ADCs of PVP‐containing tubes were additionally normalized by the median water tube ADC to account for temperature effects.Statistical Tests: Two‐way repeated measures analysis of variance and post hoc tests were used to evaluate differences in ADC, CV, and CV_M across scanners and protocols (α = 0.05).Results: ADCs were within 11% (interquartile range [IQR] 7%) of reference values and significantly improved to 2% (IQR 7%) after normalization to an internal water reference. Normalization significantly reduced interscanner variability of ADC estimates from 7% to 4%. DW‐MRI protocol did not affect ADC accuracy; however, the clinical and higher‐resolution clinical protocols resulted in the greatest (9%) and least (6%) interscanner variability, respectively. The 8‐ and 16‐channel receive coils yielded similar accuracy (PD: 12% vs. 16%) and precision (CV_M: 2.7% vs. 2.9%).Data Conclusion: Normalization by an internal reference improved interscanner ADC reproducibility. High‐resolution protocols yielded comparably accurate and significantly less variable ADCs compared to a clinical standard protocol.Evidence Level: 2Technical Efficacy: Stage 1

Purpose: Restriction spectrum imaging (RSI) decomposes the diffusion‐weighted MRI signal into separate components of known apparent diffusion coefficients (ADCs). The number of diffusion components and optimal ADCs for RSI are organ‐specific and determined empirically. The purpose of this work was to determine the RSI model for breast tissues.Methods: The diffusion‐weighted MRI signal was described using a linear combination of multiple exponential components. A set of ADC values was estimated to fit voxels in cancer and control ROIs. Later, the signal contributions of each diffusion component were estimated using these fixed ADC values. Relative‐fitting residuals and Bayesian information criterion were assessed. Contrast‐to‐noise ratio between cancer and fibroglandular tissue in RSI‐derived signal contribution maps was compared to DCE imaging.Results: A total of 74 women with breast cancer were scanned at 3.0 Tesla MRI. The fitting residuals of conventional ADC and Bayesian information criterion suggest that a 3‐component model improves the characterization of the diffusion signal over a biexponential model. Estimated ADCs of triexponential model were D_1,3 = 0, D_2,3 = 1.5 × 10⁻³, and D_3,3 = 10.8 × 10⁻³ mm²/s. The RSI‐derived signal contributions of the slower diffusion components were larger in tumors than in fibroglandular tissues. Further, the contrast‐to‐noise and specificity at 80% sensitivity of DCE and a subset of RSI‐derived maps were equivalent.Conclusion: Breast diffusion‐weighted MRI signal was best described using a triexponential model. Tumor conspicuity in breast RSI model is comparable to that of DCE without the use of exogenous contrast. These data may be used as differential features between healthy and malignant breast tissues.

Background: Diffusion magnetic resonance imaging (MRI) is integral to detection of prostate cancer (PCa), but conventional apparent diffusion coefficient (ADC) cannot capture the complexity of prostate tissues and tends to yield noisy images that do not distinctly highlight cancer. A four‐compartment restriction spectrum imaging (RSI₄) model was recently found to optimally characterize pelvic diffusion signals, and the model coefficient for the slowest diffusion compartment, RSI₄‐C₁, yielded greatest tumor conspicuity.Purpose: To evaluate the slowest diffusion compartment of a four‐compartment spectrum imaging model (RSI₄‐C₁) as a quantitative voxel‐level classifier of PCa.Study Type: Retrospective.Subjects: Forty‐six men who underwent an extended MRI acquisition protocol for suspected PCa. Twenty‐three men had benign prostates, and the other 23 men had PCa.Field Strength/Sequence: A 3 T, multishell diffusion‐weighted and axial T2‐weighted sequences.Assessment: High‐confidence cancer voxels were delineated by expert consensus, using imaging data and biopsy results. The entire prostate was considered benign in patients with no detectable cancer. Diffusion images were used to calculate RSI₄‐C₁ and conventional ADC. Classifier images were also generated.Statistical Tests: Voxel‐level discrimination of PCa from benign prostate tissue was assessed via receiver operating characteristic (ROC) curves generated by bootstrapping with patient‐level case resampling. RSI₄‐C₁ was compared to conventional ADC for two metrics: area under the ROC curve (AUC) and false‐positive rate for a sensitivity of 90% (FPR₉₀). Statistical significance was assessed using bootstrap difference with two‐sided α = 0.05.Results: RSI₄‐C₁ outperformed conventional ADC, with greater AUC (mean 0.977 [95% CI: 0.951–0.991] vs. 0.922 [0.878–0.948]) and lower FPR₉₀ (0.032 [0.009–0.082] vs. 0.201 [0.132–0.290]). These improvements were statistically significant (P < 0.05).Data Conclusion: RSI₄‐C₁ yielded a quantitative, voxel‐level classifier of PCa that was superior to conventional ADC. RSI classifier images with a low false‐positive rate might improve PCa detection and facilitate clinical applications like targeted biopsy and treatment planning.Evidence Level: 3Technical Efficacy: Stage 2

Background: Diffusion‐weighted (DW) echo‐planar imaging (EPI) is prone to geometric distortions due to B₀ inhomogeneities. Both prospective and retrospective approaches have been developed to decrease and correct such distortions.Purpose: The purpose of this work was to evaluate the performance of reduced‐field‐of‐view (FOV) acquisition and retrospective distortion correction methods in decreasing distortion artifacts for breast imaging. Coverage of the axilla in reduced‐FOV DW magnetic resonance imaging (MRI) and residual distortion were also assessed.Study Type: Retrospective.Population/Phantom: Breast phantom and 169 women (52.4 ± 13.4 years old) undergoing clinical breast MRI.Field Strength/Sequence: A 3.0 T/ full‐ and reduced‐FOV DW gradient‐echo EPI sequence.Assessment: Performance of reversed polarity gradient (RPG) and FSL topup in correcting breast full‐ and reduced‐FOV EPI data was evaluated using the mutual information (MI) metric between EPI and anatomical images. Two independent breast radiologists determined if coverage on both EPI data sets was adequate to evaluate axillary nodes and identified residual nipple distortion artifacts.Statistical Tests: Two‐way repeated‐measures analyses of variance and post hoc tests were used to identify differences between EPI modality and distortion correction method. Generalized linear mixed effects models were used to evaluate differences in axillary coverage and residual nipple distortion.Results: In a breast phantom, residual distortions were 0.16 ± 0.07 cm and 0.22 ± 0.13 cm in reduced‐ and full‐FOV EPI with both methods, respectively. In patients, MI significantly increased after distortion correction of full‐FOV (11 ± 5% and 18 ± 9%, RPG and topup) and reduced‐FOV (8 ± 4% both) EPI data. Axillary nodes were observed in 99% and 69% of the cases in full‐ and reduced‐FOV EPI images. Residual distortion was observed in 93% and 0% of the cases in full‐ and reduced‐FOV images.Data Conclusion: Minimal distortion was achieved with RPG applied to reduced‐FOV EPI data. RPG improved distortions for full‐FOV images but with more modest improvements and limited correction near the nipple.Evidence Level: 3Technical Efficacy: Stage 1

Background: Multicompartmental modeling outperforms conventional diffusion‐weighted imaging (DWI) in the assessment of prostate cancer. Optimized multicompartmental models could further improve the detection and characterization of prostate cancer.Purpose: To optimize multicompartmental signal models and apply them to study diffusion in normal and cancerous prostate tissue in vivo.Study Type: Retrospective.Subjects: Forty‐six patients who underwent MRI examination for suspected prostate cancer; 23 had prostate cancer and 23 had no detectable cancer.Field Strength/Sequence: 3T multishell diffusion‐weighted sequence.Assessment: Multicompartmental models with 2–5 tissue compartments were fit to DWI data from the prostate to determine optimal compartmental apparent diffusion coefficients (ADCs). These ADCs were used to compute signal contributions from the different compartments. The Bayesian Information Criterion (BIC) and model‐fitting residuals were calculated to quantify model complexity and goodness‐of‐fit. Tumor contrast‐to‐noise ratio (CNR) and tumor‐to‐background signal intensity ratio (SIR) were computed for conventional DWI and multicompartmental signal‐contribution maps.Statistical Tests: Analysis of variance (ANOVA) and two‐sample t‐tests (α = 0.05) were used to compare fitting residuals between prostate regions and between multicompartmental models. T‐tests (α = 0.05) were also used to assess differences in compartmental signal‐fraction between tissue types and CNR/SIR between conventional DWI and multicompartmental models.Results: The lowest BIC was observed from the 4‐compartment model, with optimal ADCs of 5.2e‐4, 1.9e‐3, 3.0e‐3, and >3.0e‐2 mm²/sec. Fitting residuals from multicompartmental models were significantly lower than from conventional ADC mapping (P < 0.05). Residuals were lowest in the peripheral zone and highest in tumors. Tumor tissue showed the largest reduction in fitting residual by increasing model order. Tumors had a greater proportion of signal from compartment 1 than normal tissue (P < 0.05). Tumor CNR and SIR were greater on compartment‐1 signal maps than conventional DWI (P < 0.05) and increased with model order.Data Conclusion: The 4‐compartment signal model best described diffusion in the prostate. Compartmental signal contributions revealed by this model may improve assessment of prostate cancer.Level of Evidence 3Technical Efficacy Stage 3J. MAGN. RESON. IMAGING 2021;53:628–639.