Purpose: To develop and cross-dataset validate a hybrid ResNet50–Transformer U-Net for automated segmentation of breast cancer (BC) lesions on 2D dynamic contrast-enhanced MRI (DCE-MRI).

Methods: A total of 20434 BC MRI images from the BC MRI Segmentation Benchmark (BC-MRI-SEG), comprising 4 independent public datasets, were included. A strict patient-level data splitting strategy was applied to avoid data leakage. The proposed model was based on a U-Net architecture integrating a ResNet50 encoder for local feature extraction and a bottleneck Transformer module for global contextual modeling. Model performance was evaluated using a leave-one-dataset-out cross-dataset external validation strategy. Segmentation performance was assessed using the Dice similarity coefficient (DSC), intersection over union (IoU), 95% Hausdorff distance (HD95), sensitivity, and precision.

Results: The proposed hybrid model achieved stable segmentation performance across the evaluated datasets. In leave-one-dataset-out external validation, patient-level DSC ranged from 0.75 ± 0.13 to 0.79 ± 0.10, while IoU ranged from 0.61 ± 0.15 to 0.66 ± 0.12. In the fixed held-out comparison cohort, the proposed model achieved a DSC of 0.82 ± 0.08 and an IoU of 0.70 ± 0.10 and outperformed the evaluated baseline models (Holm-adjusted P ≤ 0.021). Qualitative analysis showed generally close agreement between the predicted masks and reference annotations across lesions with different morphologic appearances.

Conclusion: The proposed ResNet–Transformer-based U-Net framework provided a robust approach for automated BC MRI segmentation across heterogeneous public datasets. These findings suggest that combining local feature extraction with global contextual modeling may be useful for computer-assisted breast MRI analysis and related clinical workflows.

Purpose:To investigate the effects of artificial intelligence (AI)–based reconstruction on image quality and voxel-based morphometry (VBM)–based atrophy analysis in 3D Magnetization Prepared Rapid Gradient Echo (MPRAGE).

Methods:Ten healthy volunteers underwent 3T MPRAGE imaging using varying acceleration factors with sensitivity encoding (SENSE) (SE; 2,3), compressed sensing (CS; 2,3,5,8), and AI-based reconstruction (AI; 2,3,5,8). Quantitative assessment included SNR and contrast-to-noise ratio (CNR) in both superficial and deep brain regions. Visual assessment of image quality was performed by 2 experienced neuroradiologists. VBM-based Z-score analysis of regional atrophy was performed using Voxel-based Specific Regional Analysis System for Alzheimer’s Disease (VSRAD). Images acquired using a widely recognized standard imaging protocol, SENSE acceleration factor 2 with 1.0-mm isotropic voxels, were used as the reference, and various reconstruction methods and acceleration factors were compared.

Results:Conventional SENSE and CS reconstructions showed stepwise decreases in SNR and CNR with increasing acceleration factors. SNR decrease was particularly pronounced in deep brain regions. Images with AI-based reconstruction maintained relatively stable SNR and CNR across acceleration factors, showing consistent performance in both superficial and deep regions. Visual assessment confirmed reduced noise in images with AI-based reconstruction, which were rated as more favorable for interpretation than SE or CS images. VSRAD analysis demonstrated high correlation with reference images and minimal systematic bias across all acceleration conditions.

Conclusion:AI-based reconstruction may enable faster MPRAGE acquisition (up to 5–fold acceleration) while preserving image quality and the reproducibility of VBM-based atrophy analysis.

Purpose: To evaluate whether multi-pool model (MPM)–based amide proton transfer (APT) imaging parameters can assess histological grade, histological type, lymphovascular space invasion (LVSI), and recurrence risk in endometrial carcinoma (EC), relative to conventional apparent diffusion coefficient (ADC) and magnetization transfer ratio asymmetry (MTRasym).

Methods: Preoperative 3.0T MRI was performed in 27 patients with EC at Kagawa University Hospital. Two-dimensional axial APT imaging was conducted using a single-shot fast-spin echo acquisition at 3T. ADC, MTRasym (3.5 ppm), and MPM-based APT imaging parameters were calculated using ROI analysis. Tumor grade, histological type, and LVSI status were obtained from surgical specimens. Recurrence risk was defined according to the Japan Society of Gynecologic Oncology 2023 guidelines. Between-group differences were tested using independent-samples t-tests, Mann–Whitney U tests, one-way analysis of variance, or Kruskal–Wallis tests as appropriate. Receiver operating characteristic (ROC) analysis was performed to evaluate diagnostic performance.

Results: APT_T2 was significantly higher in grade 2 than in grade 1 ECs (P < 0.05) and in type II than in type I ECs (P < 0.05), yielding the highest area under the curve (AUC) values of 0.81 and 0.83 for differentiating tumor grade and histological type, respectively. APT_T1 was significantly higher in LVSI-positive tumors than in LVSI-negative tumors (P < 0.05) and showed good performance for predicting LVSI (AUC = 0.84). APT_T1 was also significantly higher in the intermediate-risk group than in the low-risk group and achieved an AUC of 0.78 for differentiating low-risk from non–low-risk groups. MTRasym was likewise significantly higher in the non–low-risk group (P < 0.05) and showed comparable diagnostic performance. In contrast, ADC metrics showed no significant differences across these comparisons.

Conclusion: These preliminary findings indicate that APT_T2 may be useful for assessing histological grade and type, whereas APT_T1 may provide complementary information for evaluating LVSI and stratifying recurrence risk in EC.

Purpose: Atrial function is associated with exercise tolerance and survival in patients with Fontan circulation. Although atrial function is often evaluated using strain parameters, changes in atrial pressure are also important but have not been well studied. This study aimed to investigate the atrial pressure–strain relationship in patients with Fontan circulation using data from cardiac catheterization and MRI.

Methods: We retrospectively included 16 patients (median age: 12 years) who underwent cardiac MRI between 2020 and 2026, and 10 control patients (median age: 8.5 years). Atrial strain values were evaluated from 4-chamber cine MRI images. Pulmonary artery wedge pressure waveforms were used as a surrogate for atrial pressure. The pressure–strain index was calculated as an integrated measure of atrial pressure and strain.

Results: The 2 groups showed no significant differences in age or body size. Patients with Fontan circulation had higher central venous pressure (CVP) and a lower cardiac index (CVP: 11.5 vs. 4.0 mmHg, P < 0.001; cardiac index: 2.9 vs. 3.8 L/min/m², P = 0.007). Reservoir and conduit strains were lower in the Fontan group (reservoir strain: 16.7 vs. 26.2%, P < 0.001; conduit strain: 6.2 vs. 18.7%, P < 0.001); however, no significant difference was observed in pump strain between the groups. No significant difference was also noted in pressure–strain index between the 2 groups (0.2 vs. 0.2, P = 0.417).

Conclusion: In patients with Fontan physiology, reservoir and conduit strain were lower than in controls. In contrast, atrial pressure–strain relationship was not clearly associated with conventional Fontan hemodynamic parameters. These findings suggest that the physiological significance of atrial pressure–strain relationship in the Fontan circulation remains uncertain.

Purpose: Multiple sclerosis (MS) is a demyelinating disease of the central nervous system influenced by both genetic and environmental factors. Although a polygenic risk score for MS (MS-PRS) is associated with disease susceptibility, its relationship with white matter (WM) microstructure and cognition in neurologically healthy individuals remains unclear. This study investigated associations of MS-PRS with diffusion MRI metrics and cognitive function in healthy individuals.

Methods: This cross-sectional study included 35952 neurologically healthy UK Biobank participants with available MS-PRS, diffusion MRI, and cognitive data. Associations between MS-PRS and 11 cognitive measures were tested using generalized linear models adjusted for age, age squared, sex, genetic principal components, body mass index, and intracranial volume, with false discovery rate (FDR) correction. Participants were then stratified by MS-PRS quartiles. For cognitive measures showing significant associations in the highest MS-PRS group, follow-up analyses evaluated associations between MS-PRS and diffusion MRI metrics in WM regions previously associated with reaction time (RT). Mediation analysis was then performed.

Results: In the full sample, MS-PRS was not significantly associated with any cognitive measure after multiple-comparison correction. In the highest MS-PRS group, the higher MS-PRS was significantly associated with longer RT (β = 0.048, FDR-corrected P = 0.047). In the highest MS-PRS group, the higher MS-PRS was associated with higher mean diffusivity (MD) in the splenium of the corpus callosum (CC) (β = 0.021, FDR-corrected P = 0.024). MD in the splenium of the CC showed a small partial mediation effect on the association between MS-PRS and RT in the highest MS-PRS group (mediation effect 3.0%).

Conclusion: In neurologically healthy individuals, a higher genetic burden for MS was associated with slower information processing speed in the highest-risk subgroup and with altered WM microstructure in the splenium of the CC. These findings suggest that subtle WM differences may partly underlie the association between MS genetic susceptibility and cognitive performance.

Purpose: Pseudo-continuous arterial spin labeling (pCASL) MRI is a widely used, noninvasive, contrast-agent-free technique for measuring cerebral blood flow (CBF) and assessing vascular dysfunction across diverse clinical settings and murine disease models. In practice, arterial-transit artifacts that generate hyperintense signal in CBF maps warrant careful consideration. While these effects are well characterized in humans, they are less well understood in mice owing to the marked interspecies physiological differences.

Methods: To address this knowledge gap, we systematically characterized pCASL hyperintense signal as a function of post-labeling delay (PLD) and crusher-gradient strength in mice. Numerical simulations were also performed to validate the experimental findings.

Results: We found that hyperintense signals in mice extend to arteries, major veins, and ventricular structures. Such a pattern was different from human pCASL images, where hyperintense signals are predominantly present in arteries. Statistical analyses supported a PLD of 500 milliseconds as a pragmatic balance between detection sensitivity and suppression of vascular contamination. Additional experiments and numerical simulations showed that, within the tested range, stronger crusher gradients provided little extra vascular suppression—primarily because large vessel calibers relative to small voxels limit intravoxel phase dispersion.

Conclusion: These findings refine the interpretation of murine pCASL signals and facilitate more accurate perfusion imaging in preclinical pathophysiological studies.

Purpose: To develop and externally validate a non-invasive framework for quantifying brain amyloid-β (Aβ) deposition using magnetic resonance fingerprinting (MRF) and neural network–based decoding, with positron emission tomography (PET) as the reference standard.

Methods: This prospective multi-site study included 44 participants from 2 sites who had undergone, or were scheduled to undergo, Aβ PET within 1 year. MRF was performed on a 3T MR system using a 2D fast imaging with steady-state precession sequence with B₁ correction, covering the whole brain in 9.5 min. PET images were co-registered to the MRF space, and regional amyloid load was calculated using an automated template–based pipeline. An inverse mapping function was implemented to convert MRF signals into amyloid burden maps. Repeatability, agreement with PET-based centiloid values, and associations with cognitive scores were evaluated.

Results: The generated amyloid maps were visually similar to PET images. Test-retest analysis showed high repeatability, with a coefficient of variation of 1.8 ± 1.3% and an intraclass correlation coefficient of 0.84. In the external test set, MRF-based measurements correlated significantly with PET centiloid scores (Spearman’s ρ = 0.589, P = 0.015) and Montreal Cognitive Assessment scores (ρ = −0.543, P = 0.020).

Conclusion: The proposed framework enables non-invasive Aβ mapping using a clinically feasible MRI protocol and may support repeated assessment for monitoring during anti-amyloid treatment.

Purpose: Synthetic Q-space learning (synQSL), which uses only synthetic data in the training of regressors, has demonstrated promising results in parameter estimation for brain diffusion MRI (dMRI). This study aimed to evaluate the performance of synQSL in breast dMRI for intravoxel incoherent motion–diffusional kurtosis imaging (IVIM–DKI) parameter estimation, with comparing several types of regressors and conventional fitting by nonlinear least squares fitting (NL-LSF).

Methods: In synthesizing data for synQSL, IVIM–DKI parameters were sampled from uniform distributions and substituted into the signal model along with b-values to generate diffusion-weighted imaging (DWI) signals. In addition, Rician noise was mixed to the signals. We prepared datasets of 10⁵ and 10⁶ samples, and trained multi-layer perceptron (MLP), Kolmogorov–Arnold networks (KAN), and random forest (RF) regressors for synQSL. The performance of the parameter estimation methods including NL-LSF was evaluated using a digital phantom including various value combinations of IVIM–DKI parameters and clinical data of 67 cases (13 benign and 54 malignant lesions) through quantitative analysis and visual assessment.

Results: In the digital phantom, the regressors of synQSL achieved significantly lower root mean square error (RMSE) for f, D^∗, and D than NL-LSF (P < 0.05) with less variation among the parameter sets. In clinical datasets, synQSL not only improved the visual quality of parameter maps but also showed significant differences between benign and malignant lesions in parameters while the NL-LSF failed. Furthermore, the estimation times for all synQSL regressors were substantially shorter than that of NL-LSF. Among the synQSL regressors, MLP showed superior properties including computational cost.

Conclusion: synQSL demonstrated superior parameter estimation performance compared to NL-LSF in breast IVIM–DKI analysis. In this study, MLP was considered the most suitable regressor for synQSL among those we examined, based on the balance between estimation accuracy and computational costs.

Purpose: Synthetic MRI can generate multiple contrasts from a single acquisition, yet synthetic fluid-attenuated inversion recovery (FLAIR) generally shows lower quality than conventional FLAIR. We aimed to improve 3D synthetic FLAIR image quality using deep learning, without losing the scan-time advantage of synthetic MRI.

Methods: We studied 55 adults with inflammatory demyelinating diseases who underwent 3T MRI. For each participant, five 3D quantification using an interleaved Look–Locker acquisition sequence with T2 preparation (3D-QALAS) source images and a conventional 3D-FLAIR image were acquired, and a synthetic FLAIR image was generated from the 3D-QALAS data. We trained a deep learning model in which a 3D U-shaped convolutional network (U-Net)–based attention network predicted voxel-wise weights for generating FLAIR images from the five 3D-QALAS source images, using conventional 3D-FLAIR images as the reference. The model was trained with a combined loss function of mean squared error, content loss, and style loss. Agreement with the reference was assessed using image similarity and error metrics, lesion overlap using the Dice similarity coefficient, and overall image quality and focal lesion visibility using blinded radiologist ratings. Synthetic FLAIR and the deep learning–generated FLAIR images were compared using 2-sided Wilcoxon signed-rank tests; P < 0.05 was considered statistically significant.

Results: The deep learning–generated FLAIR images showed significantly higher agreement with the reference image than synthetic FLAIR images (all P < 0.001). Lesion overlap was higher with the deep learning–generated FLAIR images (median [interquartile range]: 0.642 [0.528–0.711] versus 0.487 [0.344–0.641]). Qualitatively, the deep learning–generated FLAIR images improved overall image quality and focal lesion visibility, but reader scores remained lower than those for the reference conventional 3D-FLAIR images (all P < 0.001).

Conclusion: A deep learning–based approach applied to five 3D-QALAS source images improved the image quality of 3D synthetic FLAIR. These improvements may increase the clinical utility of 3D synthetic MRI for neuroradiologic assessment.

Purpose: To investigate relationships among relaxation, susceptibility, and diffusion parameters in healthy white matter (WM) and to characterize WM MRI feature organization using an integrative multimodal quantitative MRI framework.

Methods: Twenty-two healthy volunteers underwent 3T MRI. Quantitative parameter mapping provided R₁, R₂*, R₁·R₂*, and quantitative susceptibility mapping, while diffusion kurtosis imaging provided fractional anisotropy (FA), mean kurtosis, axial kurtosis (AK), and radial kurtosis. All maps were spatially normalized to Montreal Neurological Institute (MNI) space, and mean values were extracted from WM tracts defined by the Johns Hopkins University (JHU) White Matter Atlas. Pearson correlation analysis with false discovery rate correction, principal component analysis (PCA), hierarchical clustering, and bootstrap stability analysis were performed.

Results: The first 2 principal components explained 77.7% of the total variance. The first principal component (PC1) was mainly associated with relaxation-related parameters and diffusion kurtosis metrics, whereas the second principal component (PC2) was characterized by opposing contributions from AK and FA. Quantitative susceptibility mapping showed weak correlations with other parameters (r = −0.37 to −0.08), suggesting relatively independent susceptibility-related information. The PCA structure was preserved after excluding R₁·R₂*, and bootstrap analysis supported loading stability, with mean loading correlations of 0.94 for PC1 and 0.90 for PC2. WM tracts formed 4 major cluster-like groups and showed tract-specific multimodal fingerprints.

Conclusion: Multimodal quantitative MRI provides a concise tract-level MRI feature representation of WM by integrating relaxation, susceptibility, and diffusion information. This exploratory framework may support future studies investigating subtle WM alterations, although validation in pathological cohorts is required.

Purpose: To evaluate the clinical feasibility and quantitative performance of adaptive complex signal average (ACSA) diffusion-weighted imaging (DWI) in abdominal MRI and to compare ACSA DWI with conventional non-ACSA DWI in terms of signal intensity (SI), SNR, apparent diffusion coefficient (ADC), signal intensity difference ratio (SIDR), and lesion contrast across different numbers of excitations (NEX).

Methods: This retrospective study included 82 patients who underwent free-breathing abdominal DWI on a 3T system between January and July 2025. ACSA DWI was reconstructed from the same raw data as non-ACSA DWI using a combination of adaptive averaging and complex-domain averaging. ROIs were placed in the liver, pancreas, left adrenal gland, and spinal muscles to measure SI, SNR, and ADC. SIDR and inter-lobar right–left (R–L) ratio in SI and ADC were calculated at NEX 2, 3, and 4. For 37 patients with non-cystic hepatic nodules, lesion SI, ADC, contrast ratio, and contrast-to-noise ratio (CNR) were assessed.

Results: ACSA DWI significantly increased SI and SNR in motion-prone organs, including the lateral hepatic segment, pancreas, and left adrenal gland, at all NEX levels (P < 0.05). ADC values in abdominal parenchymal organs and hepatic nodules were lower with ACSA DWI, reflecting the noise-floor suppression and adaptive average effects. R–L SI and ADC ratios were significantly different and were closer to 1 in ACSA DWI, indicating improved hepatic signal homogeneity. Lesion SI, contrast ratio, and CNR were significantly higher with ACSA DWI, enhancing nodule conspicuity.

Conclusion: ACSA DWI improves signal uniformity, SNR, and lesion contrast while maintaining reasonable ADC quantification. Without prolonging scan time or modifying acquisition parameters, ACSA provides a robust, clinically feasible post-processing technique for enhancing both image quality and quantitative reliability in abdominal DWI.

Purpose: Most existing multi-contrast MRI super-resolution (MCMSR) methods rely on spatial-domain fusion and global attention, overlooking explicit high-frequency (HF) priors while incurring high computational costs. This work addresses these limitations through a general reference-guided MCMSR framework designed for low computational cost, applicable to any contrast pairing rather than a fixed clinical protocol.

Methods: We introduce a wavelet-guided HF prior modeling block for directional-based decomposition and bounded nonlinear enhancement, enabling precise extraction and controlled amplification of anatomical details from both reference and target contrasts. We further introduce a triple cross-contrast fusion module, based on a windowed cross-contrast attention module, to efficiently transfer high-frequency information between contrasts and reduce computational complexity compared to global attention schemes. Additionally, to reduce feature differences across contrasts, a consistent feature fusion module with selective spatial adaptive modulation is incorporated.

Results: Extensive experiments on IXI and M4Raw datasets demonstrate that our proposed windowed cross-contrast attention network (WCCAN) framework consistently outperforms state-of-the-art single and multi-contrast MRI SR methods in terms of quantitative accuracy and visual fidelity. In addition, the WCCAN model achieves lower computational complexity and faster inference time compared to the other MCMSR methods.

Conclusion: The proposed WCCAN framework provides an efficient and accurate solution for MCMSR, demonstrating superior reconstruction quality with reduced computational cost compared to existing methods.

Purpose: There is a significant association between carotid artery (CA) stenosis and atherosclerosis, with those patients often presenting with carotid tortuosity. However, the relationship between atherosclerotic risk factors and carotid tortuosity remains controversial. Previous assessments of tortuosity have predominantly utilized qualitative 2D methods. This retrospective study aimed to quantitatively evaluate the tortuosity of CAs in 3D geometry and to investigate the relationship with atherosclerotic risk factors.

Methods: A total of 168 CAs in 96 patients diagnosed with at least either-side CA stenosis on magnetic resonance angiography (MRA) between 2011 and 2019 were included in this single-center study, following the exclusion of post-treatment CAs and those with poor imaging quality. The 3D geometries of CAs were segmented using MRA and the centerlines were generated. Vascular tortuosity rate (VTR) was calculated as the curved distance divided by the linear distance between the specified 2 points. The relationship between VTRs and 8 atherosclerotic risk factors—age, body mass index, systolic blood pressure, hemoglobin A1c, high-density lipoprotein, total cholesterol (TC), non-fasting serum triglyceride (TG), and Brinkman index—was analyzed.

Results: Higher VTRs were associated with ≥65 years of age, body mass index ≥25 [kg/m²], serum TC ≥220 [mg/dL], and non-fasting serum TG ≥175 [mg/dL] in common CAs, and body mass index ≥25 [kg/m²], and systolic blood pressure ≥140 [mmHg] in internal CAs.

Conclusion: Quantitative assessment of carotid tortuosity revealed significant associations with several atherosclerotic risk factors, including obesity, hypertension, advanced age, and dyslipidemia, suggesting that VTRs may reflect the presence of these risk factors.

This study aimed to demonstrate the technical feasibility of a dedicated prostate phantom for standardized qualitative evaluation of diffusion-weighted MRI sequences, focusing on signal-to-noise characteristics and image artifacts. A custom-designed phantom simulating prostate tissue, tumor lesions, and rectal gas was used to evaluate 6 diffusion-weighted imaging techniques at 1.5 and 3.0 tesla. Quantitative assessments included signal-to-noise measurements and susceptibility artifacts, while spatial distortion was evaluated by comparison with T2-weighted imaging. Qualitative image quality was assessed using corresponding clinical images as references. Clinical images were evaluated by a radiologist and a urologist, and phantom images were evaluated by 2 radiological technologists. The phantom demonstrated high inter-reader agreement for qualitative image assessment. Clinical image evaluation was performed to confirm whether artifact trends observed in the phantom were similarly observed in clinical images and showed comparable tendencies across sequences. These findings indicate that the proposed prostate phantom provides a practical and reproducible platform for qualitative evaluation.