Journal of Magnetic Resonance Imaging最新文献

Background: Pathophysiological changes affect tissue cell composition and density. For example, neurodegenerative disorders and brain tumors are associated with cell loss and abnormal accumulation, respectively. In these scenarios, if monitored and tracked, tissue cellularity might be used to inform clinical diagnosis and management.

Purpose: To propose and evaluate a new marker of tissue cellularity, called susceptibility-Derived Cellularity Index (χDCI), that would be readily available for clinical applications with fast acquisition and at high resolution.

Study type: Retrospective study.

Population/subjects: 24 healthy subjects (7/17 M/F, 70 ± 11 years) and 21 patients with IDH-wild type glioblastoma (16/5 M/F, 65 ± 8 years).

Field strength/sequence: 3 T MRI sequences including 3D T1w pre- and post-contrast agent injection, 3D T2w, 3D FLAIR, 3D multi-echo gradient recalled echo, 2D diffusion weighted imaging.

Assessment: χDCI was computed based on parameters estimated with DECOMPOSE-QSM. The Neurite Density Index (NDI) was estimated with the NODDI model. T1w images were used for region of interest (ROIs) segmentations with FreeSurfer (i.e., cortical gray matter, white matter, thalamus, caudate, putamen, pallidum, hippocampus and amygdala). For the patients with glioblastoma, regions of contrast enhancement, necrosis, and edema were also included in the analysis.

Statistical tests: Pearson's correlation analysis between mean χDCI and NDI values in the ROIs was carried out separately for the two cohorts of participants (significance level = 0.05, after correction for multiple comparisons).

Results: Significant correlations were observed between χDCI and NDI in white matter (r = 0.56) and putamen (r = 0.69) for the healthy participants. Significant positive correlations were also found in white matter (r = 0.6), pallidum (r = 0.48), putamen (r = 0.79), thalamus (r = 0.64) and edema (r = 0.69) for the patient cohort.

Data conclusion: χDCI is proposed as a marker of tissue cellularity. The significant associations between χDCI and NDI in several regions investigated in the present study support the potential of χDCI as a proxy of intracellular volume fraction.

Level of evidence: 3:

Technical efficacy: Stage 1.

Background: Accurate identification of benign and malignant bile duct dilatation (BDD) is needed to determine its management plan. Conventional imaging evaluation is subjective, whereas deep learning (DL) offers potential for automated objective assessment.

Purpose: To construct and evaluate DL models and ensemble strategies based on magnetic resonance cholangiopancreatography (MRCP) images for identifying benign and malignant BDD.

Study type: Retrospective and prospective.

Population: A retrospective cohort (n = 378; median age, 60 years [range: 14, 90]; 194 male) from two institutions and a prospective cohort (n = 60; median age, 62.5 years [range: 15, 86]; 30 male) were included. Retrospective data were randomly stratified split into training, validation, and internal test sets (2:1:1) and an independent external test set. Benign cases were downsampled to balance class distribution.

Field strength/sequence: 3 T MRCP (3D turbo spin echo: VISTA and SPACE).

Assessment: The primary retrospective endpoint was area under the curve (AUC) across DL algorithms and ensembles. Prospectively, the accuracy, sensitivity, and specificity of the model was compared with those of three radiologists.

Statistical tests: Group comparisons used Mann-Whitney U and Chi-square tests (p < 0.05). Model performance was evaluated using the Hosmer-Lemeshow test, DeLong's test with Bonferroni correction (α = 0.005), and McNemar's test.

Results: The Xception model achieved AUCs of 0.816 (95% CI, 0.788-0.844) on the internal test set and 0.807 (95% CI, 0.779-0.835) on the external test set. The ensemble model incorporating logistic regression yielded higher patient-level AUCs of 0.890 and 0.885, with good calibration (p = 0.109). No significant differences were observed among the five ensemble strategies (minimum adjusted p = 0.62). In the prospective cohort, the model showed 90.0% accuracy, sensitivity, and specificity, comparable to radiologists (76.7%-86.7%) without a significant difference (p = 0.143, 0.302, and 0.774, respectively).

Data conclusion: The Xce-LR model shows potential for automating BDD differentiation using MRCP.

Level of evidence: 2:

Technical efficacy: Stage 2.

Prostate MRI has transformed lesion detection and risk stratification in prostate cancer, but its impact is constrained by the high cost of the exam, variability in interpretation, and limited scalability. False negatives, false positives, and moderate inter-reader agreement undermine reliability, while long acquisition times restrict throughput. Artificial intelligence (AI) offers potential solutions to address many of the limitations of prostate MRI in the clinical management pathway. Machine learning-based triage can refine patient selection to optimize resources. Deep learning reconstruction enables accelerated acquisition while preserving diagnostic quality, with multiple FDA-cleared products now in clinical use. Ongoing development of automated quality assessment and artifact correction aims to improve reliability by reducing nondiagnostic exams. In image interpretation, AI models for lesion detection and clinically significant prostate cancer prediction achieve performance comparable to radiologists, and the PI-CAI international reader study has provided the strongest evidence to date of non-inferiority at scale. More recent work extends MRI-derived features into prognostic modeling of recurrence, metastasis, and functional outcomes. This review synthesizes progress across five domains-triage, accelerated acquisition and reconstruction, image quality assurance, diagnosis, and prognosis-highlighting the level of evidence, validation status, and barriers to adoption. While acquisition and reconstruction are furthest along, with FDA-cleared tools and prospective evaluations, triage, quality control, and prognosis remain earlier in development. Ensuring equitable performance across populations, incorporating uncertainty estimation, and conducting prospective workflow trials will be essential to move from promising prototypes to routine practice. Ultimately, AI could accelerate the adoption of prostate MRI toward a scalable platform for earlier detection and population-level prostate cancer management. EVIDENCE LEVEL: N/A TECHNICAL EFFICACY: 3.

Background: Cystic fibrosis (CF) monitoring relies on computed tomography (CT), but ultra-short echo time MRI (UTE-MRI) offers a radiation-free alternative. However, its clinical adoption is hindered by the laborious and subjective manual analysis, which prevents standardized quantification of bronchial abnormalities.

Purpose: To develop a deep learning (DL) system for the segmentation of CF bronchial abnormalities on UTE-MRI and assess clinical relevance in patients undergoing cystic fibrosis transmembrane conductance regulator (CFTR) modulator treatment.

Study type: Retrospective.

Population: One-hundred and sixty-six CF patients were included (age = 23 ± 11, 48% male), comprising a training set (n = 97), a test set (n = 25), and an independent clinical validation cohort (n = 44).

Field strength/sequence: 1.5T/UTE-MRI 3D gradient-echo Spiral Volume Interpolated Breath-hold Examination (VIBE) sequence.

Assessment: The RiSeNet architecture was trained using paired UTE-MRI and CT scans. Its technical performance was evaluated against expert-refined segmentations and compared to state-of-the-art segmentation models using topology-aware metrics: Normalized Surface Dice (NSD) and CenterLine Dice (clDice). Clinical validation was performed by correlating automated measurements at baseline (M0) and 1-year post-CFTR modulator treatment (M12) with Bhalla scores and pulmonary function tests (FEV1%p).

Statistical tests: Student's t-test, Mann-Whitney, Wilcoxon, and Chi-square tests were used for group comparisons. The Spearman test was used to assess correlations. A p value < 0.05 was considered statistically significant.

Results: In the test group, RiSeNet achieved significantly superior performance versus state-of-the-art with NSD scores of 0.84 for bronchiectasis, 0.90 for wall thickening, and 0.75 for mucus; and clDice scores of 0.69, 0.61, and 0.64, respectively. In the clinical validation group, significant correlations with Bhalla (ρ = -0.92/-0.85) and FEV1%p (ρ = -0.68/-0.67) were observed pre/post-CFTR modulator. Post-CFTR modulator, FEV1%p improved (69%-92%) with significant reductions in bronchiectasis (3.88-1.25), wall thickening (30.43-3.05), and mucus (53.30-11.80).

Data conclusion: RiSeNet may enable semantic segmentation of CF abnormalities on radiation-free UTE-MRI.

Evidence level: 3 TECHNICAL EFFICACY: 4.

Background: Accurate plane positioning is important for high-quality cardiac MRI images but requires specialized training, limiting accessibility.

Purpose: To evaluate an automated plane positioning tool and compare it with manual planning.

Study type: Prospective.

Population: Fifty-seven healthy volunteers (28 males; median age 42 years) and 20 consecutive patients (15 males; median age 61 years) scheduled for clinical cardiac MRI.

Field strength/sequence: Steady state free precession cine sequence at 1.5 T.

Assessment: In volunteers, short-axis (SAX), 2-chamber (2CH), 3-chamber (3CH), and 4-chamber (4CH) cine images were acquired using both automated and manual prescription. Two blinded radiologists (5 and 6 years of clinical cardiac MRI experience) rated plane quality on a Likert scale (1 = nondiagnostic to 5 = excellent). Mean plane angle differences between manual and automated prescriptions were calculated. Left and right ventricular end-systolic volume (ESV), end-diastolic volume (EDV), stroke volume (SV), and ejection fraction (EF) were compared. In patients, the number of required manual corrections to automated prescriptions was recorded.

Statistical analysis: Wilcoxon matched-pairs signed rank tests and Bland-Altman analyses, significance level at p ≤ 0.05.

Results: Automated plane positioning was successful in all volunteers. Image plane quality did not differ significantly between automated (mean score 4.64) and manual prescription (4.62, p = 0.812). Mean angle differences were 6.7° ± 4.3° (SAX), 10.3° ± 5.8° (2CH), 8.9° ± 5.1° (3CH), and 8.0° ± 4.8° (4CH). Volumetric parameters showed no significant differences between both planning methods with mean biases being -0.5 mL, p = 0.305 (LVEDV), 0.5 mL, p = 0.683 (LVESV), -1.0 mL, p = 0.168 (LVSV) and 0.4%, and p = 0.215 (LVEF). In patients, 8.8% (7/80) of automatically prescribed planes required minor corrections; 91.2% (73/80) were accepted without adjustments.

Data conclusion: Automated plane positioning for cardiac MRI may provide high-quality images and accurate volumetric assessment comparable to manual planning.

Evidence level: 2.

Technical efficacy: Stage 2.