Comparison of accelerated MRI protocol approaches for clinical assessment of the knee.

Arthroscopy-Validated Diagnostic Performance of a Deep Learning Reconstruction Pipeline for Rapid 7-Minute Five-Sequence 3-T Knee MRI.

Background. Deep learning reconstruction (DLR) methods can enhance image quality and reduce scan time of knee MRI compared with conventional approaches but require validation against independent reference standards to ensure robustness and accuracy. Objective. The purpose of this study was to assess the diagnostic performance of a two- to threefold parallel imaging-accelerated 7-minute five-sequence 3-T knee MRI protocol using a deep learning-based image reconstruction pipeline (AIR Recon DL, GE HealthCare), with arthroscopic surgery as the reference standard. Methods. A total of 117 consecutive adult patients (mean age: 44 ± 16 [SD] years; 65 men, 52 women) with painful knee conditions who underwent DLR 3-T knee MRI and arthroscopic knee surgery with a median MRI-tosurgery interval of 35 days (range: 5-88 days) between June 2021 and May 2024 were retrospectively identified and included. MRI studies were independently reviewed by seven musculoskeletal radiologists for image quality parameters using Likert scales (range: 1 = very bad to 5 = very good) and the presence of meniscus tears, cruciate and collateral ligament tears, and articular cartilage defects. Statistical analyses included interreader agreements and diagnostic performance testing. Results. Overall image quality of DLR knee MRI scans was good (median: 4 [IQR, 4-5]), with minimal image noise (4 [4-4]), good edge sharpness (4 [4-5]), absence of reconstruction artifacts (5 [4-5]), and high interreader agreement for all quality metrics (κ ≥ 0.83). Diagnostic performance for detecting arthroscopy-validated structural abnormalities was very good (AUC ≥ 0.81) with good to very good interreader agreement (κ ≥ 0.62). The sensitivity, specificity, accuracy, and AUC values were 89%, 87%, 88%, and 0.87 for medial meniscus tears (prevalence at arthroscopy: 78/117; 67%), 72%, 91%, 83%, and 0.82 for lateral meniscus tears (50/117; 43%), 97%, 98%, 97%, and 0.97 for anterior cruciate ligament tears (30/117; 26%), and 73%, 89%, 86%, and 0.81 for articular cartilage defects (182/702; 26%). Conclusion. Clinical 7-minute five-sequence 3-T knee MRI with deep learning reconstruction provides good to excellent diagnostic performance for detecting arthroscopy-validated internal derangement of the knee. Clinical Impact. Deep learning reconstruction enables rapid high-quality clinical 3-T knee MRI with high diagnostic performance for arthroscopy-validated abnormalities.

Current Medical Imaging

Deep Learning-Augmented Zero Echo Time MRI Increases Diagnostic Confidence for Osseous Assessment in Hand and Foot MRI Protocols

Background/Objectives: This study assesses the outcomes of integrating deep learning-augmented zero echo time (ZTE DL) MRI sequences into standard MRI protocols for assessment of the hands and feet. Methods: In this single-center, retrospective study, a standard MRI protocol of hands and feet at 1.5 T was compared with the same protocol with an added ZTE DL sequence. Pathological changes in the bone, including subcortical sclerosis, osteophytes, joint space narrowing and fractures, were rated as present or absent. Indeterminate intraosseous lesions (erosions/ganglia) and indeterminate extraosseous lesions (ossicles/soft tissue calcifications) were additionally assessed on a semi-quantitative 4-point Likert scale. Diagnostic confidence was rated as low, moderate, or high. Standard MRI vs. ZTE-DL-augmented MRI comparisons were evaluated using a mixed-effects model with reader as a random effect, and X-ray vs. ZTE-DL-augmented MRI comparisons using a paired Wilcoxon signed-rank test. Interreader agreement (two readers) was assessed using Kappa statistics. Results: The cohort encompassed 59 datasets (feet = 22, hands = 37) of 40 patients with an average age of 51.79 (SD ± 11.84) years (58% women (n = 34), 42% men (n = 25)). Additional ZTE DL sequences resulted in similar findings, but with higher diagnostic confidence for assessment of bone changes when compared to conventional MR sequences (p-values < 0.05) and, overall, when compared to radiographs in a subgroup (p < 0.05 for five of six pathologic bone changes). Interreader agreement of diagnostic confidence was moderate to substantial (kappa 0.59–0.71). Conclusions: Addition of ZTE DL sequences to standard MRI protocols of hands and feet at 1.5 T demonstrated similar findings in the assessment of pathological bone changes, but with higher diagnostic confidence compared to conventional MR sequences and radiographs. However, further validation against a reference standard is required to determine diagnostic accuracy.

Comparative Evaluation of Magnetic Field Strength and Fat Suppression for Magnetic Resonance Sialography in Healthy Dogs.

This prospective study aimed to evaluate the feasibility of magnetic resonance (MR) sialography in dogs, compare image quality between 1.5 and 3 T scanners, and assess the effect of fat suppression (FS). Five healthy beagle dogs underwent MR imaging (MRI) at both field strengths using three-dimensional (3D) fast spin-echo (FSE) and 3D FSE-FS sequences. Qualitative (four-point scoring of visibility and delineation) and quantitative (signal-to-noise ratio [SNR] and contrast-to-noise ratio [CNR]) assessments were performed for each salivary duct. Interobserver agreement was near-perfect (κ = 0.861; 95% confidence interval, 0.833-0.890). Qualitatively, the mandibular salivary duct was well visualized at both field strengths, whereas the sublingual salivary duct was reliably visualized only at 3 T (p = 0.031). The parotid salivary duct showed improved delineation at 3 T (p = 0.031), and the zygomatic salivary duct tended to improve visibility and delineation at 3 T. Quantitatively, the mandibular salivary duct showed increased SNR and significantly higher CNR at 3 T (p = 0.031). FS had no significant qualitative effect but increased CNR at 3 T for the mandibular, sublingual, and parotid ducts (p = 0.031). MR sialography is a feasible method for evaluating salivary ducts in dogs, with 3 T MRI providing superior visualization of small ducts such as the sublingual salivary duct. These findings indicate that MR sialography may be valuable for assessing the mandibular-sublingual salivary duct complex, a common site of salivary diseases in dogs, and may provide a normal reference for future studies on salivary gland diseases in dogs.

Accelerated 7T 3D DESS MR Neurography of the Cervical Spine Using CS-CAIPIRINHA and Deep Learning Reconstruction.

OBJECTIVE To evaluate the potential of combined coherent and incoherent undersampling with deep learning (DL) reconstruction for highly accelerated high-resolution three-dimensional (3D) double-echo steady-state (DESS) MRI of the cervical nerves and rootlets at 7T. METHODS In this prospective study, asymptomatic volunteers underwent 7T MRI of the cervical spine. A high-resolution, 3D DESS sequence (reconstructed voxel size 0.35 mm³ isotropic) with regular 2-fold GRAPPA undersampling served as reference. The sequence was repeated with coherent CAIPIRINHA undersampling alone and combined CAIPIRINHA + incoherent compressed sensing (CS)-based undersampling with DL reconstruction and varying acceleration factors (DL-CAIPIRINHA: R=2 to 12 and DL-CS-CAIPI: R=4 to 16), while all other imaging parameters remained constant. Two fellowship-trained musculoskeletal radiologists assessed image quality, noise, reconstruction and motion artifacts, and edge sharpness of the spinal cord, intradural rootlets/roots, and exiting spinal nerves, including dorsal root ganglia (DRGs), using 5-point Likert scales, with higher scores indicating better image quality or fewer artifacts. Statistics included Friedman and post hoc Wilcoxon signed-rank tests (Benjamini-Hochberg corrected) and κ statistics. RESULTS Thirty-two volunteers (mean age 30.9±7.2 y; mean BMI 23.4±2.8 kg/m²; 16 females) were included. Image quality significantly improved with 2- and 4-fold DL-CAIPIRINHA (means 4.4 to 4.6) and 4- and 8-fold DL-CS-CAIPI (3.8 to 3.9) compared with GRAPPA (3.2; all P<0.001), and remained comparable with 12-fold DL-CS-CAIPI (P=0.42) despite more than 6-fold reduction in scan time. All DL-based reconstructions reduced noise (P<0.001). Reconstruction artifacts were more pronounced for 12-fold DL-CAIPIRINHA (1.6) than for 12- and 16-fold DL-CS-CAIPI (3.1 to 3.3; P<0.001). Motion artifacts decreased with acceleration factors ≥4 (P<0.001). Sharpness of the spinal cord, intraspinal rootlets/roots, and DRGs was highest with 4-fold DL-CAIPIRINHA (4.2 to 4.7) vs. GRAPPA (2.9 to 3.2; P<0.001). Inter-reader agreement was almost perfect across all imaging features (κ=0.85 to 0.95). CONCLUSIONS Coherent CAIPIRINHA undersampling with DL reconstruction improves image quality over conventional GRAPPA while enabling 50% faster imaging in 7T cervical spine MRI. Adding incoherent CS undersampling permits substantially higher acceleration with preserved image quality and fewer reconstruction artifacts. The novel integration of coherent and incoherent undersampling via combined CAIPIRINHA and CS with DL reconstruction provides a robust framework for preserving image quality at acceleration factors beyond 8, enabling efficient high-resolution cervical spine MRI with scan time reductions exceeding 6-fold.

Deep learning-based reconstruction for multi-shot breast DWI: Quantitative and qualitative improvements.

PURPOSE To evaluate the effectiveness of deep learning (DL)-based reconstruction for improving image quality and reducing scan time in multi-shot diffusion-weighted imaging (DWI) of the breast, compared with standard readout-segmented echo-planar imaging (rs-EPI). MATERIALS AND METHODS This retrospective study included 146 women (mean age: 55.8 years) with newly diagnosed breast cancer who underwent preoperative breast MRI with both rs-EPI and prototype DL-reconstructed (DLR) multi-shot DWI. Quantitative metrics-signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), lesion contrast, and apparent diffusion coefficient (ADC)-were compared between sequences. Two radiologists independently assessed overall image quality, artifact suppression, lesion conspicuity, and fibroglandular tissue suppression using a 5-point Likert scale. An independent validation cohort of 39 consecutive patients was analyzed using the same methodology. RESULTS Compared with rs-EPI, DLR multi-shot DWI demonstrated significantly higher SNR, CNR, and lesion contrast at both b = 800 and synthetic b = 1500 s/mm2 (all p < 0.05). Tumor ADC values were slightly lower and fibroglandular tissue ADC values slightly higher with DLR (both p < 0.05). Qualitative assessments demonstrated higher scores for overall image quality and lesion conspicuity with DLR, with improved artifact suppression at b = 800 s/mm2. Interobserver agreement was substantial to almost perfect (weighted κ = 0.753-0.981). DL reconstruction reduced acquisition time by 29% compared with rs-EPI. In the validation cohort, improvements in SNR, CNR, and overall image quality were also observed, whereas ADC values did not differ significantly between techniques. CONCLUSION DL-based reconstruction improves quantitative and qualitative image quality in multi-shot breast DWI while reducing scan time. These findings support its potential integration into clinical breast MRI.