Rapid 3D breath-hold MR cholangiopancreatography using deep learning-constrained compressed sensing reconstruction

Deep Learning-accelerated MRI in Body and Chest

Deep learning reconstruction (DLR) provides an elegant solution for MR acceleration while preserving image quality. This advancement is crucial for body imaging, which is frequently marred by the increased likelihood of motion-related artifacts. Multiple vendor-specific models focusing on T2, T1, and diffusion-weighted imaging have been developed for the abdomen, pelvis, and chest, with the liver and prostate being the most well-studied organ systems. Variational networks with supervised DL models, including data consistency layers and regularizers, are the most common DLR methods. The common theme for all single-center studies on this subject has been noninferior or superior image quality metrics and lesion conspicuity to conventional sequences despite significant acquisition time reduction. DLR also provides a potential for denoising, artifact reduction, increased resolution, and increased signal-noise ratio (SNR) and contrast-to-noise ratio (CNR) that can be balanced with acceleration benefits depending on the imaged organ system. Some specific challenges faced by DLR include slightly reduced lesion detection, cardiac motion-related signal loss, regional SNR variations, and variabilities in ADC measurements as reported in different organ systems. Continued investigations with large-scale multicenter prospective clinical validation of DLR to document generalizability and demonstrate noninferior diagnostic accuracy with histopathologic correlation are the need of the hour. The creation of vendor-neutral solutions, open data sharing, and diversifying training data sets are also critical to strengthening model robustness.

Deep learning reconstruction for accelerated 3-D magnetic resonance cholangiopancreatography

PURPOSE

This study aimed to compare a conventional three-dimensional (3-D) magnetic resonance cholangiopancreatography (MRCP) sequence with a deep learning (DL)-accelerated MRCP sequence (hereafter, MRCPDL) regarding acquisition time and image quality.

MATERIALS AND METHODS

We conducted a prospective study of consecutive patients referred for MRCP between November 2023 and April 2024 at a single tertiary center. Each participant underwent 1.5T 3-D T2-weighted turbo spin echo MRCP using both a conventional sequence (threefold acceleration) and MRCPDL (eightfold acceleration). Three blinded readers independently evaluated image quality, including background signal suppression, bile and pancreatic duct visibility, artifact level, and diagnostic confidence on an ordinal four-point scale. Acquisition times were compared using a paired t-test. Image quality parameters were assessed with repeated measures ANOVA. Interreader agreement was analyzed using Fleiss' κ.

RESULTS

Out of 419 consecutive patients, 30 participants were evaluated (mean age, 63 ± 15 years; 16 men, 14 women). The mean acquisition time was 10:30 ± 03:04 min for conventional MRCP and 3:57 ± 01:13 min for MRCPDL, P < 0.001. MRCPDL reduced acquisition time by 62.4%. Artifact levels were rated at 3.17 ± 0.77 for conventional MRCP and 3.56 ± 0.66 for MRCPDL (P = 0.041). Background signal suppression, bile duct visibility, pancreatic duct visibility, and diagnostic confidence did not differ significantly (P > 0.05). Interreader agreement was substantial to almost perfect (κ: 0.64-87).

CONCLUSIONS

Deep learning-accelerated 3-D MRCP reduced acquisition time by 62%, minimized artifacts, and preserved bile and pancreatic duct visibility, supporting its adoption in routine clinical practice.

Evaluation of high-resolution pituitary dynamic contrast-enhanced MRI using deep learning-based compressed sensing and super-resolution reconstruction

OBJECTIVE

This study aims to assess diagnostic performance of high-resolution dynamic contrast-enhanced (DCE) MRI with deep learning-based compressed sensing and super-resolution (DLCS-SR) reconstruction for identifying microadenomas.

MATERIALS AND METHODS

This prospective study included 126 participants with suspected pituitary microadenomas who underwent DCE MRI between June 2023 and January 2024. Four image groups were derived from single-scan DCE MRI, which included 1.5-mm slice thickness images using DLCS-SR (1.5-mm DLCS-SR images), 1.5-mm slice thickness images with deep learning-based compressed sensing reconstruction (1.5-mm DLCS images), 1.5-mm routine images, and 3-mm slice thickness images using DLCS-SR (3-mm DLCS-SR images). Diagnostic criteria were established by incorporating laboratory findings, clinical symptoms, medical histories, previous imaging, and certain pathologic reports. Two readers assessed the diagnostic performance in identifying pituitary abnormalities and microadenomas. Diagnostic agreements were assessed using κ statistics, and intergroup comparisons for microadenoma detection were performed using the DeLong and McNemar tests.

RESULTS

The 1.5-mm DLCS-SR images (κ = 0.746-0.848) exhibited superior diagnostic agreement, outperforming 1.5-mm DLCS (κ = 0.585-0.687), 1.5-mm routine (κ = 0.449-0.487), and 3-mm DLCS-SR images (κ = 0.347-0.369) (p < 0.001 for all). Additionally, the performance of 1.5-mm DLCS-SR images in identifying microadenomas [area under the receiver operating characteristic curve (AUC), 0.89-0.94] surpassed that of 1.5-mm DLCS (AUC, 0.83-0.87; p = 0.042 and 0.011, respectively), 1.5-mm routine (AUC, 0.76-0.78; p < 0.001), and 3-mm DLCS-SR images (AUC, 0.72-0.74; p < 0.001).

CONCLUSION

The findings revealed superior diagnostic performance of 1.5-mm DLCS-SR images in identifying pituitary abnormalities and microadenomas, indicating the clinical-potential of high-resolution DCE MRI.

KEY POINTS

Question What strategies can overcome the resolution limitations of conventional dynamic contrast-enhanced (DCE) MRI, and which contribute to a high false-negative rate in diagnosing pituitary microadenomas? Findings Deep learning-based compressed sensing and super-resolution reconstruction applied to DCE MRI achieved high resolution while improving image quality and diagnostic efficacy. Clinical relevance DCE MRI with a 1.5-mm slice thickness and high in-plane resolution, utilizing deep learning-based compressed sensing and super-resolution reconstruction, significantly enhances diagnostic accuracy for pituitary abnormalities and microadenomas, enabling timely and effective patient management.

Deep learning-based reconstruction for three-dimensional volumetric brain MRI: a qualitative and quantitative assessment

BACKGROUND

To evaluate the performance of a deep learning reconstruction (DLR) based on Adaptive-Compressed sensing (CS)-Network for brain MRI and validate it in a clinical setting.

METHODS

Ten healthy volunteers and 22 consecutive patients were prospectively enrolled. Volunteers underwent 3D brain MRI including T1 without CS factor (9:16 min, reference standard); with CS factor of 2 without DLR (CS2, 4:6 min); with CS factor of 2 with DLR (DLR-CS2); with CS factor of 4 without DLR (CS4, 2:6 min); and with CS factor of 4 with DLR (DLR-CS4). The patients' MRI included the CS2 and DLR-CS4. The volumes of lateral ventricles, hippocampus, choroid plexus, and white matter hypointensity were calculated and compared among the sequences. Three radiologists independently assessed anatomical conspicuity, overall image quality, artifacts, signal-to-noise ratio (SNR), and sharpness using a 5-point scale for each sequence.

RESULTS

Applying acceleration factors of 2 and 4 reduced the scan time to 65.4% and 33.5%, respectively, of that of the reference standard. Volumes of all the measured subregions showed no significant differences among different sequences in all participants. In qualitative analysis, the interrater agreement was excellent (κ = 0.844-0.926). In volunteers, quality of DLR-CS4 were comparable to those of CS2 for all metrics except for the overall image quality and SNR despite a 51.2% scan time reduction. In patients, DLR-CS4 showed quality comparable to that of CS2 for all metrics.

CONCLUSIONS

DLR allowed the scan time reduction by at least half without sacrificing image quality and volumetric quantification accuracy, supporting its reliability and efficiency.

Utility of under-sampled scans with iterative reconstruction and high-frequency preserving transform for high spatial resolution magnetic resonance cholangiopancreatography

PURPOSE

Under-sampled scans with iterative reconstruction and high-frequency preserving transform (Us-IRHF) can increase the acquisition speed without degrading the image quality by recovering image information from under-sampled data. We investigate the clinical applicability of high spatial resolution magnetic resonance cholangiopancreatography (MRCP) images without extending the scanning time using Us-IRHF.

METHODS

A slit phantom was scanned with conventional- (without Us-IRHF), Us-IR- (without HF), and Us-IRHF scanning. The matrix size was 320 × 320 for Us-IR- and Us-IRHF- and 288 × 208 for conventional scanning. Modulation transfer function (MTF) focused on the 1.0 lp/cm gauge for each scanning was calculated. For clinical study we acquired respiratory-triggered 3D MRCP scans with and without Us-IRHF (U+-, U-MRCP) in 41 patients. The matrix size was 320 × 320 for U+- and 288 × 208 for U-MRCP. The acquisition time and the relative duct-to-periductal contrast ratios (RCs) for the right- and left intrahepatic bile-, the common bile-, and the main pancreatic duct were recorded. Visualization of each duct and overall image quality was scored on 5-point confidence scales. For visualization of each duct the score ranged from 1 (not visible) to 5 (visible with excellent details), for the image quality, it ranged from 1 (undiagnostic) to 5 (excellent). Superiority for the qualitative visualization score and non-inferiority for the RC values with prespecified margins were assessed.

RESULTS

Phantom study showed that compared to the conventional- and Us-IR (without HF) images, the MTF for the Us-IRHF image revealed the highest response. For clinical study, the mean acquisition time was 161 s for U+- and 165 s for U-MRCP. For all ducts, the RC value of U+MRCP was non-inferior to U-MRCP and the qualitative visualization score assigned to U+MRCP was superior to U-MRCP.

CONCLUSION

Us-IRHF improved the image quality of high spatial resolution MRCP without extending the scanning time.

Deep learning constrained compressed sensing reconstruction improves high-resolution three-dimensional (3D) T2-weighted turbo spin echo magnetic resonance imaging (MRI) of the lumbar spine

AIM

We sought to assess the image quality of three-dimensional (3D) T2-weighted (T2w) turbo spin echo (TSE) sequences with deep learning (DL)-constrained compressed sensing (CS) reconstruction relative to a reference two-dimensional (2D) T2w TSE sequence for routine clinical lumbar spine MRI.

MATERIALS AND METHODS

Fifty-three patients underwent imaging of the lumbar spine with a sagittal 2D T2w TSE sequence and with two CS-accelerated 3D T2w TSE sequences (voxel size of 0.4 × 0.4 × 0.5 mm) with CS factors of 7 and 11. The CS-accelerated sequences were reconstructed with iterative reconstruction with wavelet transformation (conventional CS) and secondly with a DL-constrained CS reconstruction (named CS-AI). Two readers graded image quality, based on 8 metrics (overall image quality, presence of image noise, presence of motion artifacts, delineation/conspicuity and clarity of anatomical structures such as the spinal cord, cauda equine nerve roots, cerebrospinal fluid (CSF), intervertebral disc, and bone marrow and intervertebral foramen) using Likert scales.

RESULTS

Overall inter-readout agreement was substantial (Krippendorff's α = 0.724, 95% confidence interval [CI]: 0.692-0.755). The CS7-AI and CS11-AI sequences were comparable or better than the 2D sequence in all 8 metrics (p < 0.001-p > 0.99). The CS7 and CS11 sequences were comparable or better than the 2D sequence in only 5 and 3 of the 8 metrics, respectively (p < 0.001-p > 0.99).

CONCLUSION

A DL-constrained CS reconstruction significantly improves the quality of accelerated high-resolution 3D T2w TSE imaging of the lumbar spine. Thus, high-quality imaging in a submillimeter resolution in all three imaging planes can be achieved without compromising the image quality as compared with standard 2D T2w TSE imaging.

Advanced MRI techniques in abdominal imaging

Magnetic resonance imaging (MRI) is a crucial modality for abdominal imaging evaluation of focal lesions and tissue properties. However, several obstacles, such as prolonged scan times, limitations in patients' breath-hold capacity, and contrast agent-associated artifacts, remain in abdominal MR images. Recent techniques, including parallel imaging, three-dimensional acquisition, compressed sensing, and deep learning, have been developed to reduce the scan time while ensuring acceptable image quality or to achieve higher resolution without extending the scan duration. Quantitative measurements using MRI techniques enable the noninvasive evaluation of specific materials. A comprehensive understanding of these advanced techniques is essential for accurate interpretation of MRI sequences. Herein, we therefore review advanced abdominal MRI techniques.

Deep learning-based compressed SENSE improved diffusion-weighted image quality and liver cancer detection: A prospective study

PURPOSE

To assess whether diffusion-weighted imaging (DWI) with Compressed SENSE (CS) and deep learning (DL-CS-DWI) can improve image quality and lesion detection in patients at risk for hepatocellular carcinoma (HCC).

METHODS

This single-center prospective study enrolled consecutive at-risk participants who underwent 3.0 T gadoxetate disodium-enhanced MRI. Conventional DWI was acquired using parallel imaging (PI) with SENSE (PI-DWI). In CS-DWI and DL-CS-DWI, CS but not PI with SENSE was used to accelerate the scan with 2.5 as the acceleration factor. Qualitative and quantitative image quality were independently assessed by two masked reviewers, and were compared using the Wilcoxon signed-rank test. The detection rates of clinically-relevant (LR-4/5/M based on the Liver Imaging Reporting and Data System v2018) liver lesions for each DWI sequence were independently evaluated by another two masked reviewers against their consensus assessments based on all available non-DWI sequences, and were compared by the McNemar test.

RESULTS

67 participants (median age, 58.0 years; 56 males) with 197 clinically-relevant liver lesions were enrolled. Among the three DWI sequences, DL-CS-DWI showed the best qualitative and quantitative image qualities (p range, <0.001-0.039). For clinically-relevant liver lesions, the detection rates (91.4%-93.4%) of DL-CS-DWI showed no difference with CS-DWI (87.3%-89.8%, p = 0.230-0.231) but were superior to PI-DWI (82.7%-85.8%, p = 0.015-0.025). For lesions located in the hepatic dome, DL-CS-DWI demonstrated the highest detection rates (94.8%-97.4% vs 76.9%-79.5% vs 64.1%-69.2%, p = 0.002-0.045) among the three DWI sequences.

CONCLUSION

In patients at high-risk for HCC, DL-CS-DWI improved image quality and detection for clinically-relevant liver lesions, especially for the hepatic dome.

High Resolution TOF-MRA Using Compressed Sensing-based Deep Learning Image Reconstruction for the Visualization of Lenticulostriate Arteries: A Preliminary Study

PURPOSE

To investigate the visibility of the lenticulostriate arteries (LSAs) in time-of-flight (TOF)-MR angiography (MRA) using compressed sensing (CS)-based deep learning (DL) image reconstruction by comparing its image quality with that obtained by the conventional CS algorithm.

METHODS

Five healthy volunteers were included. High-resolution TOF-MRA images with the reduction (R)-factor of 1 were acquired as full-sampling data. Images with R-factors of 2, 4, and 6 were then reconstructed using CS-DL and conventional CS (the combination of CS and sensitivity conceding; CS-SENSE) reconstruction, respectively. In the quantitative assessment, the number of visible LSAs (identified by two radiologists), length of each depicted LSA (evaluated by one radiological technologist), and normalized mean squared error (NMSE) value were assessed. In the qualitative assessment, the overall image quality and the visibility of the peripheral LSA were visually evaluated by two radiologists.

RESULTS

In the quantitative assessment of the DL-CS images, the number of visible LSAs was significantly higher than those obtained with CS-SENSE in the R-factors of 4 and 6 (Reader 1) and in the R-factor of 6 (Reader 2). The length of the depicted LSAs in the DL-CS images was significantly longer in the R-factor 6 compared to the CS-SENSE result. The NMSE value in CS-DL was significantly lower than in CS-SENSE for R-factors of 4 and 6. In the qualitative assessment of DL-CS images, the overall image quality was significantly higher than that obtained with CS-SENSE in the R-factors 4 and 6 (Reader 1) and in the R-factor 4 (Reader 2). The visibility of the peripheral LSA was significantly higher than that shown by CS-SENSE in all R-factors (Reader 1) and in the R-factors 2 and 4 (Reader 2).

CONCLUSION

CS-DL reconstruction demonstrated preserved image quality for the depiction of LSAs compared to the conventional CS-SENSE when the R-factor is elevated.

Utility of Thin-slice Fat-suppressed Single-shot T2-weighted MR Imaging with Deep Learning Image Reconstruction as a Protocol for Evaluating the Pancreas

PURPOSE

To compare the utility of thin-slice fat-suppressed single-shot T2-weighted imaging (T2WI) with deep learning image reconstruction (DLIR) and conventional fast spin-echo T2WI with DLIR for evaluating pancreatic protocol.

METHODS

This retrospective study included 42 patients (mean age, 70.2 years) with pancreatic cancer who underwent gadoxetic acid-enhanced MRI. Three fat-suppressed T2WI, including conventional fast-spin echo with 6 mm thickness (FSE 6 mm), single-shot fast-spin echo with 6 mm and 3 mm thickness (SSFSE 6 mm and SSFSE 3 mm), were acquired for each patient. For quantitative analysis, the SNRs of the upper abdominal organs were calculated between images with and without DLIR. The pancreas-to-lesion contrast on DLIR images was also calculated. For qualitative analysis, two abdominal radiologists independently scored the image quality on a 5-point scale in the FSE 6 mm, SSFSE 6 mm, and SSFSE 3 mm with DLIR.

RESULTS

The SNRs significantly improved among the three T2-weighted images with DLIR compared to those without DLIR in all patients (P < 0.001). The pancreas-to-lesion contrast of SSFSE 3 mm was higher than those of the FSE 6 mm (P < 0.001) and tended to be higher than SSFSE 6 mm (P = 0.07). SSFSE 3 mm had the highest image qualities regarding pancreas edge sharpness, pancreatic duct clarity, and overall image quality, followed by SSFSE 6 mm and FSE 6 mm (P < 0.0001).

CONCLUSION

SSFSE 3 mm with DLIR demonstrated significant improvements in SNRs of the pancreas, pancreas-to-lesion contrast, and image quality more efficiently than did SSFSE 6 mm and FSE 6 mm. Thin-slice fat-suppressed single-shot T2WI with DLIR can be easily implemented for pancreatic MR protocol.

Modified respiratory-triggered SPACE sequences for magnetic resonance cholangiopancreatography

Background

Respiratory-triggered (RT) and breath-hold are the most common acquisition modalities for magnetic resonance cholangiopancreatography (MRCP). The present study compared the three different acquisition modalities for optimizing the use of MRCP in patients with diseases of the pancreatic and biliary systems.

Materials and methods

Three MRCP acquisition modalities were used in this study: conventional respiratory-triggered sampling perfection with application-optimized contrasts using different flip evolutions (RT-SPACE), modified RT-SPACE, and breath-hold (BH)-SPACE. Fifty-eight patients with clinically suspected pancreatic and biliary system disease were included. All image data were acquired on a 1.5 T MR. Scan time and image quality were compared between the three acquisition modalities. Friedman test, which was followed by post-hoc analysis, was performed among triple-scan protocol.

Results

There was a significant difference in the mean acquisition time among conventional RT-SPACE, modified RT-SPACE, and BH-SPACE (167.41±32.11 seconds vs 50.84±73.78 seconds vs 18.00 seconds, P <0.001). Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were also significantly different among the three groups (P <0.001). The SNR and CNR were higher in the RT-SPACE group than in the BH-SPACE group (P <0.05). However, there were no statistically significant differences (P >0.05) among the 3 groups regarding quality of overall image, image clarity, background inhibition, and visualization of the pancreatic and biliary system.

Conclusions

MRCP acquisition with the modified RT-SPACE sequence greatly shortens the acquisition time with comparable quality images. The MRCP acquisition modality could be designed based on the patient's situation to improve the examination pass rate and obtain excellent images for diagnosis.

Flow independent black blood imaging with a large FOV from the neck to the aortic arch: A feasibility study at 3 tesla

PURPOSE

To investigate the feasibility of obtaining black-blood imaging with a large FOV from the neck to the aortic arch at 3 T using a newly modified Relaxation-Enhanced Angiography without Contrast and Triggering for Black-Blood Imaging (REACT-BB).

MATERIALS AND METHODS

REACT-BB provides black-blood images by adjusting the inversion time (TI) in REACT to the null point of blood. The optimal TI for REACT-BB was investigated in 10 healthy volunteers with TI varied from 200 ms to 1400 ms. Contrast ratios were calculated between muscle and three branch arteries of the aortic arch. Additionally, a comparison between REACT-BB and MPRAGE involved evaluating the depiction of high-intensity plaques in 222 patients with stroke or transient ischemic attack. Measurements included plaque-to-muscle signal intensity ratios (PMR), plaque volumes, and carotid artery stenosis rates in 60 patients with high-intensity plaques in carotid arteries.

RESULTS

REACT-BB with TI = 850 ms produced the black-blood image with the best contrast between blood and background tissues. REACT-BB outperformed MPRAGE in depicting high-intensity plaques in the aortic arch (55.4% vs 45.5%) and exhibited superior overall image quality in visual assessment (3.31 ± 0.70 vs 2.89 ± 0.73; p < 0.05). Although the PMR of REACT-BB was significantly lower than MPRAGE (2.227 ± 0.601 vs 2.285 ± 0.662; P < 0.05), a strong positive correlation existed between REACT-BB and MPRAGE (ρ = 0.935; P < 0.05), and all high-intensity plaques that MPRAGE detected were clearly detected by REACT-BB.

CONCLUSION

REACT-BB provides black-blood images with uniformly suppressed fat and blood signals over a large FOV from the neck to the aortic arch with comparable or better high-signal plaque depiction than MPRAGE.

Deep learning denoising reconstruction for improved image quality in fetal cardiac cine MRI

Purpose

This study aims to evaluate deep learning (DL) denoising reconstructions for image quality improvement of Doppler ultrasound (DUS)-gated fetal cardiac MRI in congenital heart disease (CHD).

Methods

Twenty-five fetuses with CHD (mean gestational age: 35 ± 1 weeks) underwent fetal cardiac MRI at 3T. Cine imaging was acquired using a balanced steady-state free precession (bSSFP) sequence with Doppler ultrasound gating. Images were reconstructed using both compressed sensing (bSSFP CS) and a pre-trained convolutional neural network trained for DL denoising (bSSFP DL). Images were compared qualitatively based on a 5-point Likert scale (from 1 = non-diagnostic to 5 = excellent) and quantitatively by calculating the apparent signal-to-noise ratio (aSNR) and contrast-to-noise ratio (aCNR). Diagnostic confidence was assessed for the atria, ventricles, foramen ovale, valves, great vessels, aortic arch, and pulmonary veins.

Results

Fetal cardiac cine MRI was successful in 23 fetuses (92%), with two studies excluded due to extensive fetal motion. The image quality of bSSFP DL cine reconstructions was rated superior to standard bSSFP CS cine images in terms of contrast [3 (interquartile range: 2-4) vs. 5 (4-5), P < 0.001] and endocardial edge definition [3 (2-4) vs. 4 (4-5), P < 0.001], while the extent of artifacts was found to be comparable [4 (3-4.75) vs. 4 (3-4), P = 0.40]. bSSFP DL images had higher aSNR and aCNR compared with the bSSFP CS images (aSNR: 13.4 ± 6.9 vs. 8.3 ± 3.6, P < 0.001; aCNR: 26.6 ± 15.8 vs. 14.4 ± 6.8, P < 0.001). Diagnostic confidence of the bSSFP DL images was superior for the evaluation of cardiovascular structures (e.g., atria and ventricles: P = 0.003).

Conclusion

DL image denoising provides superior quality for DUS-gated fetal cardiac cine imaging of CHD compared to standard CS image reconstruction.