Compressed Sensing 3D-GRASE for faster High-Resolution MRI

Active Learning in Brain Tumor Segmentation with Uncertainty Sampling and Annotation Redundancy Restriction

Deep learning models have demonstrated great potential in medical imaging but are limited by the expensive, large volume of annotations required. To address this, we compared different active learning strategies by training models on subsets of the most informative images using real-world clinical datasets for brain tumor segmentation and proposing a framework that minimizes the data needed while maintaining performance. Then, 638 multi-institutional brain tumor magnetic resonance imaging scans were used to train three-dimensional U-net models and compare active learning strategies. Uncertainty estimation techniques including Bayesian estimation with dropout, bootstrapping, and margins sampling were compared to random query. Strategies to avoid annotating similar images were also considered. We determined the minimum data necessary to achieve performance equivalent to the model trained on the full dataset (α = 0.05). Bayesian approximation with dropout at training and testing showed results equivalent to that of the full data model (target) with around 30% of the training data needed by random query to achieve target performance (p = 0.018). Annotation redundancy restriction techniques can reduce the training data needed by random query to achieve target performance by 20%. We investigated various active learning strategies to minimize the annotation burden for three-dimensional brain tumor segmentation. Dropout uncertainty estimation achieved target performance with the least annotated data.

Comparison of utility of deep learning reconstruction on 3D MRCPs obtained with three different k-space data acquisitions in patients with IPMN

OBJECTIVE

To compare the utility of deep learning reconstruction (DLR) for improving acquisition time, image quality, and intraductal papillary mucinous neoplasm (IPMN) evaluation for 3D MRCP obtained with parallel imaging (PI), multiple k-space data acquisition for each repetition time (TR) technique (Fast 3D mode multiple: Fast 3Dm) and compressed sensing (CS) with PI.

MATERIALS AND METHODS

A total of 32 IPMN patients who had undergone 3D MRCPs obtained with PI, Fast 3Dm, and CS with PI and reconstructed with and without DLR were retrospectively included in this study. Acquisition time, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) obtained with all protocols were compared using Tukey's HSD test. Results of endoscopic ultrasound, ERCP, surgery, or pathological examination were determined as standard reference, and distribution classifications were compared among all 3D MRCP protocols by McNemar's test.

RESULTS

Acquisition times of Fast 3Dm and CS with PI with and without DLR were significantly shorter than those of PI with and without DLR (p < 0.05). Each MRCP sequence with DLR showed significantly higher SNRs and CNRs than those without DLR (p < 0.05). IPMN distribution accuracy of PI with and without DLR and Fast 3Dm with DLR was significantly higher than that of Fast 3Dm without DLR and CS with PI without DLR (p < 0.05).

CONCLUSION

DLR is useful for improving image quality and IPMN evaluation capability on 3D MRCP obtained with PI, Fast 3Dm, or CS with PI. Moreover, Fast 3Dm and CS with PI may play as substitution to PI for MRCP in patients with IPMN.

KEY POINTS

• Mean examination times of multiple k-space data acquisitions for each TR and compressed sensing with parallel imaging were significantly shorter than that of parallel imaging (p < 0.0001). • When comparing image quality of 3D MRCPs with and without deep learning reconstruction, deep learning reconstruction significantly improved signal-to-noise ratio and contrast-to-noise ratio (p < 0.05). • IPMN distribution accuracies of parallel imaging with and without deep learning reconstruction (with vs. without: 88.0% vs. 88.0%) and multiple k-space data acquisitions for each TR with deep learning reconstruction (86.0%) were significantly higher than those of others (p < 0.05).

Sub-millisecond 2D MRI of the vocal fold oscillation using single-point imaging with rapid encoding

OBJECTIVE

The slow spatial encoding of MRI has precluded its application to rapid physiologic motion in the past. The purpose of this study is to introduce a new fast acquisition method and to demonstrate feasibility of encoding rapid two-dimensional motion of human vocal folds with sub-millisecond resolution.

METHOD

In our previous work, we achieved high temporal resolution by applying a rapidly switched phase encoding gradient along the direction of motion. In this work, we extend phase encoding to the second image direction by using single-point imaging with rapid encoding (SPIRE) to image the two-dimensional vocal fold oscillation in the coronal view. Image data were gated using electroglottography (EGG) and motion corrected. An iterative reconstruction with a total variation (TV) constraint was used and the sequence was also simulated using a motion phantom.

RESULTS

Dynamic images of the vocal folds during phonation at pitches of 150 and 165 Hz were acquired in two volunteers and the periodic motion of the vocal folds at a temporal resolution of about 600 µs was shown. The simulations emphasize the necessity of SPIRE for two-dimensional motion encoding.

DISCUSSION

SPIRE is a new MRI method to image rapidly oscillating structures and for the first time provides dynamic images of the vocal folds oscillations in the coronal plane.

Fast 3D Isotropic Proton Density-Weighted Fat-Saturated MRI of the Knee at 1.5 T with Compressed Sensing: Comparison with Conventional Multiplanar 2D Sequences

PURPOSE

 Compressed sensing (CS) is a method to accelerate MRI acquisition by acquiring less data through undersampling of k-space. In this prospective study we aimed to evaluate whether a three-dimensional (3D) isotropic proton density-weighted fat saturated sequence (PDwFS) with CS can replace conventional multidirectional two-dimensional (2D) sequences at 1.5 Tesla.

MATERIALS AND METHODS

 20 patients (45.2 ± 20.2 years; 10 women) with suspected internal knee damage received a 3D PDwFS with CS acceleration factor 8 (acquisition time: 4:11 min) in addition to standard three-plane 2D PDwFS sequences (acquisition time: 4:05 min + 3:03 min + 4:46 min = 11:54 min) at 1.5 Tesla. Scores for homogeneity of fat saturation, image sharpness, and artifacts were rated by two board-certified radiologists on the basis of 5-point Likert scales. Based on these ratings, an overall image quality score was generated. Additionally, quantitative contrast ratios for the menisci (MEN), the anterior (ACL) and the posterior cruciate ligament (PCL) in comparison with the popliteus muscle were calculated.

RESULTS

 The overall image quality was rated superior in 3D PDwFS compared to 2D PDwFS sequences (14.45 ± 0.83 vs. 12.85 ± 0.99; p < 0.01), particularly due to fewer artifacts (4.65 ± 0.67 vs. 3.65 ± 0.49; p < 0.01) and a more homogeneous fat saturation (4.95 ± 0.22 vs. 4.55 ± 0.51; p < 0.01). Scores for image sharpness were comparable (4.80 ± 0.41 vs. 4.65 ± 0.49; p = 0.30). Quantitative contrast ratios for all measured structures were superior in 3D PDwFS (MEN: p < 0.05; ACL: p = 0.06; PCL: p = 0.33). In one case a meniscal tear was only diagnosed using multiplanar reformation of 3D PDwFS, but it would have been missed on standard multiplanar 2D sequences.

CONCLUSION

 An isotropic fat-saturated 3D PD sequence with CS enables fast and high-quality 3D imaging of the knee joint at 1.5 T and may replace conventional multiplanar 2D sequences. Besides faster image acquisition, the 3D sequence provides advantages in small structure imaging by multiplanar reformation.

KEY POINTS

  · 3D PDwFS with compressed sensing enables knee imaging that is three times faster compared to multiplanar 2D sequences. · 3D PDwFS with compressed sensing provides high-quality knee imaging at 1.5 T. · Isotropic 3D sequences provide advantages in small structure imaging by using multiplanar reformations.

CITATION FORMAT

· Endler CH, Faron A, Isaak A et al. Fast 3D Isotropic Proton Density-Weighted Fat-Saturated MRI of the Knee at 1.5 T with Compressed Sensing: Comparison with Conventional Multiplanar 2D Sequences. Fortschr Röntgenstr 2021; 193: 813 - 821.

Compressed sensing and parallel imaging accelerated T2 FSE sequence for head and neck MR imaging: Comparison of its utility in routine clinical practice

PURPOSE

To directly compare the capability of compressed sensing (CS) and parallel imaging (PI) accelerated T2 FSE (Fast Spin Echo) sequence with PI for head and neck MR imaging.

METHODS

Thirty consecutive patients with various head and neck diseases (15 men and 15 women, mean age 53 ± 22 years) underwent MR imaging by PI with CS and by PI. Reduction factors were as follows: PI with CS, 3 and PI, 1.5. Examination times for PI with CS and PI were all recorded. For quantitative image quality assessment, signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated. For qualitative assessment, two investigators assessed overall image quality, artifacts and diagnostic confidence level using a 5-point scoring system, and final scores were determined by consensus of two readers. Mean examination time and all indexes were compared by means of paired t-test and Wilcoxon signed-rank test. Inter-observer agreement for each qualitative index was assessed in terms of kappa statistics.

RESULTS

Mean examination time for PI with CS (83.5 ± 11.0 s) was significantly shorter than that for PI (173.0 ± 54.4 s, p < 0.0001). SNR and CNR of PI with CS were significantly better than those with PI (mean SNR; 11.2 ± 3.6 vs 8.9 ± 2.6, median of CNR; 7.4 vs. 6.1, p < 0.0001). All inter-observer agreements were assessed as significant and substantial (0.62 < κ < 0.81).

CONCLUSION

PI with CS accelerated T2 weighted sequence performs equally well or even slightly better than its PI accelerated, conventional counterpart at reduced scan times in the context of head and neck MR imaging.

Compressed sensing and deep learning reconstruction for women's pelvic MRI denoising: Utility for improving image quality and examination time in routine clinical practice

PURPOSE

To demonstrate the utility of compressed sensing with parallel imaging (Compressed SPEEDER) and AiCE compared with that of conventional parallel imaging (SPEEDER) for shortening examination time and improving image quality of women's pelvic MRI.

METHOD

Thirty consecutive patients with women's pelvic diseases (mean age 50 years) underwent T2-weighted imaging using Compressed SPEEDER as well as conventional SPEEDER reconstructed with and without AiCE. The examination times were recorded, and signal-to-noise ratio (SNR) was calculated for every patient. Moreover, overall image quality was assessed using a 5-point scoring system, and final scores for all patients were determined by consensus of two readers. Mean examination time, SNR and overall image quality were compared among the four data sets by Wilcoxon signed-rank test.

RESULTS

Examination times for Compressed SPEEDER with and without AiCE were significantly shorter than those for conventional SPEEDER with and without AiCE (with AiCE: p < 0.0001, without AiCE: p < 0.0001). SNR of Compressed SPEEDER and of SPEEDER with AiCE was significantly superior to that of Compressed SPEEDER without AiCE (vs. Compressed SPEEDER, p = 0.01; vs. SPEEDER, p = 0.009). Overall image quality of Compressed SPEEDER with AiCE and of SPEEDER with and without AiCE was significantly higher than that of Compressed SPEEDER without AiCE (vs. Compressed SPEEDER with AiCE, p < 0.0001; vs. SPEEDER with AiCE, p < 0.0001; SPEEDER without AiCE, p = 0.0003).

CONCLUSION

Image quality and shorten examination time for T2-weighted imaging in women's pelvic MRI can be significantly improved by using Compressed SPEEDER with AiCE in comparison with conventional SPEEDER, although other sequences were not tested.

Rotated spiral RARE for high spatial and temporal resolution volumetric arterial spin labeling acquisition

Background:

Arterial Spin Labeling (ASL) MRI can provide quantitative images that are sensitive to both time averaged blood flow and its temporal fluctuations. 3D image acquisitions for ASL are desirable because they are more readily compatible with background suppression to reduce noise, can reduce signal loss and distortion, and provide uniform flow sensitivity across the brain. However, single-shot 3D acquisition for maximal temporal resolution typically involves degradation of image quality through blurring or noise amplification by parallel imaging. Here, we report a new approach to accelerate a common stack of spirals 3D image acquisition by pseudo golden-angle rotation and compressed sensing reconstruction without any degradation of time averaged blood flow images.

Methods:

28 healthy volunteers were imaged at 3T with background-suppressed unbalanced pseudo-continuous ASL combined with a pseudo golden-angle Stack-of-Spirals 3D RARE readout. A fully-sampled perfusion-weighted volume was reconstructed by 3D non-uniform Fast Fourier Transform (nuFFT) followed by sum-of-squares combination of the 32 individual channels. Coil sensitivities were estimated followed by reconstruction of the 39 single-shot volumes using an L₁-wavelet Compressed-Sensing reconstruction. Finally, brain connectivity analyses were performed in regions where BOLD signal suffers from low signal-to-noise ratio and susceptibility artifacts.

Results:

Image quality, assessed with a non-reference 3D blurring metric, of full time averaged blood flow was comparable to a conventional interleaved acquisition. The temporal resolution provided by the acceleration enabled identification and quantification of resting-state networks even in inferior regions such as the amygdala and inferior frontal lobes, where susceptibility artifacts can degrade conventional resting-state fMRI acquisitions.

Conclusion:

This approach can provide measures of blood flow modulations and resting-state networks for free within any research or clinical protocol employing ASL for resting blood flow.

Diagnostic quality assessment of IR-prepared 3D magnetic resonance neuroimaging accelerated using compressed sensing and k-space sampling order optimization

PURPOSE

To evaluate the clinical diagnostic efficacy of accelerated 3D magnetic resonance (MR) neuroimaging by radiological assessment for image quality and artefacts.

STUDY TYPE

Prospective healthy volunteer study.

SUBJECTS

Eight healthy subjects.

FIELD STRENGTH/SEQUENCE

Inversion Recovery (IR) prepared 3D Gradient Echo (GRE) sequence on a 1.5 T GE Signa HDx scanner.

ASSESSMENT

Independent radiological diagnostic quality assessments of accelerated 3D MR brain datasets were carried out by four experienced neuro-radiologists who were blinded to the acceleration factor and to the subject. The radiological grading was based on a previously reported radiological scoring key that was used for image quality assessment of human brains.

STATISTICAL TESTS

Bland-Altman analysis.

RESULTS

Optimization of the k-space sampling order was important for preserving contrast in accelerated scans. Despite having lower scores than fully sampled datasets, the majority of the compressed sensing (CS) accelerated brain datasets with k-space sampling order optimization (19/24 datasets by Radiologist 1, 24/24 datasets by Radiologist 2 and 16/24 datasets by Radiologist 3) were graded to be fully diagnostic indicating that there was adequate confidence for performing gross structural assessment of the brain.

CONCLUSION

Optimization of k-space acquisition order improves the clinical utility of CS accelerated 3D neuroimaging. This method may be appropriate for routine radiological assessment of the brain.

Fast and high-resolution myelin water imaging: Accelerating multi-echo GRASE with CAIPIRINHA

PURPOSE

Although several MRI methods have been explored to achieve in vivo myelin quantification, imaging the whole brain in clinically acceptable times and sufficiently high resolution remains challenging. To address this problem, this work investigates the acceleration of multi-echo T₂ acquisitions based on the multi-echo gradient and spin echo (GRASE) sequence using CAIPIRINHA undersampling and adapted k-space reordering patterns.

METHODS

A prototype multi-echo GRASE sequence supporting CAIPIRINHA parallel imaging was implemented. Multi-echo T₂ data were acquired from 12 volunteers using the implemented sequence (1.6 × 1.6 × 1.6 mm³ , 84 slices, acquisition time [TA] = 10:30 min) and a multi-echo spin echo (MESE) sequence as reference (1.6 × 1.6 × 3.2 mm³ , single-slice, TA = 5:41 min). Myelin water fraction (MWF) maps derived from both acquisitions were compared via correlation and Bland-Altman analyses. In addition, scan-rescan datasets were acquired to evaluate the repeatability of the derived maps.

RESULTS

Resulting maps from the MESE and multi-echo GRASE sequences were found to be correlated (r = 0.83). The Bland-Altman analysis revealed a mean bias of -0.2% (P = .24) with the limits of agreement ranging from -3.7% to 3.3%. The Pearson's correlation coefficient among MWF values obtained from the scan-rescan datasets was found to be 0.95 and the mean bias equal to 0.11% (P = .32), indicating good repeatability of the retrieved maps.

CONCLUSION

By combining a 3D multi-echo GRASE sequence with CAIPIRINHA sampling, whole-brain MWF maps were obtained in 10:30 min with 1.6 mm isotropic resolution. The good correlation with conventional MESE-based maps demonstrates that the implemented sequence may be a promising alternative to time-consuming MESE acquisitions.

Multi-spin echo T2 relaxation imaging with compressed sensing (METRICS) for rapid myelin water imaging

PURPOSE

Myelin water imaging (MWI) provides a valuable biomarker for myelin, but clinical application has been restricted by long acquisition times. Accelerating the standard multi-echo T₂ acquisition with gradient echoes (GRASE) or by 2D multi-slice data collection results in image blurring, contrast changes, and other issues. Compressed sensing (CS) can vastly accelerate conventional MRI. In this work, we assessed the use of CS for in vivo human MWI, using a 3D multi spin-echo sequence.

METHODS

We implemented multi-echo T₂ relaxation imaging with compressed sensing (METRICS) and METRICS with partial Fourier acceleration (METRICS-PF). Scan-rescan data were acquired from 12 healthy controls for assessment of repeatability. MWI data were acquired for METRICS in 9 m:58 s and for METRICS-PF in 7 m:25 s, both with 1.5 × 2 × 3 mm³ voxels, 56 echoes, 7 ms ΔTE, and 240 × 240 × 170 mm³ FOV. METRICS was compared with a novel multi-echo spin-echo gold-standard (MSE-GS) MWI acquisition, acquired for a single additional subject in 2 h:2 m:40 s.

RESULTS

METRICS/METRICS-PF myelin water fraction had mean: repeatability coefficient 1.5/1.1, coefficient of variation 6.2/4.5%, and intra-class correlation coefficient 0.79/0.84. Repeatability metrics comparing METRICS with METRICS-PF were similar, and both sequences agreed with reference values from literature. METRICS images and quantitative maps showed excellent qualitative agreement with those of MSE-GS.

CONCLUSION

METRICS and METRICS-PF provided highly repeatable MWI data without the inherent disadvantages of GRASE or 2D multi-slice acquisition. CS acceleration allows MWI data to be acquired rapidly with larger FOV, higher estimated SNR, more isotropic voxels and more echoes than with previous techniques. The approach introduced here generalizes to any multi-component T₂ mapping application.

Compressed Sensing 3D-GRASE for faster High-Resolution MRI

Purpose

High‐resolution three‐dimensional (3D) structural MRI is useful for delineating complex or small structures of the body. However, it requires long acquisition times and high SAR, limiting its clinical use. The purpose of this work is to accelerate the acquisition of high‐resolution images by combining compressed sensing and parallel imaging (CSPI) on a 3D‐GRASE sequence and to compare it with a (CS)PI 3D‐FSE sequence. Several sampling patterns were investigated to assess their influence on image quality.

Methods

The proposed k‐space sampling patterns are based on two undersampled k‐space grids, variable density (VD) Poisson‐disc, and VD pseudo‐random Gaussian, and five different trajectories described in the literature. Bloch simulations are performed to obtain the transform point spread function and evaluate the coherence of each sampling pattern. Image resolution was assessed by the full‐width at half‐maximum (FWHM). Prospective CSPI 3D‐GRASE phantom and in vivo experiments in knee and brain are carried out to assess image quality, SNR, SAR, and acquisition time compared to PI 3D‐GRASE, PI 3D‐FSE, and CSPI 3D‐FSE acquisitions.

Results

Sampling patterns with VD Poisson‐disc obtain the lowest coherence for both PD‐weighted and T2‐weighted acquisitions. VD pseudo‐random Gaussian obtains lower FWHM, but higher sidelobes than VD Poisson‐disc. CSPI 3D‐GRASE reduces acquisition time (43% for PD‐weighted and 40% for T2‐weighted) and SAR (∼45% for PD‐weighted and T2‐weighted) compared to CSPI 3D‐FSE.

Conclusions

CSPI 3D‐GRASE reduces acquisition time compared to a CSPI 3DFSE acquisition, preserving image quality. The design of the sampling pattern is crucial for image quality in CSPI 3D‐GRASE image acquisitions.