The value of Fast Dixon combined with deep learning technology in contrast agent-free high-resolution magnetic resonance imaging of the brachial plexus

Abstract

Introduction: This study aimed to investigate the application of Fast Dixon combined with the deep resolve gain (DRG) technique for enhancing the image quality of the brachial plexus on high-resolution MRI without the use of contrast agents.

Methods: Heavily T2-weighted Fast Dixon high-resolution coronal thin-slice magnetic resonance imaging was conducted on 19 social volunteers. Post-scan, the original data underwent reconstruction using deep learning-based denoising technology. Subjective quality scores were assigned to both the original and MIP images, and those processed with and without denoising technology were compared. The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) values for each segment of the bilateral brachial plexus were measured and analyzed to assess image quality.

Results: Subjective evaluations revealed that the quality of both original thin-slice and thin-MIP images processed with the DRG significantly outperformed those processed without the DRG (original thin-slice p = 0.005, thin-MIP p < 0.05). The bilateral SNRs and CNRs of each anatomical structure (root, trunk, cord, branch) of the brachial plexus in the experimental group with DRG were significantly greater than those in the control group without DRG (p < 0.05), as follows: the SNRs of the bilateral nerve roots increased by 35.1-36.2%, the SNRs of the bilateral nerve trunks increased by 40.6-40.8%, and the SNRs of the bilateral nerve cords and branches increased by about 40-45%. The CNR of the bilateral nerve roots increased by 43.1-44.6%, the CNR of the bilateral nerve trunks increased by 41.8-41.7%, and the CNR of the bilateral nerve cords and branches increased by 47.3-50.6% (root p < 0.001, trunk p < 0.001, cord p = 0.001, branch p = 0.011).

Conclusion: Fast Dixon T2WI can enhance the visibility of brachial plexus segments to a certain extent through DRG denoising technology, which may be an effective means to visualize the brachial plexus without contrast agent.

Keywords: Fast Dixon; brachial plexus; contrast agent-free; deep learning; magnetic resonance imaging.

Figure 1

Subjective rating of the original image.

Figure 2

MIP score for each segment of the brachial plexus.

Figure 3

Comparison of the control and experimental groups after the addition of thin-MIP from a 35-year-old male. A,B present comparative 10 mm thin-MIP images under identical window width/level settings. B (post-DRG) highlight the localized adaptive denoising capability of the DRG-integrated method within central regions (delineated by white circles), which contrasts with A (without-DRG) a conventional global filter. The DRG technique significantly improves the SNR and nerve-to-background contrast in anatomically complex areas, such as the cervicothoracic junction and supraclavicular fossa, while maintaining structural fidelity.

Figure 4

Comparison of the control and experimental groups after thin-MIP from a 59-year-old female. C,D present comparative 10 mm thin-MIP images under identical window width/level settings. Compared with C (without-DRG), D (post-DRG) has better denoising ability and neural background contrast.

Figure 5

Comparison of T2WI Fast Dixon measurements before and after DRG combination in displaying partial segments of the brachial plexus. A: without DRG; B: with DRG.

Figure 6

Comparative histogram analysis of the control and experimental groups.

Figure 7

Timing diagram of the T2WI Fast Dixon sequence. The diagram of the Fast Dixon sequence shows opposed phase and in phase echoes acquired in the same repetition time between a pair of refocusing pulses (row 2) and the corresponding readout gradient (row 3). ADC, Analog-to-digital converter.