Faster Elbow MRI with Deep Learning Reconstruction-Assessment of Image Quality, Diagnostic Confidence, and Anatomy Visualization Compared to Standard Imaging

Abstract

Objective: The objective of this study was to evaluate a deep learning (DL) reconstruction for turbo spin echo (TSE) sequences of the elbow regarding image quality and visualization of anatomy.

Materials and methods: Between October 2020 and June 2021, seventeen participants (eight patients, nine healthy subjects; mean age: 43 ± 16 (20-70) years, eight men) were prospectively included in this study. Each patient underwent two examinations: standard MRI, including TSE sequences reconstructed with a generalized autocalibrating partial parallel acquisition reconstruction (TSE_STD), and prospectively undersampled TSE sequences reconstructed with a DL reconstruction (TSE_DL). Two radiologists evaluated the images concerning image quality, noise, edge sharpness, artifacts, diagnostic confidence, and delineation of anatomical structures using a 5-point Likert scale, and rated the images concerning the detection of common pathologies.

Results: Image quality was significantly improved in TSE_DL (mean 4.35, IQR 4-5) compared to TSE_STD (mean 3.76, IQR 3-4, p = 0.008). Moreover, TSE_DL showed decreased noise (mean 4.29, IQR 3.5-5) compared to TSE_STD (mean 3.35, IQR 3-4, p = 0.004). Ratings for delineation of anatomical structures, artifacts, edge sharpness, and diagnostic confidence did not differ significantly between TSE_DL and TSE_STD (p > 0.05). Inter-reader agreement was substantial to almost perfect (κ = 0.628-0.904). No difference was found concerning the detection of pathologies between the readers and between TSE_DL and TSE_STD. Using DL, the acquisition time could be reduced by more than 35% compared to TSE_STD.

Conclusion: TSE_DL provided improved image quality and decreased noise while receiving equal ratings for edge sharpness, artifacts, delineation of anatomical structures, diagnostic confidence, and detection of pathologies compared to TSE_STD. Providing more than a 35% reduction of acquisition time, TSE_DL may be clinically relevant for elbow imaging due to increased patient comfort and higher patient throughput.

Keywords: MRI; acceleration; deep learning reconstruction; elbow; image processing.

Figure 1

Non-contrast elbow MRI acquired at 1.5 T in a 51-year-old female patient with the institution’s standard turbo spin echo (TSE) sequences (TSE_STD, upper row) and deep learning-reconstructed TSE sequences (TSE_DL, lower row). Increased edge sharpness in the TSE_DL images is primarily seen in the coronal and sagittal T1-weighted sequences (T1 cor and T1 sag). In the fat-saturated coronal and axial PD-weighted images (PD cor and PD ax) the DL reconstruction shows decreased noise.

Figure 2

Non-contrast elbow MRI acquired at 3 T in a 70-year-old male patient with the institution’s standard turbo spin echo (TSE) sequences (TSE_STD, upper row) and deep learning-reconstructed TSE sequences (TSE_DL, lower row). In this example the standard and DL-reconstructed images show very similar image quality. In the TSE_DL images, all anatomic details are well depicted, showing that the reduced acquisition time did not lead to a loss in detail. Additionally, edge sharpness in the coronal T1-weighted images (T1 cor) is slightly increased in the TSE_DL image.

Figure 3

Elbow MRI acquired at 1.5 T in a 36-year-old male patient with clinically suspected biceps tendon tear after trauma with the institution’s standard turbo spin echo (TSE) sequences (TSE_STD, upper row) and deep learning-reconstructed TSE sequences (TSE_DL, lower row). The MRI examination confirmed the suspicion of a partial tear of the biceps’ tendon at its insertion with a peritendinous hematoma. Noise is reduced in the DL-reconstructed axial and sagittal fat-suppressed PD-weighted images (PD ax and PD sag). The intact ulnar collateral ligament is depicted well in the standard and DL-reconstructed coronal T1-weighted images; the difference in angulation between the two images is due to movement between acquisitions.

Figure 4

Elbow MRI acquired at 1.5 T with the institution’s standard turbo spin echo (TSE) sequences (TSE_STD, upper row) and deep learning reconstructed TSE sequences (TSE_DL, lower row) in a 66-year-old female patient who had suffered an elbow dislocation. The patient could not extend the elbow for the examination. The images show an avulsion of the common flexor tendon with hemorrhage into the flexor muscles (PD cor and PD cor), a complete tear of the radial collateral ligament (PD cor, third column) and a dislocated annular ligament (PD sag). The fat-supressed PD-weighted TSE_DL images show improved noise, while the DL-reconstructed coronal T1-weighted image shows very similar image quality compared to the standard image.