Effects of Deep Learning-Based Reconstruction on the Quality of Accelerated Contrast-Enhanced Neck MRI

Abstract

Objective: To compare the quality of deep learning-reconstructed turbo spin-echo (DL-TSE) and conventionally interpolated turbo spin-echo (Conv-TSE) techniques in contrast-enhanced MRI of the neck.

Materials and methods: Contrast-enhanced T1-weighted DL-TSE and Conv-TSE images were acquired using 3T scanners from 106 patients. DL-TSE employed a closed-source, 'work-in-progress' (WIP No. 1062, iTSE, version 10; Siemens Healthineers) algorithm for interpolation and denoising to achieve the same in-plane resolution (axial: 0.26 × 0.26 mm²; coronal: 0.29 × 0.29 mm²) while reducing scan times by 15.9% and 52.6% for axial and coronal scans, respectively. The full width at half maximum (FWHM) and percent signal ghosting were measured using stationary and flow phantom scans, respectively. In patient images, non-uniformity (NU), contrast-to-noise ratio (CNR), and regional mucosal FWHM were evaluated. Two neuroradiologists visually rated the patient images for overall quality, sharpness, regional mucosal conspicuity, artifacts, and lesions using a 5-point Likert scale.

Results: FWHM in the stationary phantom scan was consistently sharper in DL-TSE. The percent signal ghosting outside the flow phantom was lower in DL-TSE (0.06% vs. 0.14%) but higher within the phantom (8.92% vs. 1.75%) compared to Conv-TSE. In patient scans, DL-TSE showed non-inferior NU and higher CNR. Regional mucosal FWHM was significantly better in DL-TSE, particularly in the oropharynx (coronal: 1.08 ± 0.31 vs. 1.52 ± 0.46 mm) and hypopharynx (coronal: 1.26 ± 0.35 vs. 1.91 ± 0.56 mm) (both P < 0.001). DL-TSE demonstrated higher overall image quality (axial: 4.61 ± 0.49 vs. 3.32 ± 0.54) and sharpness (axial: 4.40 ± 0.56 vs. 3.11 ± 0.53) (both P < 0.001). In addition, mucosal conspicuity was improved, especially in the oropharynx (axial: 4.41 ± 0.67 vs. 3.40 ± 0.69) and hypopharynx (axial: 4.45 ± 0.58 vs. 3.58 ± 0.63) (both P < 0.001). Extracorporeal ghost artifacts were reduced in DL-TSE (axial: 4.32 ± 0.60 vs. 3.90 ± 0.71, P < 0.001) but artifacts overlapping anatomical structures were slightly more pronounced (axial: 3.78 ± 0.74 vs. 3.95 ± 0.72, P < 0.001). Lesions were detected with higher confidence in DL-TSE.

Conclusion: DL-based reconstruction applied to accelerated neck MRI improves overall image quality, sharpness, mucosal conspicuity in motion-prone regions, and lesion detection confidence. Despite more pronounced ghost artifacts overlapping anatomical structures, DL-TSE enables substantial scan time reduction while enhancing diagnostic performance.

Keywords: Deep learning reconstruction; Full width at half maximum; Head and neck magnetic resonance imaging; Image quality; Image sharpness; Motion artifact.

Fig. 1. Evaluation of image sharpness and flow-related ghost artifacts using stationary and flow phantoms. A, B: Full width at half maximum measurements were performed along the yellow lines across resolution pattern hole arrays (diameters: 0.6, 0.7, and 0.8 mm). The holes appear more sharply visualized in DL-TSE (A) compared to Conv-TSE (B). C-F: In scans of the custom-made flow phantom, flow-related ghost artifacts adjacent to the silicone tube (arrows) are more discrete in DL-TSE (C) than in Conv-TSE (D). Ghost artifacts outside the phantom, including flow-related ghosts (arrows), are less prominent in DL-TSE (E) compared to Conv-TSE (F). Circles in the flow phantom images indicate regions of interest used for signal intensity measurements. DL-TSE = deep learning-reconstructed turbo spin-echo, Conv-TSE = conventional turbo spin-echo

Fig. 2. In vivo measurement of mucosal FWHM across multiple anatomical sites. A: Axial scans of the Rosenmüller fossa show an FWHM of 1.32 mm in DL-TSE (left) and 1.46 mm in Conv-TSE (middle). B: Coronal scans of the soft palate show an FWHM of 1.74 mm in DL-TSE (left) and 2.29 mm in Conv-TSE (middle). C: Axial scans of the pyriform sinus reveal an FWHM of 1.13 mm in DL-TSE (left) compared to 2.03 mm in Conv-TSE (middle). The corresponding signal intensity profiles for each panel are displayed on the right side. FWHM = full width at half maximum, DL-TSE = deep learning-reconstructed turbo spin-echo, Conv-TSE = conventional turbo spin-echo

Fig. 3. Comparison of overlapping and extracorporeal ghost artifacts between DL-TSE and Conv-TSE. A, B: In coronal images, ghosts overlapping the spine and paraspinal structures (arrows) are more accentuated in DL-TSE (A) compared to Conv-TSE (B). However, extracorporeal ghost artifacts (arrows) are less prominent in DL-TSE. C, D: Another coronal DL-TSE (C) shows accentuated streaks in the tongue (arrows) relative to Conv-TSE (D), but the conspicuity of the epiglottis (arrows) is improved, and extracorporeal ghosts (arrows) are absent. E, F: Axial scans show dark streaks caused by bulk motion, more prominent in DL-TSE (E) than in Conv-TSE (F), overlapping the left level III lymphadenopathy and anterior neck structures (arrows). Extracorporeal ghosts (arrows) remain less present in DL-TSE. DL-TSE = deep learning-reconstructed turbo spin-echo, Conv-TSE = conventional turbo spin-echo

Fig. 4. Enhanced lesion depiction using DL-TSE compared to Conv-TSE in clinical cases. A, B: Axial scans of a 22-year-old male with nasopharyngeal carcinoma. DL-TSE (A) provides clearer visualization of both the primary tumor and retropharyngeal lymphadenopathy (arrows) compared to Conv-TSE (B). C, D: Axial scans of a 67-year-old male with hoarseness reveal vocal cord evaluation. DL-TSE (C) highlights a focal surface irregularity (arrows) not detected in Conv-TSE (D) due to motion-related edge jitter. Glottic squamous cell carcinoma at the site indicated by DL-TSE was later confirmed by laryngoscopy (C, right upper) and subsequent surgery. E, F: Coronal scans of a 54-year old male post-partial glossectomy. In Conv-TSE (F), the patchy contrast-enhancing lesion (arrows) shows less volume effect, potentially leading to misinterpretation as postoperative change. DL-TSE (E) more clearly delineates the mass, later confirmed as recurrent squamous cell carcinoma of the tongue. DL-TSE = deep learning-reconstructed turbo spin-echo, Conv-TSE = conventional turbo spin-echo