Deep learning–accelerated contrast-enhanced 3D MRI for acute optic neuritis: faster acquisition and improved lesion conspicuity
- 1Department of Radiology, Renmin Hospital of Wuhan University
- 2MR Research, GE Healthcare
- 3Department of Ophthalmology, Renmin Hospital of Wuhan University
Abstract
Objective: To compare a deep learning-accelerated contrast-enhanced 3D T1-weighted sequence (T1-CUBE-DLS) with standard parallel imaging-accelerated T1-CUBE for detecting acute optic neuritis (ON).
Methods: This prospective study enrolled 59 patients with suspected acute ON (38 females; mean age 41.2 ± 12.8 years). All underwent contrast-enhanced whole-brain 3.0 T MRI including both T1-CUBE-DLS and standard T1-CUBE sequences. Two blinded neuroradiologists independently scored motion artifacts, subjective noise, and optic nerve clarity. Consensus readings recorded enhancement laterality, involved segments, and visual conspicuity. Enhancement length was measured per optic nerve, with inter-observer reliability assessed via intraclass correlation coefficient (ICC) and Bland–Altman analysis. Clinical diagnosis by a senior neuro-ophthalmologist served as the reference standard.
Results: Acute ON was confirmed in 43 of 59 patients (72.9%). T1-CUBE-DLS reduced acquisition time by 36.2%. Subjective noise was equivalent between sequences ( p > 0.05), but T1-CUBE-DLS showed significantly fewer motion artifacts and superior optic nerve clarity (both p < 0.05). No significant difference was detected in diagnostic performance between the two sequences: sensitivity (88.4% vs. 86.0%), specificity (93.8% vs. 93.8%), and accuracy (89.8% vs. 88.1%; all p > 0.05). T1-CUBE-DLS demonstrated stronger visual conspicuity ( p = 0.011) and slightly longer measured enhancement length (mean difference, 0.836 mm; p < 0.0001). ICC for length measurement was 0.91 for T1-CUBE-DLS vs. 0.83 for standard T1-CUBE, with narrower 95% limits of agreement (−1.37 to 1.51 mm vs. − 3.1 to 2.4 mm).
Conclusion: T1-CUBE-DLS reduces acquisition time by 36.2%, with no significant difference in diagnostic accuracy detected, while improving enhancement conspicuity and measurement reproducibility, supporting its potential as a practical and efficient alternative for ON evaluation.
Related
- AI-accelerated compressed sensing and deep learning reconstruction for colon MRI: prospective validation of image quality and preliminary assessment of staging accuracy versus contrast-enhanced CT
- Optimizing Portal Venous Phase Acquisition Quality of Contrast-Enhanced CT in Oncology
- Prediction of gross extrathyroidal extension in papillary thyroid carcinoma using a deep learning radiomics model based on post-processing techniques for multi-phase contrast-enhanced MRI images
- Deep learning-based segmentation and subtype prediction of renal cell carcinoma on contrast-enhanced CT
- Predicting microvascular invasion in hepatocellular carcinoma using contrast-enhanced CT-based deep learning models: a comparative study of transformer, CNN, and MIL
BibTeX
@article{liu2026deep,
title = {{Deep learning–accelerated contrast-enhanced 3D MRI for acute optic neuritis: faster acquisition and improved lesion conspicuity}},
author = {Ying Liu and Weiyin Vivian Liu and Zhi Wen and Min Lu and Changsheng Liu and Liang Li and Ting Chen and Renjie Yang and Yunfei Zha},
journal = {Frontiers in Neurology},
year = {2026},
month = oct,
doi = {10.3389/fneur.2026.1954788},
url = {https://www.frontiersin.org/articles/10.3389/fneur.2026.1954788/full},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Deep learning–accelerated contrast-enhanced 3D MRI for acute optic neuritis: faster acquisition and improved lesion conspicuity
AU - Ying Liu
AU - Weiyin Vivian Liu
AU - Zhi Wen
AU - Min Lu
AU - Changsheng Liu
AU - Liang Li
AU - Ting Chen
AU - Renjie Yang
AU - Yunfei Zha
PY - 2026
DA - 2026/10/15
JO - Frontiers in Neurology
DO - 10.3389/fneur.2026.1954788
UR - https://www.frontiersin.org/articles/10.3389/fneur.2026.1954788/full
AB - Objective: To compare a deep learning-accelerated contrast-enhanced 3D T1-weighted sequence (T1-CUBE-DLS) with standard parallel imaging-accelerated T1-CUBE for detecting acute optic neuritis (ON). Methods: This prospective study enrolled 59 patients with suspected acute ON (38 females; mean age 41.2 ± 12.8 years). All underwent contrast-enhanced whole-brain 3.0 T MRI including both T1-CUBE-DLS and standard T1-CUBE sequences. Two blinded neuroradiologists independently scored motion artifacts, subjective noise, and optic nerve clarity. Consensus readings recorded enhancement laterality, involved segments, and visual conspicuity. Enhancement length was measured per optic nerve, with inter-observer reliability assessed via intraclass correlation coefficient (ICC) and Bland–Altman analysis. Clinical diagnosis by a senior neuro-ophthalmologist served as the reference standard. Results: Acute ON was confirmed in 43 of 59 patients (72.9%). T1-CUBE-DLS reduced acquisition time by 36.2%. Subjective noise was equivalent between sequences ( p > 0.05), but T1-CUBE-DLS showed significantly fewer motion artifacts and superior optic nerve clarity (both p < 0.05). No significant difference was detected in diagnostic performance between the two sequences: sensitivity (88.4% vs. 86.0%), specificity (93.8% vs. 93.8%), and accuracy (89.8% vs. 88.1%; all p > 0.05). T1-CUBE-DLS demonstrated stronger visual conspicuity ( p = 0.011) and slightly longer measured enhancement length (mean difference, 0.836 mm; p < 0.0001). ICC for length measurement was 0.91 for T1-CUBE-DLS vs. 0.83 for standard T1-CUBE, with narrower 95% limits of agreement (−1.37 to 1.51 mm vs. − 3.1 to 2.4 mm). Conclusion: T1-CUBE-DLS reduces acquisition time by 36.2%, with no significant difference in diagnostic accuracy detected, while improving enhancement conspicuity and measurement reproducibility, supporting its potential as a practical and efficient alternative for ON evaluation.
SN - 1664-2295
ER -