Utilizing 3D Fast Spin Echo Anatomical Imaging to Reduce the Number of Contrast Preparations in $T_{1\rho}$ Quantification of Knee Cartilage Using Learning-Based Methods

• DOI: 10.48550/arxiv.2502.08973
• Publication Date: 2025-02-13
• Authors: Junru Zhong; Chun‐Ying Huang; Zhifeng Yu; Fan Xiao; Siyue Li; Michael Tim‐Yun Ong; Kevin Ki‐Wai Ho; Queenie Chan; James F. Griffith; Weitian Chen
• Keywords: Knee cartilage; Echo (communications protocol); Fast spin echo

Purpose: To propose and evaluate an accelerated $T_{1\rho}$ quantification method that combines $T_{1\rho}$-weighted fast spin echo (FSE) images proton density (PD)-weighted anatomical FSE images, leveraging deep learning models for mapping. The goal is to reduce scan time facilitate integration into routine clinical workflows osteoarthritis (OA) assessment. Methods: This retrospective study utilized MRI data from 40 participants (30 OA patients 10 healthy volunteers). A volume of PD-weighted a acquired at non-zero spin-lock were used as input train models, including 2D U-Net multi-layer perceptron (MLP). maps generated by these compared with ground truth derived traditional non-linear least squares (NLLS) fitting using four images. Evaluation metrics included mean absolute error (MAE), percentage (MAPE), regional (RE), (RPE). Results: Deep achieved RPEs below 5% across all evaluated scenarios, outperforming NLLS methods, especially in low signal-to-noise conditions. best results obtained the U-Net, which effectively leveraged spatial information accurate fitting. proposed demonstrated compatibility shorter TSLs, alleviating RF hardware specific absorption rate (SAR) limitations. Conclusion: approach enables efficient mapping reducing while maintaining standards. has potential quantitative techniques practice, benefiting diagnosis monitoring.