Utilisation of Clinical Knee MRI to Accelerate Quantitative T1ρ Imaging of Knee

Motivation: To accelerate the acquisition of quantitative knee T1ρ MRI by using clinical proton density (PD)-weighted image. Goal(s): Achieve an accurate T1ρ quantification on knee cartilage using clinical PD-weighted image against the traditional mono-exponential fitting. Approach: 30 knee osteoarthritis (OA) patients and 10 healthy volunteers underwent turbo spin echo (TSE) PD and T1ρ MRI. A multi-layer perceptron (MLP) neural network produced T1ρ values from one T1ρ dynamic scan (spin-lock time, TSL=50ms) and PD-weighted intensities. Results: We achieved an agreement against the ground truth with a regional-averaged absolute and percentage error within the cartilage region at 3.80 ms and 8.08%. Impact: Our proposed technology reduces the T1ρ contrast needed for fitting using the clinical PD-weighted MRI. This can contribute to clinical adoption of T1ρ.