GACELLE: GPU-AcCELerated toolbox for high-throughput multidimensionaL quantitative parameter Estimation

Motivation: Parameter mapping is a computationally intensive task in quantitative MRI with demands increase with model complexity and higher-resolution data, hindering their applicability in research and clinical applications. Goal(s): To develop a tool facilitating GPU acceleration with deep learning optimisers in parameter mapping, enabling researchers to quickly apply it to their specific domain. Approach: We introduce GACELLE, a Matlab package for GPU-accelerated minimisation-based and Markov-Chain-Monte-Carlo estimations, facilitating high-throughput processing of hundreds of thousands of voxels simultaneously. Results: GACELLE achieves between 360- and 2210-fold acceleration in our demonstration, with high-quality results comparable to conventional methods. Full documentation and tutorials are provided online for easy dissemination. Impact: GACELLE removes the obstacle of time-consuming data processing associated with quantitative MRI multi-parametric non-linear problems, promoting wider adoption of quantitative MRI for the scientific community.