Local response function estimation in spherical deconvolution for comprehensive tissue representation using diffusion MRI

Diffusion MRI (dMRI) plays a crucial role in studying tissue microstructure and fibre orientation. Due to the intricate nature of the dMRI signal, end users require representations that provide a straightforward interpretation. Currently, these representations rely on tissue-average estimations or simplified tissue models and are hence limited in their applicability to pathology. In this study, we propose a novel approach called LoRE-SD-a local response function estimation in spherical deconvolution. LoRE-SD minimises assumptions about tissue microstructure to improve the reconstruction of dMRI data in the presence of pathology. This is achieved by introducing a general signal representation that spans the most common multi-compartment microstructure models used in neuroimaging. Leveraging spherical deconvolution, LoRE-SD provides accurate estimations of the local fibre orientations, allowing tractography in the healthy and pathological brain. We evaluate this approach using simulations and in vivo data from a healthy volunteer and from patients with glioma. Comparing the results quantitatively with the state-of-the-art, we find that LoRE-SD accurately reconstructs fibre orientations across the brain while also significantly improving glioma reconstruction and fibre bundle estimation. Additionally, the tissue representation in LoRE-SD facilitates the generation of various image contrasts, including response function anisotropy and contrasts accentuating intra-axonal, extra-axonal, and free water spaces, which enables a more flexible approach for tractography. In conclusion, LoRE-SD introduces a framework for estimating a data-driven, local representation of tissue microstructure with minimal prior assumptions. This approach provides a new way to represent the human brain, pathology, and other organs using dMRI and opens avenues for defining novel image contrasts, which may benefit tractography.