DMRI shell synthesis improves quality using volumetric harmonics and DTI guidance
DTI-Guided Volumetric Spherical Harmonics Regression for Single-to-Multi-Shell dMRI Synthesis
Computer Vision and Pattern Recognition
Summary
Multi-shell diffusion MRI scans provide richer details about the brain than single-shell scans but take longer to acquire. The authors propose a method to create missing multi-shell scan data from quicker single-shell scans, using mathematical harmonics and diffusion tensor images for guidance. Their approach uses a 3D neural network that predicts the detailed signals and ensures accuracy by checking consistency in the resulting images. They tested their method on large brain scan datasets and found it maintains important brain structure measures better than other techniques.
What this means in practice
- •For clinical imaging teams: Generate multi-shell diffusion MRI signals from single-shell scans to speed up clinical workflows while preserving key brain microstructure details.
- •For medical image processing developers: Incorporate a volumetric spherical harmonics regression model conditioned on diffusion tensor imaging for accurate synthesis of missing diffusion MRI shells.
Authors
Binghua Li, Christina Andica, Tong Liang, Ziqing Chang, Chao Li, Wataru Uchida, Kaito Takabayashi, Qibin Zhao, Toshihisa Tanaka, Zhe Sun, Shigeki Aoki
Abstract
Multi-shell diffusion MRI (dMRI) unlocks more expressive microstructural modeling than single-shell scans, yet its longer acquisition time hinders deployment in large-scale cohorts and time-constrained clinical settings. Synthesizing an unobserved shell from a single-shell input is fundamentally ill-posed and further complicated by protocol mismatch, where source and target gradient direction sets may not align. We propose DTI-SHNet, a single-to-multi-shell synthesis framework that operates in the real symmetric spherical harmonics (SH) coefficient domain and performs spatially aware volumetric regression. Given a source shell, we estimate diffusion tensor imaging (DTI) and use direction-agnostic parametric maps along with a brain mask as conditioning priors to guide a 3D U-Net regressor from source-shell to target-shell SH coefficients. To couple coefficient accuracy with signal fidelity, we introduce a signal consistency regularization that reconstructs signals on randomly sampled canonical directions from predicted coefficients and enforces agreement in the signal domain. Experiments on UK Biobank and Cam-CAN data for b=1000 to b=2000 dMRI synthesis show that DTI-SHNet achieves competitive visual quality compared to advanced methods, while better preserving downstream diffusion measures. Our code is available at https://github.com/xiaovhua/dti-shnet.