Reduced NEXI protocol for the quantification of human gray matter microstructure on the Connectome 2.0 scanner

2026-07-08 · Source: cs.CV updates on arXiv.org · Field: Science & Research — Health & Medical Research, Mathematics & Computational Sciences, Artificial Intelligence & Machine Learning · Depth: Expert, extended

Summary

This study presents a reduced 8-feature diffusion MRI protocol for Neurite Exchange Imaging (NEXI) on the Connectome 2.0 scanner, significantly shortening scan duration from 27 to 14 minutes. Developed using an explainable AI framework (XGBoost-SHAP-RFE), the protocol identifies an optimal subset from a 15-feature scheme. Validation in vivo demonstrated that this optimized protocol preserves NEXI parameter accuracy, anatomical contrast, and test-retest reproducibility, yielding estimates and cortical maps comparable to the full acquisition. It notably reduced deviation in water exchange time (t_ex) estimates by over two-fold compared to theory-driven and heuristic reduction schemes. This advancement supports broader application of exchange-sensitive diffusion MRI in neuroscience and clinical research.

Key takeaway

For research scientists designing advanced diffusion MRI protocols, this XAI-driven optimization framework offers a robust method to significantly reduce scan times. You can achieve comparable parameter accuracy and reproducibility to full protocols, even for challenging gray matter microstructure. Consider adapting this XGBoost-SHAP-RFE pipeline to your specific scanner and biophysical models to accelerate data acquisition and enhance clinical translation.

Key insights

Explainable AI (XAI) can optimize complex MRI protocols, reducing scan time without losing data fidelity.

Principles

• Data-driven optimization outperforms purely theoretical methods.
• XAI reveals crucial features for robust biophysical model estimation.
• Sampling extremes of acquisition parameter space is key.

Method

The XGBoost-SHAP-RFE pipeline trains regressors on synthetic signals, quantifies feature relevance via SHAP values, and iteratively removes least informative features to identify an optimal subset.

In practice

• Apply XGBoost-SHAP-RFE for dMRI protocol design.
• Retrain XAI for different scanner hardware or biophysical models.
• Prioritize features at b-value and diffusion time extremes.

Topics

• Diffusion MRI
• Neurite Exchange Imaging
• Protocol Optimization
• Explainable AI
• Connectome 2.0 Scanner
• Gray Matter Microstructure

Code references

• Mic-map/graymatter_swissknife
• QuentinUhl/XAI-dMRI-protocol-optimization

Best for: AI Scientist, Research Scientist

Related on AIssential

• A physics-informed foundation model for quantitative diffusion MRI — PIGMENT uses a physics-informed generative prior to enable reliable quantitative diffusion MRI from sparse, heterogeneous data.
• Decision processes in 3D structural MRI schizophrenia classification evaluated with saliency maps — Explainable AI methods are crucial for validating deep learning models in medical diagnostics, revealing decision plausibility beyond performance metrics.
• AI algorithm enables tracking of vital white matter pathways — An AI tool precisely segments brainstem white matter bundles in live MRI, revealing disease-specific structural changes and recovery.
• Sensory Restoration via Brain-Computer Interfaces: A Unified 2 x 2 Framework and Convergence Roadmap — The article proposes a 2x2 framework and roadmap to unify fragmented BCI research for sensory restoration.
• Anatomy-Guided Residual Motion Diffusion for Controllable 4D Cardiac MRI Synthesis — A cascaded latent diffusion model synthesizes controllable, anatomically consistent 4D cardiac MRI, improving AI model robustness with limited data.
• PhyMRI-SR: Toward Physics-Aware MRI Image Super-Resolution — MRI super-resolution benefits from physics-aware reconstruction that optimizes resolution-SNR trade-offs, not just upscaling fixed inputs.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CV updates on arXiv.org.