MICCAI-CDMRI 2023 QuantConn Challenge Findings on Achieving Robust Quantitative Connectivity through Harmonized...

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Computational Neuroscience

Background:

• White matter alterations are key indicators in neurological disease progression.
• Diffusion-weighted magnetic resonance imaging (DW-MRI) is vital for studying white matter microstructure and connectivity.
• Inconsistent DW-MRI acquisition protocols hinder quantitative analysis and cross-study comparisons.

Purpose of the Study:

• To harmonize DW-MRI preprocessing to ensure robust quantitative diffusion metrics across different acquisitions.
• To evaluate methods for minimizing acquisition-specific differences while preserving biological variation in DW-MRI data.
• To assess the impact of harmonization on bundle-wise microstructure, shape features, and connectomics.

Main Methods:

• The QuantConn challenge provided raw DW-MRI data from the same individuals with two acquisition protocols.
• Participants were tasked with preprocessing DW-MRI data to harmonize cross-acquisition differences.
• Harmonized data were evaluated based on the reproducibility and comparability of microstructure, shape, and connectomics measures.

Main Results:

• Measures like bundle surface area and fractional anisotropy were most susceptible to acquisition bias.
• Machine learning voxel-wise correction, RISH mapping, and NeSH methods effectively reduced these biases.
• Microstructure measures (AD, MD, RD), bundle length, and connectome metrics were least biased by acquisition differences.

Conclusions:

• A machine learning approach was most effective for harmonizing connectomic, microstructure, and macrostructure features, but requires co-registered data.
• The NeSH resampling method also proved effective and offers a generalizable framework independent of co-registration.
• These harmonization techniques are essential for reliable large-scale DW-MRI studies in neurological diseases.