DT-MRI fiber tracking: a shortest paths approach

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biomedical Engineering

Background:

• Conventional Diffusion Tensor Magnetic Resonance Imaging (DT-MRI) fiber tracking algorithms often fail due to their "greedy" nature, leading to accumulated local errors.
• Reconstructing complex or low signal-to-noise ratio (SNR) fiber trajectories remains a significant challenge in the field.

Purpose of the Study:

• To develop a novel fiber tracking algorithm for DT-MRI that overcomes the limitations of conventional "greedy" approaches.
• To demonstrate the algorithm's capability in precisely reconstructing diverse fiber trajectories, including those where conventional methods fail.

Main Methods:

• The new algorithm frames fiber tracking as a shortest path problem on a weighted directed graph, where voxels are vertices and edges connect neighboring voxels.
• Edge weights are probabilities derived from a Bayesian framework, favoring edges aligned with local fiber orientations.
• Computationally scalable shortest path algorithms are employed to find optimal paths of maximum probability.

Main Results:

• The algorithm achieves global optimality, preventing error accumulation and enabling accurate reconstruction even with low signal-to-noise ratios.
• It demonstrates precise reconstruction of complex fiber trajectories in both authentic and synthetic DT-MRI data where conventional algorithms falter.
• The method shows impartiality to seed point selection and offers faster computation times compared to traditional all-paths tracking.

Conclusions:

• This novel shortest path-based algorithm provides a robust and accurate solution for DT-MRI fiber tracking, surpassing conventional methods.
• Its global optimality and ability to handle low SNR data open new possibilities for detailed brain connectomics and neurological disorder research.