Axisymmetric diffusion kurtosis imaging with Rician bias correction: A simulation study

Investigating apparent differences between standard DKI and axisymmetric DKI and its consequences for biophysical parameter estimates

PURPOSE

The purpose of the study is to identify differences between axisymmetric diffusion kurtosis imaging (DKI) and standard DKI, their consequences for biophysical parameter estimates, and the protocol choice influence on parameter estimation.

METHODS

Noise-free and noisy, synthetic diffusion MRI human brain data is simulated using standard DKI for a standard and the fast "199" acquisition protocol. First the noise-free "baseline" difference between both DKI models is estimated and the influence of fiber complexity is investigated. Noisy data is used to establish the signal-to-noise ratio at which the baseline difference exceeds noise variability. The influence of protocol choices and denoising is investigated. The five axisymmetric DKI tensor metrics (AxTM), the parallel and perpendicular diffusivity and kurtosis and mean of the kurtosis tensor are used to compare both DKI models. Additionally, the baseline difference is also estimated for the five parameters of the WMTI-Watson model.

RESULTS

The parallel and perpendicular kurtosis and all of the WMTI-Watson parameters had large baseline differences. Using a Westin or FA mask reduced the number of voxels with large baseline difference, that is, by selecting voxels with less complex fibers. For the noisy data, precision was worsened by the fast "199" protocol but adaptive denoising can help counteract these effects.

CONCLUSION

For the diffusivities and mean of the kurtosis tensor, axisymmetric DKI with a standard protocol delivers similar results as standard DKI. Fiber complexity is one main driver of the baseline differences. Using the "199" protocol worsens precision in noisy data but adaptive denoising mitigates these effects.

Insights and improvements in correspondence between axonal volume fraction measured with diffusion-weighted MRI and electron microscopy

Biophysical diffusion-weighted imaging (DWI) models are increasingly used in neuroscience to estimate the axonal water fraction (f AW ), which in turn is key for noninvasive estimation of the axonal volume fraction (f A ). These models require thorough validation by comparison with a reference method, for example, electron microscopy (EM). While EM studies often neglect the unmyelinated axons and solely report the fraction of myelinated axons, in DWI both myelinated and unmyelinated axons contribute to the DWI signal. However, DWI models often include simplifications, for example, the neglect of differences in the compartmental relaxation times or fixed diffusivities, which in turn might affect the estimation off AW . We investigate whether linear calibration parameters (scaling and offset) can improve the comparability between EM- and DWI-based metrics off A . To this end, we (a) used six DWI models based on the so-called standard model of white matter (WM), including two models with fixed compartmental diffusivities (e.g., neurite orientation dispersion and density imaging, NODDI) and four models that fitted the compartmental diffusivities (e.g., white matter tract integrity, WMTI), and (b) used a multimodal data set including ex vivo diffusion DWI and EM data in mice with a broad dynamic range of fibre volume metrics. We demonstrated that the offset is associated with the volume fraction of unmyelinated axons and the scaling factor is associated with different compartmentalT 2 and can substantially enhance the comparability between EM- and DWI-based metrics off A . We found that DWI models that fitted compartmental diffusivities provided the most accurate estimates of the EM-basedf A . Finally, we introduced a more efficient hybrid calibration approach, where only the offset is estimated but the scaling is fixed to a theoretically predicted value. Using this approach, a similar one-to-one correspondence to EM was achieved for WMTI. The method presented can pave the way for use of validated DWI-based models in clinical research and neuroscience.

Validation of deep learning techniques for quality augmentation in diffusion MRI for clinical studies

The objective of this study is to evaluate the efficacy of deep learning (DL) techniques in improving the quality of diffusion MRI (dMRI) data in clinical applications. The study aims to determine whether the use of artificial intelligence (AI) methods in medical images may result in the loss of critical clinical information and/or the appearance of false information. To assess this, the focus was on the angular resolution of dMRI and a clinical trial was conducted on migraine, specifically between episodic and chronic migraine patients. The number of gradient directions had an impact on white matter analysis results, with statistically significant differences between groups being drastically reduced when using 21 gradient directions instead of the original 61. Fourteen teams from different institutions were tasked to use DL to enhance three diffusion metrics (FA, AD and MD) calculated from data acquired with 21 gradient directions and a b-value of 1000 s/mm2. The goal was to produce results that were comparable to those calculated from 61 gradient directions. The results were evaluated using both standard image quality metrics and Tract-Based Spatial Statistics (TBSS) to compare episodic and chronic migraine patients. The study results suggest that while most DL techniques improved the ability to detect statistical differences between groups, they also led to an increase in false positive. The results showed that there was a constant growth rate of false positives linearly proportional to the new true positives, which highlights the risk of generalization of AI-based tasks when assessing diverse clinical cohorts and training using data from a single group. The methods also showed divergent performance when replicating the original distribution of the data and some exhibited significant bias. In conclusion, extreme caution should be exercised when using AI methods for harmonization or synthesis in clinical studies when processing heterogeneous data in clinical studies, as important information may be altered, even when global metrics such as structural similarity or peak signal-to-noise ratio appear to suggest otherwise.