Denoising Diffusion & Functional MRI for Neuroimaging

MICSI-RMT

High-end imaging on low-end scanners

High-end Imaging Across Magnetic Field Strengths

Our findings demonstrate that MICSI-RMT consistently delivers high-quality images across different magnetic field strengths, from 1.5T to 3.0T MRI systems. This capability enables the use of advanced diffusion imaging across MRI systems, significantly enhancing image quality without the need for expensive hardware upgrades.

Field Strength Comparison in MRI Scans of the Same Patient

These images showcase mean diffusivity (MD) and directionally encoded color fractional anisotropy (cFA) for the same patient scanned at different magnetic field strengths using MICSI-RMT technology. The top row features scans from a 1.5T MRI system, while the bottom row displays results from a 3.0T system.

The SNR values, calculated over white matter, illustrate how MICSI-RMT significantly improves image quality and detail across different magnetic field strengths, demonstrating its efficacy in providing high-quality diagnostic imaging without costly hardware upgrades.

1.5T
SNR = 10.9

1.5T MICSI
SNR = 53.6

3.0T
SNR = 14.3

3.0T MICSI
SNR = 43.6

ADC 1.5
ADC 1.5 MICSI
ADC 3
FA 3 MICSI

ADC

FA 1.5
FA 1.5 MICSI
FA 3
FA 3 MICSI

FA

1.5T
SNR = 10.9

1.5T MICSI
SNR = 53.6

ADC 1.5
ADC 1.5 MICSI

3.0T
SNR = 14.3

3.0T MICSI
SNR = 43.6

ADC 3
ADC 3 MICSI

ADC

1.5T
SNR = 10.9

1.5T MICSI
SNR = 53.6

FA 1.5
FA 1.5 MICSI

3.0T
SNR = 14.3

3.0T MICSI
SNR = 43.6

FA 3
FA 3 MICSI

FA

Diffusion MRI

Visualize Smaller Brain Structure

Enabling Anatomical Diffusion Parameter Maps[1, 7, 8]

Neuroimaging through dMRI traditionally grapples with imaging resolution against noise. Through a combination of denoising and Bayesian parameter estimation, it is possible to maximize the precision while minimizing bias to produce maps high fidelity quantitative maps.

Standard vs MICSI-dMRI Parameter Mapping

Parameter maps of mean diffusivity (MD), fractional anisotropy (FA), directionally encoded color FA (cFA), and radial kurtosis (RK) are compared between standard and MICSI-DMRI parameter estimation.

MP-PCA was used to preprocess both of these datasets, therefore the artifacts of the standard approach are unrelated to noise. The white arrows on standard cFA show regions within the cerebrospinal fluid (CSF), which incorrectly show fibrous structure, as though this fluid contains dense white matter.

The white arrows on standard cFA show regions within the cerebrospinal fluid (CSF).

MD

FA

cFA

RK

dMRI Param Mapping Original 1 Diagram
dMRI Param Mapping Original 2 Diagram
dMRI Param Mapping Original 3 Diagram
dMRI Param Mapping Original 4 Diagram

Standard dMRI[10]

dMRI Param Mapping RMT 1 Diagram
dMRI Param Mapping RMT 2 Diagram
dMRI Param Mapping RMT 3 Diagram
dMRI Param Mapping RMT 4 Diagram

Bayesian dMRI[8]

Standard dMRI[10]

Bayesian dMRI[8]

dMRI Param Mapping Original 1 Diagram
dMRI Param Mapping RMT 1 Diagram

MD

dMRI Param Mapping Original 2 Diagram
dMRI Param Mapping RMT 2 Diagram

FA

dMRI Param Mapping Original 3 Diagram
dMRI Param Mapping RMT 3 Diagram

cFA

dMRI Param Mapping Original 4 Diagram
dMRI Param Mapping RMT 4 Diagram

RK

1-mm Resolution Visualize Structures not at 2-mm Isotropic Resolution

1-mm Resolution Visualize Structures not at 2-mm Isotropic Resolution

A. Axial section of the brainstem section at the level of the trigeminal nerve, clearly visualized at high resolution.

2x2x2

Anatomy A2x

1x1x1

Anatomy A1x

B. Axial section of the thalamus, showing various tracts and thalamic nuclei.

Anatomy B2x Anatomy B1x

C. Axial section of the motor and somatosensory cortex, revealing green u-fibers connecting nearby cortical areas.

Anatomy C2x Anatomy C1x

D. Axial section of the cerebral peduncle and substantia nigra.

Anatomy D2x Anatomy D1x

E. Para-sagittal image of callosum – stria medullaris ventral to the fornix.

Anatomy E2x Anatomy E1x

F. Coronal image through anterior left medial temporal lobe – alveus & fimbria (arrow) / stratum lacunosum of hippocampus proper.

Anatomy F2x Anatomy F1x

Refining Fiber Tractography[1, 9]

MP-PCA increases the precision of fiber tracts derived from diffusion MRI compared to those of the original dataset.

Traco Side A1 Diagram

Original

Tracto Side B1 Diagram

MP-PCA

Functional MRI

Overcoming Technical Obstacles in Mapping Neural Activity

Functional MRI (fMRI) has been one of the greatest success stories of neuroimaging and has transformed our understanding of the brain by allowing us to visualize and measure brain activity. However, its application in medical imaging faces challenges related to:

  • Spatial Resolution
  • Scan time
  • Limitations to the the variety of task paradigms within a single session


As an enabling technology for clinical functional MRI, MP-PCA is capable of boosting the signal-to-noise ratio (SNR) commensurate with the number of temporal frames in the fMRI time-series. The extra SNR provided by MP-PCA can be strategically utilized to mitigate issues stated above.

Task-Based FMRI

We demonstrate the feasibility of measuring cortical activation from a sentence completion task using MP-PCA. Areas of activation are shown for Z>3, no spatial filtering was applied to calculate activation z-scores.[2, 4, 5]

Original

MP-PCA

Neural Activity Original Diagram
Neural Activity RMT Diagram

Related Publications

Initial publication describing the method:

Denoising of Diffusion MRI Using Random Matrix Theory

Denoising functional MRI for pre-operative planning:

Improved Task-based Functional MRI Language Mapping in Patients with Brain Tumors through Marchenko-Pastur Principal Component Analysis Denoising

Denoising neuroimaging diffusion MRI:

Evaluation of the accuracy and precision of the diffusion parameter EStImation with Gibbs and NoisE removal pipeline

Advanced diffusion metrics in neuroimaging:

Quantifying Brain Microstructure with Diffusion MRI: Theory and Parameter Estimation

Improved Fiber Orientations via MPPCA confirmed on simulation and real data:

Brain Fiber Structure Estimation Based on Principal Component Analysis and RINLM Filter

Gold Standard weights for estimation of the diffusion cumulant expansion:

Weighted Linear Least Squares Estimation of Diffusion MRI Parameters: Strengths, Limitations, and Pitfalls

Bayesian approach to fitting diffusion MRI data:

Disentangling Micro from Mesostructure by Diffusion MRI: A Bayesian Approach

Nature Communications Paper applying MP-PCA onto normalized noise levels for high-resolution functional MRI:

Lowering the Thermal Noise Barrier in Functional Brain Mapping with Magnetic Resonance Imaging

Interested in learning more?

Schedule a demo or contact us for more information

Contact Us