Validation of DWI pre-processing procedures for reliable differentiation between human brain gliomas.

Area of Science:

• Neuroimaging
• Medical Physics
• Radiology

Background:

• Diffusion magnetic resonance imaging (dMRI) is crucial for brain tumor detection and characterization in oncology.
• Clinical dMRI data are prone to artifacts (motion, noise, distortions) that compromise image quality and analysis accuracy.
• Effective pre-processing is vital for reliable diffusion scalar metrics and subsequent data interpretation.

Purpose of the Study:

• To evaluate the impact of various dMRI pre-processing techniques on the accuracy of brain glioma differentiation.
• To identify optimal pre-processing strategies for enhancing sensitivity and specificity in glioma malignancy grading.

Main Methods:

• Assessed effects of noise correction, smoothing algorithms (anisotropic diffusion filtering), and spatial interpolation (cubic-order spline) on raw dMRI data.
• Utilized diffusion tensor imaging (DTI), diffusion kurtosis imaging (DKI), and neurite orientation dispersion and density imaging (NODDI) biophysical models.
• Quantified accuracy using derived scalar metrics for glioma malignancy grading.

Main Results:

• Noise correction, anisotropic diffusion filtering, and cubic-order spline interpolation significantly improved glioma malignancy grading.
• These selected pre-processing steps demonstrated the highest sensitivity and specificity in differentiating glioma grades.
• The chosen biomarkers (DTI, DKI, NODDI metrics) proved sensitive to pre-processing effects.

Conclusions:

• Specific pre-processing steps, including noise correction, anisotropic diffusion filtering, and cubic-order spline interpolation, are recommended for dMRI analysis in brain tumor studies.
• Optimized pre-processing enhances the accuracy of glioma malignancy grading using advanced diffusion imaging models.
• These findings support the clinical utility of dMRI in oncology by improving diagnostic reliability.