Classification algorithms using multiple MRI features in mild traumatic brain injury.

Area of Science:

• Neuroimaging
• Radiology
• Medical Diagnostics

Background:

• Mild traumatic brain injury (mTBI) presents diagnostic challenges.
• Objective classification of mTBI is crucial for effective patient management.
• Current diagnostic methods may not fully capture the complexities of mTBI.

Purpose of the Study:

• To develop and validate a classification algorithm for mTBI using Magnetic Resonance Imaging (MRI) metrics.
• To identify optimal MRI features for distinguishing mTBI patients from healthy controls.

Main Methods:

• Prospective study involving mTBI patients and healthy controls.
• Acquisition of multimodal MRI data including diffusion-weighted imaging, resting-state fMRI, and magnetic field correlation (MFC).
• Application of machine learning classifiers with feature selection (minimal-redundancy maximal-relevance) on various MRI metrics.

Main Results:

• Single-feature classification using thalamic mean kurtosis (MK) achieved 74% accuracy.
• Multifeature analysis yielded up to 86% accuracy.
• Key features included thalamic MK, anterior cingulate volume, thalamic thickness, resting-state networks, and MFC.

Conclusions:

• Multifeature analysis integrating diffusion-weighted imaging, MFC, fMRI, and volumetrics shows promise for mTBI classification.
• Optimal feature selection and classification methods enhance diagnostic accuracy.
• This approach provides Class III evidence for accurate mTBI identification.