Advantage of grading classification using volumetric artificial intelligence for periventricular hyperintensity and...

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Periventricular hyperintensity (PVH) and deep subcortical white matter hyperintensity (DWMH) are markers of small vessel disease.
• Accurate grading of these lesions is crucial for diagnosis and monitoring.
• Current grading relies on subjective expert assessment of magnetic resonance imaging (MRI).

Purpose of the Study:

• To develop and validate an artificial intelligence (AI) algorithm for automated grading of PVH and DWMH.
• To compare the AI's performance against human expert assessments using objective metrics.
• To evaluate the AI's consistency and reliability in lesion classification.

Main Methods:

• An AI algorithm was trained and tested on MRI scans from 246 patients.
• AI-predicted Fazekas scale grades were compared to expert assessments using accuracy, F1-score, and mean absolute error.
• Inter-rater agreement was assessed using Fleiss' kappa and Cohen's kappa.

Main Results:

• The AI achieved superior multi-class accuracy for PVH classification (0.798) compared to human experts (0.743).
• For DWMH classification, the AI outperformed experts in distinguishing severe from mild/moderate lesions (0.954 vs. 0.927).
• The AI demonstrated good agreement with human raters for both PVH and DWMH, exceeding human inter-rater agreement for PVH.

Conclusions:

• The developed AI algorithm effectively grades PVH and DWMH with accuracy comparable to human performance.
• The AI shows potential for objective, consistent, and reliable automated analysis of white matter hyperintensities.
• This AI tool could aid in the diagnosis and management of cerebrovascular diseases.