Machine Learning and Multiparametric Brain MRI to Differentiate Hereditary Diffuse Leukodystrophy with Spheroids from...

Area of Science:

• Neuroimaging
• Machine Learning
• Differential Diagnosis

Background:

• Hereditary diffuse leukoencephalopathy with spheroids (HDLS) and multiple sclerosis (MS) are challenging to differentiate due to overlapping symptoms.
• Misdiagnosis of HDLS as MS can lead to inappropriate treatment and management.

Purpose of the Study:

• To develop and validate a machine learning method for robust classification of HDLS versus MS using quantitative MRI.
• To improve the diagnostic accuracy for rare neurological disorders like HDLS.

Main Methods:

• A quantitative brain MRI protocol (synthetic MRI) was applied to 4 HDLS and 14 MS patients at 3 Tesla.
• Machine learning models were trained using quantitative MRI features.
• Repeatability analysis was performed on healthy controls across different field strengths (3T and 1.5T).

Main Results:

• The machine learning model achieved 100% correct classification in cross-validation using 5-11 quantitative MRI features.
• High true positive rates were observed even with added measurement noise (97.2% for HDLS, 99.6% for MS).
• Predicting features showed high measurement confidence (1.7% at 3T, 2.3% at 1.5T).

Conclusions:

• Quantitative MRI combined with machine learning shows promise for computer-assisted differential diagnosis of HDLS and MS.
• This approach can aid clinicians in identifying patients with a high probability of HDLS, guiding further genetic testing.