Support Vector Machine-based Spontaneous Intracranial Hypotension Detection on Brain MRI.

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Spontaneous intracranial hypotension (SIH) is a condition often diagnosed through invasive methods.
• Accurate and non-invasive identification of SIH is crucial for timely patient management.

Purpose of the Study:

• To develop a fully automatic algorithm for identifying spontaneous intracranial hypotension (SIH) using magnetic resonance imaging (MRI).
• To improve the diagnostic accuracy and efficiency of SIH detection.

Main Methods:

• A support vector machine (SVM) was trained using structured reports from 140 patients with suspected SIH.
• A convolutional neural network (CNN) segmented venous sinuses and basal cisterns on contrast-enhanced T1-weighted MPRAGE sequences.
• 56 radiomic features from segmented regions were extracted and used as input for the SVM.

Main Results:

• The algorithm achieved a high diagnostic accuracy, with an area under the curve (AUC) of 0.91.
• Venous sinuses and suprasellar cistern features showed the strongest discriminative power between SIH and non-SIH patients.
• The combined SVM score effectively differentiated SIH (mean score 1.41) from non-SIH patients (mean score 0.30, p < 0.001).

Conclusions:

• A fully automatic algorithm analyzing a single MRI sequence can accurately differentiate SIH from non-SIH patients.
• This AI-driven approach may assist in determining the necessity of invasive diagnostics and patient transfer to specialized centers.