Identifying Patients with CSF-Venous Fistula Using Brain MRI: A Deep Learning Approach.

Area of Science:

• Neurology
• Radiology
• Artificial Intelligence in Medicine

Background:

• Spontaneous intracranial hypotension (SIH) is increasingly recognized.
• It is caused by cerebrospinal fluid (CSF) leaks, often linked to CSF-venous fistulas.
• Brain MRI in SIH patients may show dural enhancement, brain sag, and pituitary engorgement.

Purpose of the Study:

• To develop a deep learning model for accurate diagnosis of CSF-venous fistulas.
• To utilize brain Magnetic Resonance Imaging (MRI) for identifying CSF leaks.
• To improve diagnostic accuracy in patients with suspected SIH.

Main Methods:

• A deep learning model was trained on brain MRI scans from 129 patients.
• Patients were categorized based on digital subtraction myelogram findings (definite fistula, no fistula, indeterminate).
• A 5-fold cross-validation was employed to assess model reliability and predictive value (Area Under the Receiver Operating Characteristic Curve).

Main Results:

• The study included 129 patients (median age 54 years; 51.2% with CSF-venous fistula).
• The deep learning model achieved an average Area Under the Receiver Operating Characteristic Curve of 0.8668 (±0.0254).
• The model demonstrated strong performance in discriminating between positive and negative cases for CSF-venous fistulas.

Conclusions:

• A deep learning model was developed to predict spinal CSF-venous fistulas from brain MRI in SIH patients.
• Further refinement and external validation are needed for clinical adoption.
• Deep learning shows significant potential for diagnosing CSF-venous fistulas using brain MRI.