CSF Flow Waveform Morphology: AI Biomarkers from Automated Analysis of Aqueduct Cine PC-MRI for Spontaneous Intracranial Hypotension

Background: /Objectives: Spontaneous intracranial hypotension (SIH) is caused by spinal cerebrospinal fluid (CSF) leakage and is typically diagnosed by clinical presentation and characteristic MRI signs; however, objective tools for monitoring physiological alterations and treatment response remain limited. Cine phase-contrast MRI (PC-MRI) enables noninvasive quantification of aqueductal CSF dynamics, yet reliable analysis is challenging since the cerebral aqueduct is extremely small and susceptible to low contrast, partial volume effects, and ROI-dependent measurement variability—particularly in SIH where CSF pulsatility is often reduced. Methods: We propose an end-to-end automated framework that integrates (1) a cascade localization–segmentation strategy (Tiny YOLOv4 detection followed by MultiResUNet segmentation on a cropped ROI form YOLOv4 result; (2) physiology-informed pulsatility-based segmentation (PUBS) to refine anatomical masks into functional flow ROIs, and (3) one-dimensional convolutional neural networks (1D-CNNs) to learn waveform morphology biomarkers from 32-phase cardiac-cycle velocity waveforms. The study includes 39 participants (11 controls; 28 pre-treatment SIH; 20 post-treatment recovery). Results: The cascade model significantly improves segmentation robustness compared with a full-image baseline, achieving higher Dice scores and markedly lower boundary errors across cohorts (e.g., pre-treatment SIH HD95: 1.66 ± 0.74 px vs. 15.37 ± 44.98 px). PUBS refinement reduced quantification deviation from expert manual references in SIH (mean relative error: 7.4% to 5.6%) and improved diagnostic performance for multiple hemodynamic parameters (e.g., downward mean flow AUC: 0.747 to 0.792). Importantly, waveform morphology learning substantially outperformed conventional scalar metrics: peak systolic CSF velocity (PSV) showed limited discrimination (AUC 0.69), whereas the full morphology AI model achieved AUC 0.96. A single trait-like morphology feature reached AUC 0.98 with 100% specificity, while a state-dependent feature normalized after recovery, indicating complementary utility for diagnosis and longitudinal monitoring. Conclusions: This fully automated, reproducible, and physiologically informed pipeline demonstrates that SIH-related information is not only reflected in flow magnitude but also encoded in subtle aqueductal CSF waveform shape patterns, supporting morphology-based biomarkers as a promising tool for SIH assessment and follow-up.