A prospective study of transorbital ultrasound multimodal imaging in the detection and evaluation of increased intracranial pressure

Peng He, Lanlan Yu, Xingheng Wang, Bo Ren, Peng Gu, Fei Yang, Yuan ZouBMC Medical Imaging 2025 · DOI

TL;DR

Monitoring increased intracranial pressure (IICP) traditionally requires invasive, risky procedures like lumbar punctures or drilling into the skull. This paper introduces a highly accurate, non-invasive alternative using a multimodal ultrasound model that combines structural measurements of the optic nerve with blood flow metrics from the eye. By integrating these parameters, the researchers achieved a 95% diagnostic accuracy, offering a safer, rapid screening tool for critical neurology patients.

Introduction

Increased intracranial pressure (IICP) is a life-threatening neurological emergency that can result from brain tumors, hemorrhages, or infections. If left unchecked, it can lead to brain herniation and death. The gold standard for diagnosing IICP is invasive monitoring—such as a lumbar puncture or an intraventricular catheter—but these procedures carry significant risks of bleeding and infection and require specialized medical setups.

To avoid these risks, clinicians have increasingly turned to non-invasive transorbital ultrasound, specifically measuring the swelling of the Optic Nerve Sheath Diameter (ONSD). Because the optic nerve sheath is directly connected to the brain's subarachnoid space, it expands when intracranial pressure rises. However, using ONSD alone is often unreliable, with sensitivity fluctuating between 50% and 90% due to natural anatomical differences between patients. This study addresses this reliability gap by combining structural measurements with real-time vascular (blood flow) data to create a robust, multimodal predictive model.

Methods

The researchers conducted a prospective diagnostic study on 136 neurology patients scheduled for lumbar punctures. Just before the invasive procedure, patients underwent a non-invasive transorbital ultrasound applied over closed eyelids. The ultrasound operators were blinded to the patients' clinical status to ensure objective data collection.

Flowchart of patient screening and enrollment for the prospective study.

The team measured two types of parameters:

  1. Anatomical/Structural: The raw Optic Nerve Sheath Diameter (ONSD) and the ONSD-to-eyeball transverse diameter ratio (ONSD/ETD). The ratio accounts for individual anatomical variations in eye size.
  2. Hemodynamic (Blood Flow): Using Doppler ultrasound, they measured the Central Retinal Artery (CRA) located within the optic nerve sheath, recording Peak Systolic Velocity (PSV) and the Resistance Index (RI).
Comparison of invasive lumbar puncture and non-invasive transorbital ultrasound for monitoring intracranial pressure.

Immediately after the ultrasound, patients received their scheduled lumbar puncture to establish the "ground truth" intracranial pressure (where >200 mmH₂O was classified as IICP). Finally, the team used multivariate logistic regression to fuse these ultrasound parameters into a single predictive mathematical model.

Key Insights and Ideas

  • Pressure Chokes Blood Flow: The study demonstrated that IICP doesn't just stretch tissue; it physically compresses the blood vessels feeding the eye. As brain pressure rises, blood flow velocity in the central retinal artery drops (lower PSV), and vascular resistance spikes (higher RI).
  • Normalization is Crucial: Using the ONSD/ETD ratio proved vastly superior to using raw ONSD measurements. By dividing the nerve sheath diameter by the width of the patient's eyeball, the model successfully corrected for individual anatomical variations, drastically reducing false positives.
  • Synergy Over Isolation: Structural changes (sheath swelling) and hemodynamic changes (blood flow resistance) provide complementary, independent information. Combining them yields a holistic view of the eye's reaction to brain pressure, resulting in a classifier significantly more accurate than its individual components.

Key Results

  • Superior Combined Accuracy: The integrated multimodal model (ONSD + ONSD/ETD + PSV + RI) achieved an outstanding Area Under the Curve (AUC) of 0.95, with 88.1% sensitivity and 98.0% specificity.
  • Poor Single-Metric Performance: Relying on single parameters was highly flawed. For example, raw ONSD alone had an AUC of only 0.81 (with 59.7% sensitivity), and blood flow velocity (PSV) alone had an AUC of 0.67.
  • Clear Hemodynamic Shifts: Patients with IICP showed statistically significant drops in blood flow velocity (9.45 cm/s vs 10.86 cm/s in normal patients) and higher resistance indices (0.66 vs 0.56).
  • Deployable Equation: The study yielded a specific logistic regression formula (Logit(P) = 1.94×ONSD – 84.6×ONSD/ETD + 0.79×PSV – 3.8×RI + 11.05) that could theoretically be programmed directly into ultrasound machine software for automated, real-time risk scoring.

Limitations and Future Work

The authors acknowledge several limitations. First, the sample size is relatively small (136 patients) and originates from a single hospital, meaning the model requires validation across larger, multi-center, and multi-ethnic cohorts. Second, the study did not stratify results based on the specific cause of the pressure (e.g., tumor vs. infection), which could subtly alter ultrasound readings.

Future research may incorporate additional optical assessments, such as optic disc elevation (papilledema) and Optical Coherence Tomography (OCT), to further enhance diagnostic efficacy.

Summary updated

BibTeX
@article{he2025prospective,
  author = {Peng He and Lanlan Yu and Xingheng Wang and Bo Ren and Peng Gu and Fei Yang and Yuan Zou},
  title = {{A prospective study of transorbital ultrasound multimodal imaging in the detection and evaluation of increased intracranial pressure}},
  journal = {BMC Medical Imaging},
  year = {2025},
  doi = {10.1186/s12880-025-01893-w},
}
Figure