Marc LaFonte, MD, FACS
Assistant Professor of Surgery
Robert Wood Johnson University Hospital, United States
Disclosure information not submitted.
Ilker Hacihaliloglu, PhD
Assistant Professor, department of Biomedical Engineering
Rutgers University, United States
Disclosure information not submitted.
.jpg "Matthew Lissauer, MD, FCCM photo")
Matthew Lissauer, MD, FCCM
Acute Care Surgeon
Robert Wood Johnson University Hospital
Hartford, Connecticut
Disclosure information not submitted.
Title: Automated Image Processing with Point-of-care Ocular Ultrasound for Real-time ICP Monitoring
Objectives:
To determine whether computational image analysis is superior to manual ultrasound measurement of optic sheath nerve diameter (OSND) in brain injury patients with invasive ICP monitoring.
DESIGN, PATIENTS, SETTING:
This prospective observational study included 61 images obtained from 16 patients in the surgical intensive care unit (SICU) at a surgical intensive care unit in a level 1 tertiary care hospital who required invasive intracranial monitoring in the setting of traumatic or non-traumatic brain injury.
INTERVENTIONS:
Utilization of advanced computational image analysis to refine OSND after manual acquisition using ultrasound.
Results:
A total of 61 optic nerve sheath images were obtained from 16 patients. We compared automated measurements to manually obtained ones by assessing the gross measurement differences in millimeters, the percent difference, mean absolute error, and statistical significance between clinically intervenable ICP range groups. Algorithm-based ONSD measurements were on average 2.75 mm larger, or a 7.37% difference, than those measured manually. Algorithmically derived ONSD measurements also demonstrated a statistically significant difference between a cohort of patients with ICP of 16-20 mmHg versus those with an ICP of 21-25 mmHg groups (95% CI, 56.2 to 67.2, P< 0.05). Algorithmic measurements were also able to discriminate between patients with ICP in the 11-15 mmHg from 16-20 mmHg (95% CI, 53.0 to 62.7, P< 0.05), which manual measurements were not able to accomplish.
Conclusions:
Automated algorithmic image processing may be able to help eliminate operator error, creating a more robust method for non-invasive ICP monitoring. In conjunction with ocular ultrasound for real-time intracranial pressure monitoring, there’s tremendous potential for obviating the risks associated with invasive ICP monitoring, as well as providing a reliable clinical tool in austere environments, such as emergency departments, ambulatory settings, or battlefield hospitals.