G. Herrnstadt, G. Oland, I.O. Emeruwa, N. Parikh, P. Vespa
CranioView,
United States
Keywords: ocular ultrasound, intracranial pressure, medical device, traumatic brain injury, non-invasive diagnostics, image processing, machine learning
Summary:
Elevated intracranial pressure (ICP) can be a life-threatening consequence for the 3.6 million patients that suffer from traumatic brain injury, cardiac arrest, and hemorrhagic stroke annually in the U.S. Invasive ICP monitoring is the standard of care for management of elevated ICP and prevention of secondary brain injury. While highly accurate, invasive ICP monitoring requires drilling a hole through the skull to insert a catheter through brain tissue and is associated with risks of life-threatening infections and bleeding, and often still requires surgical revision, and can increase the cost per patient by >100%. Given these risks, invasive monitoring is only performed for an estimated 3% of the 900,000 patients who are considered at-risk for elevated ICP after their brain injury. There is an urgent unmet need for an innovative, non-invasive, and accurate brain pressure monitoring solution to significantly increase access to ICP monitoring, reduce complications, and improve outcomes. Ocular ultrasound is an emerging non-invasive approach where the optic nerve sheath diameter (ONSD) can be measured to give a binary assessment of whether ICP is elevated or normal. While ONSD has shown good clinical performance for detection of elevated ICP, its specific clinical application still varies widely between providers and it cannot estimate absolute, quantitative ICP. CranioView improves on this technology through commercialization of a novel, lightweight, wearable device that automates clinical processes and accurately estimates brain pressure. The CranioView wearable improves image stability over prolonged insonation, and when coupled with CranioView’s image processing approach, enables accurate detection of novel dynamic signals in the ultrasound videos for improved correlation with ICP. CranioView is advancing machine learning (ML) methods for autosegmentation and handsfree extraction of variables from anatomic structures surrounding the optic nerve (i.e., ONSD, eyeball transverse diameter, and optic disc elevation), and integrating them with CranioView’s novel dynamic signals to create a multivariate model for improving non-invasive ICP estimation at the bedside. CranioView has demonstrated the wearable device and ML autosegmentation approach through a pilot study of 12 healthy subjects and ~1,000 images, achieving ONSD accuracy within 97% of expert measurement. CranioView also demonstrated motion artifacts reduction with the wearable device and accurate identification of dynamic brain tissue motions. Commercialization of CranioView’s ICP monitor would expand monitoring to mild- and moderate-risk patient populations without the complications associated with invasive monitoring, as well as reduce the time to intervention in severe-risk patients in whom delays in treatment result in worsening of brain injury and even death.