UCLA researchers in the Department of Neurosurgery have developed a novel framework to constrain noises in the measurements of vital physiological signals from neurosurgical patients.
Cerebral blood flow velocity (CBFV) and intracranial pressure (ICP) remain the pivotal physiological signals for managing neurosurgical patients in both acute states, such as brain injury, and in chronic states such as hydrocephalus. Despite this importance, signal processing capabilities in existing commercial CBFV and ICP monitoring devices remain poor, providing clinicians with a limited amount of information. One of the major challenges in accurate continuous monitoring of these signals is the interference from various types of noises. It has been known that ICP is intrinsically related with CBFV, which might provide some important information for the denoising problem. However, this complex relationship has only been partially explored, and not been applied to constraining the noises.
UCLA researchers have learned that the changes in the CBFV and ICP waveforms exhibit temporal trends. They have developed a framework, based on machine learning, which learns the temporal relationship between CBFV and ICP, and then used the data to denoise/reconstruct the signal. This framework significantly reduces the amount of noise and artifacts presented in the signal.
cerebral blood flow, intracranial pressure, machine learning, traumatic brain injury, droppler ultrasound blood flow