The invention relates to the technical field of
intracranial pressure monitoring, in particular to a multi-
modal fusion-based
intracranial pressure dynamic prediction method for a patient with
craniocerebral injury, which comprises the following steps of: acquiring an electroencephalogram
signal, extracting a
peak value to a termination section path, judging whether a waveform slope is consistent with an
intracranial pressure direction or not, identifying an unsynchronized response segment, and obtaining a neural migration section mark set. According to the method, the non-cooperative state between the
neural activity and the pressure change is distinguished by recognizing the periodic waveform which does not cause the intracranial
pressure response in the electroencephalogram
signal, and the time period with the inconsistent regulation
rhythm is positioned by combining the continuous expression of the
ventricular volume and the arterial pressure direction deviation; the method comprises the following steps of: extracting response sequence and direction characteristics of arterial pressure,
cerebral blood flow and neural signals in a continuous section, enhancing a
rhythm corresponding relation among multi-
modal signals, adjusting a
signal alignment structure according to offset direction difference and a
rhythm starting point, forming a linkage fragment sequence under a unified time reference, and obtaining a multi-
modal signal sequence; and linkage and deduction of asynchronous response information on a
structural level are promoted.