CN-121129314-B - PD motion subtype classification method based on transcranial Doppler and Bayesian analysis

CN121129314BCN 121129314 BCN121129314 BCN 121129314BCN-121129314-B

Abstract

The invention discloses a PD movement subtype classification method based on transcranial Doppler and Bayesian analysis, which relates to the technical field of medical diagnosis, and comprises the following steps of S1, acquiring brain blood flow velocity signals of a subject in different physiological states through transcranial Doppler equipment, and synchronously acquiring arterial blood pressure signals through continuous noninvasive blood pressure monitoring equipment; S2, processing the signals acquired in the step S1, extracting characteristic indexes reflecting the automatic regulation function of dynamic cerebral blood flow, S3, combining the characteristic indexes with basic hemodynamic parameters to form characteristic vectors, S4, inputting the characteristic vectors into a pre-trained Bayesian discrimination model, and S5, classifying the subjects into different Parkinson disease movement subtypes according to the output of the Bayesian discrimination model. The invention has the advantages of improving classification objectivity, capturing dynamic physiological signal characteristics, realizing accurate subtype identification and the like.

Inventors

LI RUI
XING YINGQI
CHEN YUJUN
YANG CHAO
CHEN HONGXIU
LUO YUMENG

Assignees

首都医科大学宣武医院

Dates

Publication Date: 20260508
Application Date: 20250902

Claims (5)

1. A method for classifying PD motion subtypes based on transcranial doppler and bayesian analysis, comprising: S1, acquiring brain blood flow velocity signals of a subject in different physiological states through a transcranial Doppler device, and synchronously acquiring arterial blood pressure signals through a continuous noninvasive blood pressure monitoring device; S2, processing the signals acquired in the S1, and extracting characteristic indexes reflecting the automatic regulation function of dynamic cerebral blood flow; the step S2 comprises the following steps: Converting the time domain signals of the cerebral blood velocity signals and the arterial blood pressure signals into frequency domain signals by adopting Fourier transformation; Calculating frequency domain characteristic parameters between brain blood flow velocity signals and arterial blood pressure signals in one or more preset frequency bands by adopting a transfer function analysis method based on the converted frequency domain signals, wherein the frequency domain characteristic parameters comprise phase, gain and consistency functions; Meanwhile, calculating an average correlation coefficient between the average cerebral blood flow speed and arterial blood pressure according to the average value of the frequency domain characteristic parameters of the bilateral hemispheres; The frequency domain characteristic parameters and the average correlation coefficient are used as characteristic indexes for representing the automatic regulating function of dynamic cerebral blood flow; s3, combining the characteristic indexes with basic hemodynamic parameters to form characteristic vectors; S4, inputting the feature vector into a pre-trained Bayesian discrimination model; And S5, classifying the subjects into different parkinsonism movement subtypes according to the output of the Bayesian discrimination model.
2. The method according to claim 1, wherein S1 comprises: continuously acquiring brain blood flow speed signals of a first preset time period through a transcranial Doppler device and synchronously acquiring arterial blood pressure signals through a continuous noninvasive blood pressure monitoring device under a static lying state of a subject; and continuously acquiring the cerebral blood flow velocity signal and the arterial blood pressure signal within a second preset time period under the condition that the subject experiences posture change induction.
3. The method of claim 2, wherein the acquiring of the transcranial doppler device is performed under the following conditions: the probe of the 1.6 MHz transcranial Doppler equipment is fixed on the bilateral temporal window of the subject, and the detection depth is 45-60 mm; The subjects fasted for no less than 2 hours and deactivated the autonomic nerve drug for no less than 12 hours prior to harvest.
4. The method of claim 1, wherein the baseline hemodynamic parameter is a cerebral arterial blood flow parameter directly acquired from a raw craniofacial doppler signal monitored by a transcranial doppler device, the cerebral arterial blood flow parameter comprising a systolic peak flow rate, a end diastole flow rate, a pulsation index, and a resistance index.
5. The method of claim 1, wherein the bayesian discriminant model is a classification function that is learned from training samples of known subtypes to distinguish between tremor-dominant and dysposture-unstable forms of parkinson's disease.

Description

PD motion subtype classification method based on transcranial Doppler and Bayesian analysis Technical Field The invention relates to the technical field of medical diagnosis, in particular to a PD motion subtype classification method based on transcranial Doppler and Bayesian analysis. Background Parkinson's Disease (PD) is a complex neurodegenerative disease whose clinical manifestations are markedly heterogeneous. The current clinical practice mainly depends on the unified Parkinson disease rating scale (MDS-UPDRS scale) of the dyskinesia association to classify subtypes, and the method has obvious limitations that firstly, the scale evaluation process takes longer time and depends on subjective judgment of doctors, the change of fine movement characteristics is difficult to capture, secondly, the requirement of periodic re-diagnosis makes the dynamic evolution of the symptoms of patients difficult to track in time, and furthermore, the existing method cannot reflect the potential abnormal function of nerve blood vessels. In recent years, it has been found that patients with different motor subtypes of parkinson's disease may have characteristic cerebral vascular dysfunction. Dynamic cerebral blood flow autoregulation function (dCA function) serves as an important mechanism for maintaining cerebral perfusion stable, and its abnormality is closely related to the pathological course of parkinson's disease. However, the prior art has not established an effective objective index to characterize this association, resulting in a clinical lack of reliable biomarker-assisted subtype identification. Although transcranial Doppler (TCD) techniques are capable of noninvasively monitoring cerebral hemodynamic parameters, traditional analytical methods have difficulty in efficiently extracting characteristic information related to motor subtypes. The dynamic relationship between blood pressure and brain blood flow velocity implies important physiological regulatory information, but the prior art fails to fully exploit the underlying classification features in these signals. In addition, the conventional statistical method has the problem of insufficient model adaptability when processing multidimensional physiological parameters, and cannot meet the accuracy requirement of clinical classification. In view of the above, there is a need in the art for improvements. Disclosure of Invention In view of the above, the invention provides a PD motion subtype classification method based on transcranial Doppler and Bayesian analysis, which has the advantages of improving classification objectivity, capturing dynamic physiological signal characteristics and realizing accurate subtype identification. The invention provides a PD motion subtype classification method based on transcranial Doppler and Bayesian analysis, which comprises the following steps: S1, acquiring brain blood flow velocity signals of a subject in different physiological states through a transcranial Doppler device, and synchronously acquiring arterial blood pressure signals through a continuous noninvasive blood pressure monitoring device; S2, processing the signals acquired in the S1, and extracting characteristic indexes reflecting the automatic regulation function of dynamic cerebral blood flow; s3, combining the characteristic indexes with basic hemodynamic parameters to form characteristic vectors; S4, inputting the feature vector into a pre-trained Bayesian discrimination model; And S5, classifying the subjects into different parkinsonism movement subtypes according to the output of the Bayesian discrimination model. In an alternative embodiment, the S1 includes: continuously acquiring brain blood flow speed signals of a first preset time period through a transcranial Doppler device and synchronously acquiring arterial blood pressure signals through a continuous noninvasive blood pressure monitoring device under a static lying state of a subject; and continuously acquiring the cerebral blood flow velocity signal and the arterial blood pressure signal within a second preset time period under the condition that the subject experiences posture change induction. In an alternative embodiment, the acquisition performance conditions of the transcranial doppler device are: The probe of the 1.6MHz transcranial Doppler equipment is fixed on the bilateral temporal window of the subject, and the detection depth is 45-60mm; The subjects fasted for no less than 2 hours and deactivated the autonomic nerve drug for no less than 12 hours prior to harvest. In an alternative embodiment, the S2 includes: Converting the time domain signals of the cerebral blood velocity signals and the arterial blood pressure signals into frequency domain signals by adopting Fourier transformation; Calculating frequency domain characteristic parameters between brain blood flow velocity signals and arterial blood pressure signals in one or more preset frequency bands by adopting a transfer func