CN-122000030-A - Parkinson's disease Cheng Dingliang assessment method and system based on multi-scale feature fusion

CN122000030ACN 122000030 ACN122000030 ACN 122000030ACN-122000030-A

Abstract

The application provides a parkinsonism Cheng Dingliang assessment method and system based on multi-scale feature fusion, comprising simulating continuous disease progression from healthy to severe in parkinsonism state based on a constructed cortical-basal segment-thalamus nerve loop model to obtain neuron pulse sequence signals of each brain region in different progression stages, extracting multi-scale features from the neuron pulse sequence signals to obtain micro discharge features, macro oscillation features and network synchronization features, sorting contribution degrees of the micro discharge features, the macro oscillation features and the network synchronization features through a recursive feature elimination algorithm to screen out a core feature subset which is most sensitive to disease progression, and inputting the screened core feature subset into a constructed machine learning model to output an assessment result of parkinsonism. The application overcomes the saturation effect of a single index in the later stage of the disease course, realizes the linear evaluation of the whole disease course, and can more comprehensively and accurately quantify the severity of the disease.

Inventors

ZHANG SHUAI
WANG YANBIN
YUE KE
AN YUCHEN
LIU JIZHOU
SUN JIAQI

Assignees

河北工业大学

Dates

Publication Date: 20260508
Application Date: 20260127

Claims (10)

1. A method for evaluating parkinson's disease Cheng Dingliang based on multi-scale feature fusion, comprising: Simulating continuous disease progress of the Parkinson state from healthy to severe based on the constructed cortical-basal ganglia-thalamus nerve loop model so as to acquire neuron pulse sequence signals of each brain region in different progress stages; performing multi-scale feature extraction on the neuron pulse sequence signal to obtain microscopic discharge features, macroscopic oscillation features and network synchronization features; Sorting contribution degrees of the micro discharge features, the macro oscillation features and the network synchronization features through a recursive feature elimination algorithm so as to screen out a core feature subset which is most sensitive to disease progress; and inputting the screened core feature subsets into a constructed machine learning model to output an evaluation result of the parkinsonism process.
2. The method according to claim 1, characterized in that: The key conductivity parameters in the model are regulated to simulate the pathological evolution of basal ganglia network caused by striatal dopamine deficiency, and the disease course coefficient is set, so that neuron pulse sequence signals of each brain region in different progress stages are obtained; Among these, the parameter regulation mechanisms include cortical-striatal synaptic efficacy modulation, striatal neuronal excitability modulation and lateral pallidosis enhancement.
3. The method according to claim 1, characterized in that: And analyzing the change trend of the two parkinsonism treatment markers along with the disease course coefficient by selecting the STN average discharge rate and the Beta frequency band power so as to carry out validity verification on the generated simulation data.
4. The method according to claim 1, characterized in that: the microcosmic discharge characteristics comprise average discharge rate, local variation coefficient and Fano factor, and are used for quantifying the discharge mode and variability of single neurons; wherein the average firing rate is used to characterize the overall neuronal excitability level by calculating the total number of pulses sent by the neurons per unit time; the local variation coefficient is used for quantifying the irregularity of the discharge pattern by calculating the variability of the adjacent pulse intervals; The Fano factor is used to measure the variability of the pulse sequence on a time scale by calculating the ratio of the variance to the mean of the pulse counts over a fixed time window.
5. The method according to claim 1, characterized in that: The macroscopic oscillation feature is to convolve a discrete pulse sequence with a Gaussian kernel function to generate an LEP-like signal, and calculate relative power spectral density, sample entropy and Hjorth parameters of a key frequency band, wherein the Hjorth parameters comprise activity, mobility and complexity of the signal.
6. The method according to claim 1, characterized in that: The network synchronization features include coherence, phase lock values, cross-correlation between subthalamic nucleus-lateral pallidum regions, and phase-amplitude coupling between subthalamic nucleus-skin layers to reveal pathological connection features at the network level.
7. The method according to claim 1, characterized in that: in the machine learning model training process, 5-fold cross validation is adopted to combine with a grid search strategy to carry out super-parameter optimization, and the determined optimal super-parameter combination is decision tree number, leaf node minimum sample number, splitting minimum sample number and maximum depth.
8. A parkinsonism stroke quantitative assessment system based on multi-scale feature fusion, comprising: a signal acquisition module configured to simulate continuous course progression of parkinson's state from healthy to severe based on the constructed cortical-basal segment-thalamous neural loop model to acquire neuronal pulse sequence signals for each brain region at different stages of progression; the feature extraction module is configured to obtain microscopic discharge features, macroscopic oscillation features and network synchronization features by performing multi-scale feature extraction on the neuron pulse sequence signals; The core feature screening module is configured to rank contribution degrees of the micro discharge features, the macro oscillation features and the network synchronization features through a recursive feature elimination algorithm so as to screen out a core feature subset which is most sensitive to the progress of the disease; And the course evaluation module is configured to input the screened core feature subsets into the constructed machine learning model so as to output an evaluation result of the parkinsonism course.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any of claims 1-7 when the program is executed by the processor.
10. A non-transitory computer readable storage medium, wherein the non-transitory computer readable storage medium stores computer instructions for causing a computer to perform the method of any one of claims 1-7.

Description

Parkinson's disease Cheng Dingliang assessment method and system based on multi-scale feature fusion Technical Field The application belongs to the technical field of human brain simulation research, and particularly relates to a method and a system for evaluating Parkinson's disease Cheng Dingliang based on multi-scale feature fusion. Background Parkinson's disease is a chronic progressive neurodegenerative disease. The current clinical evaluation of the course of disease mainly depends on the unified Parkinson disease score scale (UPDRS), and the mode has extremely strong subjectivity and cannot realize continuous and objective monitoring. In the field of electrophysiological studies, existing evaluation methods mostly depend on a single Beta-band (13-30 Hz) oscillation power. However, the prior art suffers from the following significant drawbacks: 1. The index saturation problem is that Beta power often enters a plateau after entering a middle-late stage (PD coefficient > 0.5) of a disease course, and the moderate disease course and the severe disease course cannot be distinguished. 2. Feature scale singleness-only concern macroscopic field potentials (LFP) or microscopic pulses (Spike), ignoring the inherent correlation between neuron discharge patterns, population oscillations, and inter-region synchronization. 3. The nonlinear capture is insufficient, and the traditional linear model is difficult to describe the complex nonlinear pathological evolution in the course of disease. Disclosure of Invention In view of the foregoing, the present application aims to provide a method and a system for evaluating parkinson's disease Cheng Dingliang based on multi-scale feature fusion, so as to solve at least one of the above problems. In order to achieve the above purpose, the technical scheme of the application is realized as follows: in a first aspect, the application provides a method for evaluating parkinson's disease Cheng Dingliang based on multi-scale feature fusion, comprising: Simulating continuous disease progress of the Parkinson state from healthy to severe based on the constructed cortical-basal ganglia-thalamus nerve loop model so as to acquire neuron pulse sequence signals of each brain region in different progress stages; performing multi-scale feature extraction on the neuron pulse sequence signal to obtain microscopic discharge features, macroscopic oscillation features and network synchronization features; Sorting contribution degrees of the micro discharge features, the macro oscillation features and the network synchronization features through a recursive feature elimination algorithm so as to screen out a core feature subset which is most sensitive to disease progress; and inputting the screened core feature subsets into a constructed machine learning model to output an evaluation result of the parkinsonism process. In a second aspect, based on the same inventive concept, the application also provides a parkinsonism process quantitative evaluation system based on multi-scale feature fusion, which comprises: a signal acquisition module configured to simulate continuous course progression of parkinson's state from healthy to severe based on the constructed cortical-basal segment-thalamous neural loop model to acquire neuronal pulse sequence signals for each brain region at different stages of progression; the feature extraction module is configured to obtain microscopic discharge features, macroscopic oscillation features and network synchronization features by performing multi-scale feature extraction on the neuron pulse sequence signals; The core feature screening module is configured to rank contribution degrees of the micro discharge features, the macro oscillation features and the network synchronization features through a recursive feature elimination algorithm so as to screen out a core feature subset which is most sensitive to the progress of the disease; And the course evaluation module is configured to input the screened core feature subsets into the constructed machine learning model so as to output an evaluation result of the parkinsonism course. In a third aspect, based on the same inventive concept, the present application also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method according to the first aspect when executing the program. In a fourth aspect, based on the same inventive concept, the present application also provides a non-transitory computer readable storage medium, wherein the non-transitory computer readable storage medium stores computer instructions for causing the computer to perform the method according to the first aspect. Compared with the prior art, the Parkinson disease Cheng Dingliang evaluation method and system based on multi-scale feature fusion have the following beneficial effects: The application discloses a multi-scale featu