Search

CN-122004862-A - Automatic auxiliary diagnosis system for depression based on heart rate variability and machine learning

CN122004862ACN 122004862 ACN122004862 ACN 122004862ACN-122004862-A

Abstract

The invention discloses an automatic auxiliary diagnosis system for depression based on heart rate variability and machine learning, which comprises a portable heart rate variability analyzer and a computer, wherein the portable heart rate variability analyzer acquires photoplethysmography signals and uploads the photoplethysmography signals to the computer, a basic analysis module calculates time-frequency domain indexes, a dynamic induction control module controls the portable heart rate variability analyzer to output visual guidance graphics to indicate a subject to change a basic respiratory state, a multidimensional characteristic decoupling module extracts an instantaneous heart rate sequence, a photoelectric derived respiratory sequence and a microvascular tension sequence from the photoplethysmography signals, a dynamic coupling characteristic calculation module performs cross-spectrum analysis and cross-correlation calculation to obtain physical coupling parameters, and the machine learning diagnosis module constructs the physical coupling parameters and the time-frequency domain indexes into multimode dynamic characteristic vectors and outputs a judgment result by using a classification model. The invention introduces a dynamic load induction paradigm and a single source signal decoupling technology, eliminates respiratory interference and improves the specificity of depression identification.

Inventors

  • YAO XIAOPENG

Assignees

  • 西南医科大学

Dates

Publication Date
20260512
Application Date
20260327

Claims (10)

  1. 1. The automatic auxiliary diagnosis system for the depression based on heart rate variability and machine learning is characterized by comprising a portable heart rate variability analyzer and a computer; the portable heart rate variability analyzer is connected with the computer in a data communication way; The portable heart rate variability analyzer comprises a data acquisition module and a display interface, at least acquires a photoplethysmography wave signal through the interface, performs data processing and packaging through the data acquisition module after filtering, amplifying and analog-to-digital conversion, and uploads the data to the computer; the software running on the computer is internally provided with a basic analysis module, a dynamic induction control module, a multidimensional feature decoupling module, a dynamic coupling feature calculation module and a machine learning diagnosis module; The basic analysis module extracts time data of the interval between two adjacent heartbeats in the photoplethysmograph wave signal, performs time domain analysis and frequency domain analysis, and calculates and outputs time-frequency domain indexes; the dynamic induction control module generates an induction control signal with specific frequency and transmits the induction control signal to the portable heart rate variability analyzer, and controls the portable heart rate variability analyzer to output a visual guidance graph on the display interface so as to instruct a subject to change a basic respiratory state; the multidimensional feature decoupling module receives the photoplethysmograph signal and extracts an instantaneous heart rhythm sequence, a photoelectrically derived respiration sequence and a microvascular tension sequence; the dynamic coupling characteristic calculation module receives the decoupled instantaneous rhythm sequence, the photoelectric derived breathing sequence and the microvascular tension sequence, and performs cross spectrum analysis and cross correlation calculation to obtain physical coupling parameters; the machine learning diagnosis module extracts the physical coupling parameters and the time-frequency domain indexes to construct a multi-mode dynamic feature vector, predicts a judging result of the mental state of the subject by using a classification model, and generates a report containing the judging result.
  2. 2. The automatic auxiliary diagnosis system for depression based on heart rate variability and machine learning according to claim 1, wherein the external connection state of the portable heart rate variability analyzer comprises: The portable heart rate variability analyzer is connected with a photoelectric wave cable through a preset photoelectric plethysmograph wave interface, and the tail end of the photoelectric wave cable is connected with a pulse finger probe which is specifically a short plethysmograph; During measurement, the short plethysmograph is clamped to the subject's left index finger with the subject's left index finger nail facing up and the short plethysmograph's cable is walked over the subject's back of the hand.
  3. 3. The automatic aided diagnosis system of depression based on heart rate variability and machine learning of claim 1, wherein the dynamic evoked control module performs dynamic test window timing control, in particular comprising: Within a baseline test window, the system indicates that the subject is in a resting state, the data acquisition module continuously acquires the photoplethysmograph wave signal at the baseline state; Within a load induction window, the portable heart rate variability analyzer outputs the visual guidance pattern to instruct the subject to breathe deeply and slowly, and the data acquisition module continuously acquires the photoplethysmograph wave signals in an induction state; And in a recovery test window, the portable heart rate variability analyzer stops outputting the visual guide graph to indicate the subject to recover the natural respiratory state, and the data acquisition module continuously acquires the photoplethysmograph wave signals in the recovery state.
  4. 4. The automatic auxiliary diagnosis system for depression based on heart rate variability and machine learning according to claim 3, wherein the portable heart rate variability analyzer is further integrated with an audio playing unit, and the process of performing visual and auditory joint guidance by the portable heart rate variability analyzer comprises: the portable heart rate variability analyzer analyzes the received guiding control instruction to extract breathing guiding frequency, and drives the display interface to output the visual guiding graph; the visual guide pattern changes the alternation of an enlarged state and a reduced state of the pattern according to the breathing guide frequency, wherein the enlarged state is used for indicating the testee to execute the inspiration action, and the reduced state is used for indicating the testee to execute the expiration action; The portable heart rate variability analyzer synchronously drives the audio playing unit to output an auditory guidance rhythm which is kept time-phase synchronous with the visual guidance pattern morphological change.
  5. 5. The automatic aided diagnosis system of depression based on heart rate variability and machine learning of claim 1, wherein the process of extracting the instantaneous heart rate sequence and the photoelectrically derived respiration sequence by the multidimensional feature decoupling module comprises: The multidimensional feature decoupling module performs morphological feature analysis on the discretized photoplethysmograph wave signals, identifies peak points in the systolic period and records corresponding time values of the peak points in the systolic period; calculating the difference value of the time values of the peak points of two adjacent systoles, and constructing the instant heart rhythm sequence equivalent to the electrocardio R-R interval; And filtering the discretized photoplethysmograph wave signals by using a low-pass filter with preset cut-off frequency, extracting a low-frequency baseline fluctuation signal, and resampling the low-frequency baseline fluctuation signal at the time value of the peak point of the systolic period to construct the photoelectric derivative respiratory sequence.
  6. 6. The automatic aided diagnosis system of depression based on heart rate variability and machine learning of claim 5, wherein the process of extracting the microvascular tension sequence by the multidimensional feature decoupling module comprises: Taking two adjacent peak point time values in the systolic period as a search interval, and traversing the discretized photoplethysmograph wave signals in the search interval; extracting a local minimum value in the search interval as a direct current component; extracting a local maximum value in the search interval, and calculating a difference value between the local maximum value and the local minimum value as an alternating current component; and calculating the ratio of the alternating current component to the direct current component, and constructing the microvascular tension sequence.
  7. 7. The automatic auxiliary diagnosis system for depression based on heart rate variability and machine learning according to claim 3, wherein the process of calculating the physical coupling parameters by the dynamic coupling feature calculation module comprises calculation of heart lung transfer function gain: calculating a self-power spectral density of the photoelectrically derived respiration sequence within the load induction window; Calculating cross-power spectral densities of the photoelectrically derived respiratory sequence and the instantaneous cardiac rhythm sequence within the load induction window; dividing the cross-power spectral density by the self-power spectral density to calculate a heart-lung transfer function; And extracting a transfer function amplitude at a preset respiratory guidance frequency, and taking the transfer function amplitude as a heart lung transfer function gain in the physical coupling parameters.
  8. 8. The automatic aided diagnosis system of depression based on heart rate variability and machine learning of claim 7, wherein the process of calculating the physical coupling parameters by the dynamic coupling feature calculation module further comprises cross-correlation calculation: performing first-order differential calculation on the instantaneous rhythm sequence in the recovery test window to construct a parasympathetic rebound slope sequence; performing first-order differential calculation on the microvascular tension sequence in the recovery test window to construct a sympathogenic damping attenuation sequence; Calculating a cross-correlation function of said parasympathetic bounce slope sequence and said sympatholytic decay sequence according to a finite discrete sequence calculation rule; And extracting the maximum value of the cross-correlation function as a maximum cross-correlation coefficient, extracting the delay time corresponding to the maximum cross-correlation coefficient as an optimal delay time, and adding the maximum cross-correlation coefficient and the optimal delay time into the physical coupling parameter.
  9. 9. The automatic auxiliary diagnosis system for depression based on heart rate variability and machine learning according to claim 8, wherein the process of constructing the multi-modal dynamic feature vector by the machine learning diagnosis module comprises: calculating the dynamic change rate of a preset evaluation index, wherein the preset evaluation index is limited to an RMSSD index and an HF index; Calculating the difference value between the evaluation index in the load induction window and the evaluation index in the baseline test window, and dividing the difference value by the evaluation index in the baseline test window to respectively obtain the dynamic change rate of the RMSSD index and the dynamic change rate of the HF index; And performing cascade concatenation on the dynamic change rate of the RMSSD index, the dynamic change rate of the HF index, the heart-lung transfer function gain, the maximum cross-correlation coefficient and the optimal delay time to construct the multi-mode dynamic feature vector.
  10. 10. The automatic auxiliary diagnosis system for depression based on heart rate variability and machine learning according to claim 9, wherein the process of generating a report containing the determination result by the machine learning diagnosis module comprises: Inputting the constructed multi-mode dynamic feature vector into the classification model, wherein the classification model specifically adopts a support vector machine model or a random forest model; the classification model is constructed based on the mapping relation between the multidimensional heart rate variability characteristics of the known samples and the corresponding mental state categories, wherein the corresponding mental state categories comprise a healthy state, a moderate depression state and a major depression state; The classification model processes the multimodal dynamic feature vector and outputs a predicted classification label indicating that the subject is in one of the healthy state, the moderately depressed state, or the severely depressed state as a final decision.

Description

Automatic auxiliary diagnosis system for depression based on heart rate variability and machine learning Technical Field The invention relates to the technical field of medical auxiliary diagnosis, in particular to an automatic auxiliary diagnosis system for depression based on heart rate variability and machine learning. Background In objective physiological assessment of mental diseases, heart rate variability analysis is widely used as a non-invasive detection technique reflecting the function of the autonomic nervous system. Existing computer-aided diagnosis schemes typically utilize physiological signal acquisition devices to acquire the subject's electrocardiographic or photoplethysmographic signals. After the computer receives the physiological signals, the heartbeat interval time sequence is extracted, and the time domain parameters and the frequency domain parameters are calculated. Then, the system inputs the calculated time domain parameters and frequency domain parameters into a pre-trained machine learning model, tries to establish a mapping relation between heart rate variability characteristics and depression, and further outputs a mental state evaluation result for the subject. However, the existing automatic auxiliary diagnosis system for depression has the technical defect of insufficient physiological data identification specificity. The existing heart rate variability measurement and feature extraction process is completely limited to a subject in a resting state. The conventional resting state evaluation mode can only reflect the basic tension of the autonomic nervous system under no external physiological load, and can not excite and record the dynamic regulation process of the autonomic nervous system when the autonomic nervous system responds to the stimulus. An important feature of a patient suffering from depression is a decrease in neuromodulation elasticity caused by an impaired function of the autonomic neuromodulation center. Depending on the data acquired in the resting state, abnormal manifestations of the patient's sympatholytic and parasympathetic antagonism mechanisms cannot be effectively exposed, resulting in insufficient discrimination of the characteristic parameters extracted by the existing system in objectively discriminating between healthy and depressed states. Disclosure of Invention Aiming at the defects of the prior art, the invention provides an automatic auxiliary diagnosis system for depression based on heart rate variability and machine learning, which solves the problems that the physiological evaluation of depression in the prior art is highly dependent on resting state data, is easily interfered by respiratory rhythm and lacks objective dynamic regulation and control characteristics. The automatic auxiliary diagnosis system for the depression based on heart rate variability and machine learning comprises a portable heart rate variability analyzer and a computer. And a data communication connection is established between the portable heart rate variability analyzer and the computer. The portable heart rate variability analyzer comprises a data acquisition module and a display interface. The portable heart rate variability analyzer at least acquires photoplethysmographic signals through an interface, performs data processing and packaging through a data acquisition module after filtering, amplifying and analog-to-digital conversion, and uploads the photoplethysmographic signals to a computer. The software running on the computer is internally provided with a basic analysis module, a dynamic induction control module, a multidimensional feature decoupling module, a dynamic coupling feature calculation module and a machine learning diagnosis module. The basic analysis module extracts time data of the interval between two adjacent heartbeats in the photoplethysmograph wave signal, performs time domain analysis and frequency domain analysis, and calculates and outputs time-frequency domain indexes. The dynamic induction control module generates an induction control signal with specific frequency and transmits the induction control signal to the portable heart rate variability analyzer, and the portable heart rate variability analyzer is controlled to output a visual guidance graph on a display interface so as to instruct the subject to change the basic respiratory state. The multidimensional feature decoupling module receives the photoplethysmograph signal and extracts the instantaneous heart rhythm sequence, the photoelectrically derived respiration sequence, and the microvascular tension sequence. The dynamic coupling characteristic calculation module receives the decoupled instantaneous rhythm sequence, the photoelectric derivative breathing sequence and the microvascular tension sequence, and performs cross spectrum analysis and cross correlation calculation to obtain physical coupling parameters. The machine learning diagnosis module extracts the physical