A device for screening dysphagia after stroke based on voice acoustic parameters
The predictive model constructed using voice acoustic parameters solves the accuracy problem of screening for dysphagia after stroke, achieving early identification and reducing the risk of aspiration.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHANGZHOU DEAN HOSPITAL
- Filing Date
- 2026-03-27
- Publication Date
- 2026-06-09
AI Technical Summary
Existing technologies are insufficient for effectively screening for dysphagia after stroke, leading to high aspiration rates and the risk of related complications. There is a lack of rapid and accurate screening methods.
A predictive model is built based on voice acoustic parameters (VAP). The patient's voice parameters are collected by a noise detection device, and a predictive model is established using logistic regression analysis to output the risk probability of swallowing dysfunction in stroke patients.
It provides a reliable and rapid screening method, improves the screening accuracy of dysphagia after stroke, enables early identification of dysphagia and timely intervention, and reduces the risk of aspiration.
Smart Images
Figure CN122163154A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a device for screening dysphagia after stroke based on voice acoustic parameters, belonging to the field of biomedicine. Background Technology
[0002] According to the Global Burden of Disease Stroke Collaboration report, there were 101 million stroke cases globally in 2019, with stroke-related disability-adjusted life years (DAWs) reaching 143 million. Looking at the 30 years from 1990 to 2019, the absolute number of stroke events increased by 70%, stroke-related deaths increased by 43%, and DAWs increased by 32%. With the increasing global elderly population, the burden of stroke is expected to continue to rise annually worldwide. The burden of stroke in China is also concerning. According to the 2022 Stroke Center Report, the standardized prevalence of stroke among Chinese residents aged ≥40 years reached 2.64% in 2021. The average age of first-time stroke was between 60.9 and 63.4 years, with the 40-64 age group accounting for over 66.6%, indicating that the age of stroke onset in China is increasingly younger. The report also shows that the recovery rate for hospitalized stroke patients is only 44.11%; among stroke survivors who underwent a 3-month follow-up using the modified Rankin Scale, 52.9% had residual symptoms of varying degrees, with 27.86% meeting the criteria for disability. This demonstrates the immense pressure that stroke rehabilitation and prevention face in China.
[0003] Dysphagia refers to the inability of patients to safely and effectively transport food from the mouth to the stomach to obtain sufficient water and nutrients. Because maintaining normal swallowing function requires coordinated control of the central and peripheral nervous systems, dysphagia has a high incidence in neurological disorders, especially in stroke patients. According to relevant studies both domestically and internationally, the overall prevalence of dysphagia after stroke ranges from 27% to 64%. The incidence of dysphagia is as high as 72.4% during the acute phase of stroke, while over 50% of patients recover safe swallowing function during the subacute phase. Approximately 5% of patients still experience dysphagia six months after stroke. Among stroke patients with dysphagia, over 40% are at risk of aspiration, 4% may die directly from aspiration, and approximately 37% may develop aspiration pneumonia, an independent risk factor for increased mortality in stroke patients. Due to the high aspiration rate and high mortality risk associated with dysphagia, it has become a key focus in global stroke rehabilitation management. The Chinese guidelines for the management of dysphagia rehabilitation also strongly recommend that early screening and timely intervention be carried out for high-risk groups of dysphagia, including those with stroke, in order to prevent the risk of related complications. Summary of the Invention
[0004] The purpose of this invention is to provide a device for screening dysphagia after stroke based on voice acoustic parameters. It is used to measure objective voice parameters of stroke patients with or without dysphagia, explore the predictive role of VAP on dysphagia after stroke using logistic regression analysis, and establish a regression model to enrich the rapid screening methods for dysphagia after stroke and further promote early screening.
[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a device for screening dysphagia after stroke based on voice acoustic parameters, comprising a noise detection device, an input device, a processing device and an output device.
[0006] The noise detection device is used to collect the measured values of the patient's longest vocalization time (MPT), highest sound intensity (I_high), highest fundamental frequency (F0_high), lowest sound intensity (I_low), and fundamental frequency jitter. The input device is used to collect brainstem lesion values from patients.
[0007] The processing device is connected to both the noise detection device and the input device. The measured values of the Voice Disorder Index (DSI) are calculated using the measured values of the longest vocalization time, the highest fundamental frequency, the lowest sound intensity, and the fundamental frequency perturbation as variables. The total score of the prediction model is calculated using the brainstem lesion assignment, the measured values of the longest vocalization time, the highest sound intensity, and the measured DSI as variables. The risk probability value of dysphagia in stroke patients is obtained based on the total score of the prediction model.
[0008] The output device is connected to the processing device and is used to output the risk probability value of stroke patients developing swallowing dysfunction.
[0009] According to the above technical solution, the processing device is used to draw a nomogram predicting the risk of dysphagia after stroke, and the output device is also used to output the total score of the prediction model and the nomogram predicting the risk of dysphagia after stroke.
[0010] According to the above technical solution, the nomogram for predicting the risk of dysphagia after stroke includes a scaled-down scoring axis, brainstem lesion assignment, measured value of longest vocalization time, measured value of highest vocal intensity, measured value of DSI, total score axis, and risk axis for dysphagia after stroke. A vertical line is drawn upwards from the corresponding variable axis for brainstem lesion assignment, measured value of longest vocalization time, measured value of highest vocal intensity, and measured value of DSI. The intersection of this line with the scoring axis represents the score value corresponding to that variable. The sum of the score values corresponding to the four variables is the total score of the prediction model. A vertical line is drawn downwards from the total score axis. The intersection of this line with the risk axis for dysphagia after stroke represents the probability of dysphagia in stroke patients.
[0011] According to the above technical solution, the DSI in the processing device is calculated based on the formula: DSI=0.13×MPT+0.0053×F0_high-0.26×I_low-1.18×Jitter+12.4 to obtain the measured DSI value.
[0012] According to the above technical solution, the noise detection device includes a recording module and an environmental noise monitoring module. The recording module is used to collect the user's audio, and the environmental noise monitoring module is used to collect ambient sound and determine the noise intensity.
[0013] A method of using a device for screening post-stroke dysphagia based on voice acoustic parameters includes the following steps: S1: The input device prompts the patient to input brainstem lesion values, gender, and age; S2: Noise detection device detects ambient noise; S3: A pop-up window prompts the user to sit upright, keeping the lower lip about 30cm away from the phone microphone, and a demonstration video is played simultaneously to help the user correct their posture; S4: The user is prompted to pronounce words via voice, and the recording module simultaneously collects audio signals; S5: The processing device calculates the risk probability value of stroke patients developing dysphagia based on the data collected by the input device and the noise detection device and transmits it to the output device. S6: The output device outputs the probability value of the risk of swallowing dysfunction in stroke patients.
[0014] According to the above technical solution, S2 further includes a pop-up window prompting the user to change the collection environment when the environmental noise monitoring module detects environmental noise ≥45dB.
[0015] According to the above technical solution, step S4 further includes the following steps: S41: The user is prompted by voice to continuously pronounce long vowels from low to high frequencies, and the recording module simultaneously collects the audio signal and extracts F0_high; S42: The voice prompts the user to pronounce long vowels in a comfortable tone, with the volume gradually increasing from weak to strong. The recording module simultaneously collects the audio signal and extracts I_high and I_low. S43: After prompting the user to take a deep breath, the user should continuously pronounce a long vowel with a steady pitch and loudness until there is no airflow. The recording module will record the sound three times, with an interval of ≥15 seconds between each recording. The longest duration will be automatically selected as the MPT, and the signal will be extracted from the steady segment of 3-4 seconds of pronunciation to obtain the Jitter.
[0016] According to the above technical solution, the output device is integrated into the application software and is used to output the risk probability value of stroke patients developing swallowing dysfunction in the application software. Beneficial effects
[0017] The nomogram model based on VAP can effectively predict the risk of dysphagia after stroke, has reliable clinical value, and can be used as an effective supplementary means to identify dysphagia after stroke in clinical practice. It can also be cross-validated with other swallowing detection methods to further improve the accuracy of dysphagia screening after stroke.
[0018] A nomogram prediction model for dysphagia after stroke was constructed based on voice acoustic parameters (VAP), enriching the means of rapid screening for dysphagia after stroke and providing a new method for early screening. Attached Figure Description
[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 It is a nomogram model for predicting the risk of dysphagia after stroke; Figure 2 It is the ROC curve of the training set post-stroke dysphagia risk nomogram prediction model; Figure 3 It is the calibration curve of the nomogram prediction model for the risk of dysphagia after stroke; Figure 4 It is the DCA curve of the nomogram prediction model for dysphagia risk after stroke; Figure 5 It is the ROC curve of the nomogram prediction model for dysphagia risk after stroke; Figure 6 It is a calibration curve of the nomogram prediction model for the risk of dysphagia after stroke; Figure 7 It is a validation set of the DCA curve of the nomogram prediction model for dysphagia risk after stroke. Detailed Implementation
[0020] The present invention will be specifically described below through embodiments. It should be noted that the following embodiments are only used to further illustrate the present invention and should not be construed as limiting the scope of protection of the present invention. Those skilled in the art can make some non-essential improvements and adjustments to the present invention based on the above description. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those skilled in the art.
[0021] MPT refers to the longest duration an individual can sustain a single vowel sound after a deep inhalation. It directly reflects changes in laryngeal dynamics. Previous studies have shown that MPT is closely related to laryngeal and airway resistance, reflecting not only the pathological state of the vocal cords but also indirectly the roughness and hoarseness of the voice, making it a key parameter objectively reflecting voice quality. I_high refers to one of the main attributes of the human voice, representing the objective physical strength of sound, reflecting the amount of energy carried during sound propagation. Physiologically, sound intensity primarily depends on the subglottic pressure and the intensity of activity of the vocal organs' muscle groups during phonation. Increased muscle activity and a larger exhaled airflow during phonation increase the pressure of the airflow on the vocal organs, resulting in greater amplitude of vocal cord vibration and a stronger sound; conversely, a weaker sound results in a weaker sound. DSI refers to a speech impairment severity index. It is modeled based on multiple linear regression analysis and finally selected four VAPs, F0_high, I_low, Jitter and MPT, to construct the regression equation.
[0022] F0 refers to the fundamental frequency of the periodic vibration of the vocal cords, and its level is affected by subglottic pressure, cricothyroid muscle, and thyroarytenoid muscle.
[0023] Jitter belongs to the VAP (Voice Amplitude Perturbation) type of perturbation. It reflects the degree of variation in vocal cord vibration, that is, the degree of change in the fundamental frequency of sound waves between adjacent cycles. The smaller the value, the more stable the vocal cord vibration. When an individual's vocal cords have pathological changes, the physiological vibration pattern of the vocal cords will be damaged, leading to an increased level of perturbation in phonation. Sound intensity is the strength of sound, and its level is mainly related to the amplitude of vocal cord vibration; the larger the amplitude, the higher the sound intensity.
[0024] Example This study included 350 stroke patients from the Rehabilitation Center of Changzhou De'an Hospital. Of these, 322 were initially included, with 22 dropping out. A total of 300 stroke patients were ultimately included.
[0025] The acoustic testing methods for noise are as follows: 1. Collect basic characteristics of each patient, including gender, age, smoking status (smokers are defined as those who smoke continuously or cumulatively for ≥6 months), alcohol consumption status (those who consume ≥25g of pure ethanol per month are defined as drinkers), stroke type (cerebral infarction / cerebral hemorrhage), and brainstem lesion (brainstem / non-brainstem).
[0026] 2. Further testing was conducted using Ling-WAVES from ATMOS, Germany: During the data collection process, the ambient noise level should be <45dB. The patient should be seated upright. Adjust the microphone height so that the patient's lower lip is at the same height as the microphone. Use the calipers provided with the assessment system to maintain the distance between the patient's lower lip and the microphone at 30cm. Instruct the patient to pronounce the vowel / a: / from low to high frequencies. The machine will automatically collect the lowest fundamental frequency F0_low and the highest fundamental frequency F0_high. Instruct the patient to pronounce / a: / in a comfortable tone, gradually increasing the volume. The machine will automatically collect the lowest intensity I_low and the highest intensity I_high. Instruct the patient to take a deep breath and pronounce the vowel / a: / for as long as possible with a comfortable and stable pitch and loudness until there is no airflow. Repeat the test 3 times, with an interval of more than 15 seconds between each test. The longest pronunciation time is recorded as the longest pronunciation time. The MPT system will automatically capture the patient's most stable and strongest signal segment of 3-4 seconds from the above recordings and analyze the Jiitter and Shimmer. After the test, the system analysis software will automatically provide the measured DSI value.
[0027] The DSI calculation method is as follows: DSI = 0.13 × MPT + 0.0053 × F0_high - 0.26 × I_low - 1.18 × Jitter + 12.4. The calculated DSI value is the actual measured value. When DSI is negative, the smaller the value, the more severe the hoarseness. When DSI is positive, the larger the positive value, the milder the hoarseness and the closer the voice is to normal.
[0028] The clinical screening methods for dysphagia are as follows: 1. FEES testing was conducted. Before the FEES test, the necessary items and food were prepared, including tissues, three disposable paper cups, a spoon, and a 10ml syringe. According to JSDR 2013 thickening standards, the liquid was prepared into three thicknesses: low, medium, and high. The thickener used in this test was Shushisu S. The preparation methods for different thicknesses were as follows: Low thickness – 1g of Shushisu S added to 100ml of warm water; Medium thickness – 2g of Shushisu S added to 100ml of warm water; High thickness – 3g of Shushisu S added to 100ml of warm water. The food was then stained with green dye.
[0029] 2. Patients with indwelling nasogastric tubes must have the tube removed before the examination. The endoscope is inserted into the pharynx through the nasal cavity. The patient is instructed to ingest 3ml, 5ml, and 10ml of medium-thickness food, high-thickness food, and low-thickness food in sequence. Endoscopic observation includes: the condition of the patient's tongue base, epiglottis, piriform sinus, pharyngeal wall, larynx, and vocal cords; the patient's swallowing initiation speed and the amount of food remaining in the piriform sinus and vallecula after swallowing; and whether there is staining of the lower respiratory tract (signs such as food bolus entering the laryngeal vestibule and aspiration).
[0030] 3. A VF (Vacuum-Fluid) examination can also be used. Before the VF examination, prepare the necessary items and food, including tissues, three disposable paper cups, a spoon, and a 10ml syringe. Iohexol is used as the contrast agent. Prepare low-, medium-, and high-viscosity foods using a contrast agent like iohexol S. The preparation methods for different consistencies are as follows: 40ml of iohexol + 80ml of water to make a stock solution; low-viscosity – 45ml stock solution; medium-viscosity – 40ml stock solution with 0.8g of contrast agent S added; high-viscosity – 35ml stock solution with 1g of contrast agent S added. Patients with indwelling nasogastric tubes need to have the tube removed before the examination. The patient is seated, and a lateral view is taken. The image shows the area from the lips anteriorly, to the nasal cavity superiorly, to the spine posteriorly, and below the upper esophageal sphincter. Instruct the patient to eat 3ml, 5ml, and 10ml of medium-viscosity, high-viscosity, and low-viscosity foods respectively. During the examination, the patient should be instructed to swallow the entire bolus in one gulp, ensuring it passes completely down the esophagus before attempting the next bolus. If choking or aspiration occurs, immediate chest percussion or phlegm removal is necessary.
[0031] The diagnostic methods for dysphagia are as follows: The swallowing initiation speed (normal initiation / delayed initiation), leakage / aspiration (based on the Rosenbek leakage / aspiration score), and residue (normal / mild / moderate / severe residue) are evaluated based on the observed conditions. If the patient's swallowing initiation is normal and no leakage, aspiration (Rosenbek score of 2-4 indicates leakage, 5-8 indicates aspiration), or residue is observed, the patient is considered to have no swallowing disorder. All other conditions are considered swallowing disorders.
[0032] The statistical analysis methods are as follows: 1. Compile the basic characteristics, acoustic system test data, and FEES or VFSS diagnostic results for each patient. Classify patients according to the presence or absence of swallowing disorders. Enter the data into an Excel spreadsheet twice, allowing Excel to automatically compare the two entries to improve accuracy and avoid data errors or omissions caused by human error. After verification, import the data into SPSS statistical software for analysis.
[0033] 2. Data were processed and analyzed using SPSS 27.0 statistical software and R4.4.2 software. Categorical data were described using frequencies or rates (%). The chi-square test was used to compare rates, and the chi-square test was used to compare differences between groups. Continuous data were first tested for KS normality and homogeneity of variance. Data that were normally distributed and had homogeneous variances were described as mean ± standard deviation, and the independent samples t-test was used to compare two groups. Data that were not normally distributed or had unequal variances were described using M(Q1, Q3), and nonparametric tests were used to compare two groups. Based on the variables selected from the difference analysis, a logistic regression was fitted. The "rms" package in R4.4.2 software was used to create a nomogram predicting the risk of dysphagia after stroke, thus visualizing the predictive model.
[0034] The bootstrap self-sampling method (1000 times) was used for internal validation and evaluation of the constructed predictive model. Receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) was calculated to evaluate the model's discriminative power. A calibration plot was plotted to evaluate calibration, and a decision curve analysis (DCA) was plotted to evaluate the model's clinical efficacy. External validation of the predictive model was performed using data from the validation set, further evaluating the model's AUC, calibration, and clinical efficacy. All differences were considered statistically significant at P < 0.05.
[0035] 2.1 General information of the dysphagia group and the non-dysphagia group was compared. The results are shown in Table 1. There were statistically significant differences between the two groups in age, sex, and presence or absence of brainstem injury (P<0.05). There were no statistically significant differences between the two groups in other general information.
[0036] Table 1. Analysis of basic characteristics of patients with dysphagia after stroke.
[0037] 2.2 VAP analysis was performed on patients with dysphagia after stroke. The results are shown in Table 2. The levels of F0_high, MPT, I_high, and DSI in the dysphagia group were significantly lower than those in the non-dysphagia group (P<0.05). The levels of Jitter and Shimmer in the dysphagia group were significantly higher than those in the non-dysphagia group (P<0.05). There was no statistically significant difference in the levels of F0_low and I_low between the two groups (P>0.05).
[0038] Table 2. VAP analysis of patients with dysphagia after stroke
[0039] Where F0_high is the highest fundamental frequency; F0_low is the lowest fundamental frequency; Jitter is the fundamental frequency perturbation; Shimmer is the amplitude perturbation; MPT is the longest vocalization time; I_low is the lowest sound intensity; I_high is the highest sound intensity; and DSI is the voice impairment index.
[0040] 3. Univariate logistic regression results on the relationship between each variable and patients with dysphagia after stroke.
[0041] 3.1 The results of univariate logistic regression analysis on the relationship between basic information and dysphagia after stroke are shown in Table 3. Age, gender, and presence or absence of brainstem injury may be relevant influencing factors for dysphagia after stroke (P<0.05).
[0042] Table 3. Univariate Logistic Regression Results of the Relationship between Basic Information and Patients with Post-Stroke Dysphagia
[0043] 3.2 The univariate logistic regression results of VAP analysis and the relationship between dysphagia after stroke are shown in Table 4. F0_high, MPT, I_high, DSI, Jitter, and Shimmer may be related influencing factors in dysphagia after stroke, and the differences are statistically significant (P<0.05).
[0044] Table 4. Univariate Logistic Regression Results of VAP Analysis and its Relationship with Post-Stroke Dysphagia
[0045] 4. Univariate logistic regression analysis of the relationships between variables in the training set and patients with dysphagia after stroke. Regression results.
[0046] 4.1 The results of univariate logistic regression analysis on the relationship between basic information in the training set and patients with dysphagia after stroke are shown in Table 5.
[0047] Table 5. Relationship between basic information in the training program and patients with dysphagia after stroke: univariate logistic regression results.
[0048] 4.2 The univariate logistic regression results of the relationship between VAP analysis in the training set and patients with dysphagia after stroke are shown in Table 6.
[0049] Table 6. VAP Analysis of Post-Stroke Dysphagia Patients in the Training Set
[0050] 5. Binary Logistic Regression Analysis of Dysphagia after Stroke 5.1 As shown in Table 7, dysphagia was used as the dependent variable, and factors with P < 0.10 in the basic characteristic analysis and VAP analysis (gender, age, brainstem lesion, F0_high, Jitter, Shimmer, MPT, I_high, DSI) were used as independent variables (variable assignment, see Table 1-2). A binary logistic regression analysis was performed using forward stepwise regression. The regression model was established, and the results are shown in Table 8. Brainstem lesion, MPT, I_high, and DSI were included in the regression equation. Compared with non-brainstem lesions, brainstem lesions increased the risk of dysphagia by 7.169 times. The higher the MPT, I_high, and DSI values, the lower the risk of dysphagia after stroke. For each unit increase, the risk of dysphagia decreased by 0.925 times, 0.96 times, and 0.848 times, respectively.
[0051] Table 7. Variable assignments for binary logistic regression analysis.
[0052] Table 8. Results of multivariate logistic regression analysis of dysphagia in stroke patients.
[0053] The Omnibus test for the model coefficients showed P < 0.05, indicating that the model was effective; the Hosmer-Lemmash test showed a chi-square value of 6.341 and a significance level of 0.609, indicating that the model had a good fit.
[0054] 6. Logistic regression analysis of the training set identified brainstem lesions, MPT, I-high, and DSI as independent predictors of dysphagia after stroke. Based on these results, software was used to plot... Figure 1 A nomogram model for predicting the risk of dysphagia after stroke.
[0055] The nomogram predicting the risk of dysphagia after stroke includes a scaled-down scoring axis, brainstem lesion assignment, measured values of longest vocalization time, highest vocal intensity, DSI, total score axis, and risk axis for dysphagia after stroke. Vertical lines are drawn upwards from the corresponding variable axes for brainstem lesion assignment, longest vocalization time, highest vocal intensity, and DSI. The intersection of these lines with the scoring axis represents the score for that variable. The sum of the scores for the four variables is the total score of the predictive model. A vertical line is drawn downwards from the total score axis, intersecting with the risk axis for dysphagia after stroke. This intersection represents the probability of dysphagia in stroke patients.
[0056] 6.1 Internal validation and evaluation of the nomogram model for predicting the risk of dysphagia after stroke.
[0057] See 6.1.1 Figure 2 The predictive model's discriminative power was 0.789 (95% CI: 0.726–0.852) on the training set, indicating that the nomogram model has good predictive ability for dysphagia after stroke. The constructed predictive model was internally validated and evaluated using the Bootstrap self-sampling method with 1000 iterations.
[0058] See 6.1.2 Figure 3 The model calibration degree is predicted by comparing the model calibration curve with the ideal curve in the training set. The consistent trend of the two curves indicates that the model has good prediction accuracy.
[0059] See 6.1.3 Figure 4 The clinical decision curve, as shown by the DCA curve plotted on the training set, indicates that the net benefit of the model is consistently greater than the threshold probabilities of the two extreme strategies over a wide range, suggesting its potential clinical application.
[0060] 6.2 External validation and evaluation of the nomogram model for predicting the risk of dysphagia after stroke.
[0061] See 6.2.1 Figure 5 The constructed nomogram model was validated using a validation set. The validation set AUC was 0.739 (95% CI: 0.631–0.846), indicating that the nomogram model has good predictive ability for dysphagia after stroke.
[0062] See 6.2.2 Figure 6 In the validation set, the model calibration curve and the ideal curve are basically consistent, indicating that the model has good prediction accuracy. However, compared with the calibration curve in the training set, the model prediction curve in the validation set is further away from the ideal curve, indicating that the consistency of the calibration curve in the validation set is slightly worse than that in the training set.
[0063] See 6.2.3 Figure 7 DCA in the validation queue shows that, over a wide range of threshold probabilities, the net return of this model is consistently greater than the net returns of the two extreme strategies.
[0064] Obviously, the above embodiments are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, these obvious variations or modifications derived from the spirit of the present invention are still within the scope of protection of the present invention.
Claims
1. A device for screening post-stroke dysphagia based on voice acoustic parameters, characterized in that, It includes a noise detection device, an input device, a processing device, and an output device; The noise detection device is used to collect the measured values of the patient's longest vocalization time (MPT), highest sound intensity (I_high), highest fundamental frequency (F0_high), lowest sound intensity (I_low), and fundamental frequency jitter. The input device is used to collect brainstem lesion values from patients. The processing device is connected to the noise detection device and the input device respectively; The measured values of the Voice Disorder Index (DSI) were calculated using the measured values of the longest vocalization time, the highest fundamental frequency, the lowest sound intensity, and the fundamental frequency perturbation as variables. The total score of the prediction model was calculated using the brainstem lesion assignment, the measured values of the longest vocalization time, the highest sound intensity, and the measured DSI as variables. The risk probability value of stroke patients developing dysphagia was obtained based on the total score of the prediction model. The output device is connected to the processing device and is used to output the risk probability value of swallowing dysfunction in stroke patients.
2. The device for screening post-stroke dysphagia based on voice acoustic parameters according to claim 1, characterized in that, The processing device is used to draw a nomogram predicting the risk of dysphagia after stroke, and the output device is also used to output the total score of the prediction model and the nomogram predicting the risk of dysphagia after stroke.
3. The device for screening post-stroke dysphagia based on voice acoustic parameters according to claim 1, characterized in that, The nomogram predicting the risk of dysphagia after stroke includes a scaled-down scoring axis, brainstem lesion assignment, measured values of longest vocalization time, highest vocal intensity, DSI, total score axis, and risk axis for dysphagia after stroke. Vertical lines are drawn upwards from the corresponding variable axes for brainstem lesion assignment, longest vocalization time, highest vocal intensity, and DSI. The intersection of these lines with the scoring axis represents the score for that variable. The sum of the scores for the four variables is the total score of the predictive model. A vertical line is drawn downwards from the total score axis, intersecting with the risk axis for dysphagia after stroke. This intersection represents the probability of dysphagia in stroke patients.
4. The device for screening post-stroke dysphagia based on voice acoustic parameters according to claim 1, characterized in that, The DSI in the processing device is calculated based on the formula: DSI=0.13×MPT+0.0053×F0_high-0.26×I_low-1.18×Jitter+12.4, and the actual measured value of DSI is obtained after calculation.
5. The device for screening post-stroke dysphagia based on voice acoustic parameters according to claim 1, characterized in that, The noise detection device includes a recording module and an environmental noise monitoring module. The recording module is used to collect the user's audio, and the environmental noise monitoring module is used to collect ambient sound and determine the noise intensity.
6. The method of using the device for screening post-stroke dysphagia based on voice acoustic parameters as described in claim 1, comprising the following steps: S1: The input device prompts the patient to input brainstem lesion values, gender, and age; S2: Noise detection device detects ambient noise; S3: A pop-up window prompts the user to sit upright, keeping the lower lip about 30cm away from the phone microphone, and a demonstration video is played simultaneously to help the user correct their posture; S4: The user is prompted to pronounce words via voice, and the recording module simultaneously collects audio signals; S5: The processing device calculates the risk probability value of stroke patients developing dysphagia based on the data collected by the input device and the noise detection device and transmits it to the output device. S6: The output device outputs the probability value of the risk of swallowing dysfunction in stroke patients.
7. The method for screening post-stroke dysphagia based on voice acoustic parameters according to claim 6, characterized in that, The S2 also includes a pop-up window prompting the user to change the data collection environment when the environmental noise monitoring module detects environmental noise ≥45dB.
8. The method for screening post-stroke dysphagia based on voice acoustic parameters according to claim 6, characterized in that, S4 further includes the following steps: S41: The user is prompted by voice to continuously pronounce long vowels from low to high frequencies, and the recording module simultaneously collects the audio signal and extracts F0_high; S42: The voice prompts the user to pronounce long vowels in a comfortable tone, with the volume gradually increasing from weak to strong. The recording module simultaneously collects the audio signal and extracts I_high and I_low. S43: After prompting the user to take a deep breath, the user should continuously pronounce a long vowel with a steady pitch and loudness until there is no airflow. The recording module will record the sound three times, with an interval of ≥15 seconds between each recording. The longest duration will be automatically selected as the MPT, and the signal will be extracted from the steady segment of 3-4 seconds of pronunciation to obtain the Jitter.
9. The device for screening post-stroke dysphagia based on voice acoustic parameters according to claim 1, characterized in that, The output device is integrated into the application software and is used to output the risk probability value of stroke patients developing swallowing dysfunction in the application software.