Recognition method, system and equipment for difficult intubation based on formant frequency change and medium

Abstract

The first aspect of the invention is to provide a method for identifying difficult intubation based on formant frequency change. The second aspect of the invention is to disclose a difficult intubation identification system based on formant frequency change. The third aspect of the invention is to provide an electronic device. The fourth aspect of the invention is to provide a computer readable storage medium, an executable instruction is stored on the computer readable storage medium, and when the instruction is executed by the processor, the processor executes the method. According to the technical scheme disclosed by the invention, when the artificial airway is established, the difference existing in normal ventilation can be prompted according to the sound difference, so that the principle of existence of some difficult ventilation can be predicted, and an artificial intelligence algorithm is combined by analyzing the acoustic characteristics of the specific voice-Chinese phonetic alphabet produced by a patient, so that the artificial airway is established. And the technical scheme of rapid non-invasive identification of the difficult airway is realized.

Description

Technical Field

[0001] This invention relates to a rapid, non-invasive method for predicting difficult intubation through voiceprint feature signal analysis, a system, device, and medium for implementing the method, belonging to the field of medical artificial intelligence. Background Technology

[0002] Difficult airway prediction and management is one of the major challenges in anesthesiology, emergency medicine, and critical care medicine. Existing methods for assessing difficult airways (such as the Mallampati classification, thyromental distance measurement, SARI scoring, Wilson scoring, etc.) can provide airway assessment and help predict difficult intubation.

[0003] The Mallampati classification is a system for assessing airway anatomy, predicting the ease of intubation based on the number of oral structures visible with the mouth fully open. The currently widely used modified Mallampati classification comprises five classes: Class 0: any part of the epiglottis is visible in the pharyngeal field; Class I: the soft palate, uvula, and palatopharyngeal pillars are visible; Class II: the soft palate and part of the uvula are visible, with the palatopharyngeal pillars obscured by the base of the tongue; Class III: only the soft palate is visible, with both the uvula and palatopharyngeal pillars obscured; Class IV: only the hard palate is visible. A higher Mallampati classification indicates fewer visible structures in the oral cavity, potentially leading to greater intubation difficulty. This classification system is widely used in clinical practice to predict difficult intubation, especially in patients under general anesthesia or for pre-intubation assessment. However, it is important to note that the Mallampati classification alone has limited accuracy in predicting difficult airways; it is primarily used as an adjunct tool, and its diagnostic value is higher when combined with other clinical assessments.

[0004] The SARI scoring method includes seven variables: mouth opening, thyromental distance, Mallampati classification, neck mobility, mandibular advancement ability, body weight, and history of difficult intubation. The total score ranges from 0 to 12, with each variable assigned a score of 0-1 or 0-2. The total score is used to predict the likelihood of difficult intubation.

[0005] Wilson's scoring method includes five variables: body weight, head and neck mobility, mandibular mobility, mandibular retraction, and incisor protrusion. Each variable is assigned a score of 0, 1, or 2, for a total score of 10. The higher the score, the greater the likelihood of difficult endotracheal intubation.

[0006] The MACOCHA scoring method includes patient-related factors (Mallampati classification III or IV, obstructive sleep apnea, limited cervical spine mobility, limited mouth opening), pathophysiological factors (coma, severe hypoxemia <80%), and operator factors (non-anesthesiologist operation). The total score is 0-12 points, with each factor assigned a value of 1, 2, or 5 points. A score ≥3 points predicts difficulty in intubation in critically ill patients.

[0007] Arne scoring method: includes airway disease, history of difficult intubation, diseases related to difficult intubation, mouth opening and temporomandibular joint mobility, thyromental distance, head and neck mobility, and modified Mallampati classification, totaling 48 points. A total score >11 points indicates difficult intubation.

[0008] Upper lip bite test (ULBT): Assess the likelihood of airway difficulty by the degree to which the patient bites their upper lip.

[0009] Thyroid-mental distance (TMD): The distance from the thyroid cartilage to the tip of the mandible when the head is in an extended position. These distances are measured to predict difficult airways.

[0010] Imaging methods, such as CT, X-ray, MRI, and ultrasound, can provide visual assessment of the airway and help predict difficult airways.

[0011] The aforementioned traditional difficult airway assessment tools and methods can be used individually or in combination to improve prediction accuracy. In practice, physicians will choose the most appropriate predictive tool based on the patient's specific condition and available resources. Each method has its limitations, some of which cannot be avoided even with multiple methods. Inadequate assessment and prediction failure can have serious consequences. The specific limitations of traditional difficult airway assessment tools and methods are as follows: 1) Low accuracy: Traditional physical examinations, such as Mallampati classification, thyromental distance (TMD), and mouth opening, are simple and easy to perform, but mainly focus on the airway in front of the tongue base. They cannot accurately reflect the deeper abnormal anatomical structures inside the airway, such as epiglottis and tonsil abnormalities, resulting in low accuracy in predicting difficult airways.

[0012] 2) Reliance on subjective judgment: Many assessment methods rely on the subjective judgment of clinicians, such as the Mallampati classification. Due to the influence of patients’ differences in mouth opening, this may lead to inconsistencies and reliability issues in the assessment results.

[0013] 3) Limited predictive ability: Traditional assessment methods have limited ability to predict difficult airways, with positive predictive value (PPV) lower than the actual value; another issue is that the predicted airway may not actually be a difficult airway. This means that the assessment result may differ from the probability of matching the actual difficult airway.

[0014] 4) Lack of dynamic assessment: Traditional assessment methods are mostly static assessments, which cannot provide information on dynamic changes in the airway. However, dynamic changes in the airway are crucial for predicting and managing difficult airways and directly affect ventilation. In other words, ventilation status under different physiological and pathological conditions cannot be detected by these conventional methods.

[0015] 5) Equipment dependence and radiation risks: Some auxiliary examination methods, such as X-ray and CT scans, can provide more information about airway structure, but they are equipment-dependent and pose radiation risks.

[0016] 6) Cost issues: Some advanced imaging assessment methods, such as CT scans and MRI, are expensive and may not be suitable for routine screening.

[0017] 7) Sample size and study quality limitations: Many studies on difficult airway assessment have small sample sizes and are mostly non-randomized, open-label case-control studies, which may lead to bias in the results.

[0018] 8) Technological limitations: Although virtual laryngoscopes and 3D printing technology provide a three-dimensional view of the airway, these technologies have only been studied in small sample patients, the evidence is insufficient, and the cost is high, so they cannot be used as routine screening methods.

[0019] 9) Lack of comprehensive evaluation: Traditional evaluation methods often consider a single factor in isolation without taking into account the combined effect of multiple factors, which may limit the comprehensiveness and accuracy of the evaluation.

[0020] 10) Limited applicability to special cases: In some special cases, such as obesity, neck deformity, and scar adhesion after burns, traditional assessment methods may not be applicable or effective.

[0021] In summary, existing methods for assessing difficult airways mainly rely on surface anatomical indicators or subjective experience, resulting in insufficient accuracy and sensitivity, and often leading to unforeseen difficult intubation. Especially in specific populations (such as obese patients, patients with obstructive sleep apnea, and patients with temporomandibular joint dysfunction), while adding examinations such as CT, X-ray, MRI, and ultrasound can help improve the detection rate of difficult airways, these examinations involve radiation hazards, significant additional costs, and limited examination sites, making them unsuitable for routine use. Therefore, existing methods are prone to inadequate assessment and predictive failure, which can have serious consequences.

[0022] Recently, a new method for predicting difficult airways has emerged, based on patient facial features and big data analytics using artificial intelligence. This method predicts difficult airways by taking facial photographs of patients from different angles. However, it also has some limitations in difficult airway assessment, mainly in the following aspects: Significant individual differences exist: Facial features vary considerably among individuals, and even patients with similar facial appearances may have vastly different airway conditions. For example, patients with similarly small jaws may have only mild airway restriction in some cases, while others may have severe airway difficulties. Relying solely on facial features makes it difficult to accurately quantify these differences, thus affecting the accurate assessment of difficult airways.

[0023] Lack of dynamic assessment capability: Assessments are mostly static observations conducted when the patient is at rest, which cannot reflect the dynamic changes in the airway during anesthesia induction or changes in surgical position. For example, after anesthesia induction, muscle relaxation may cause changes in the position and tension of airway tissues, and what may have seemed like a normal airway may present unexpected difficulties. In addition, the airway is an active process for ventilation and does not reflect differences during ventilation.

[0024] Other factors can interfere with the accuracy of difficult airway assessment based on facial features. For example, facial fat deposits in obese patients may obscure the true jaw structure, leading assessors to misjudge the airway condition; swelling and deformities in patients with facial trauma can also make facial feature assessment inaccurate, thus failing to accurately assess the degree of difficult airway.

[0025] High accuracy is crucial for predicting difficult airways; otherwise, unexpected airway difficulties can lead to fatal risks. Traditional assessment methods rely heavily on facial features, such as the Mallampati classification, which, while valuable and primarily focused on oral and pharyngeal structures, cannot effectively assess the condition of the airway below the pharynx or its compliance, risking the omission of crucial information. Other indicators also have limitations, failing to comprehensively cover all factors that might influence airway difficulty; furthermore, they are primarily static assessments and heavily influenced by the subjective experience of medical personnel. Overall, these factors contribute to an incomplete and inaccurate assessment of difficult airways. Summary of the Invention

[0026] The technical problem to be solved by this invention is that in recent years, the development of speech signal analysis technology and artificial intelligence has provided new ideas for the assessment of difficult airways. Speech signals can reflect the structure and vocal function of the vocal cords, glottis and airway, airflow exhalation obstruction, and even lower respiratory tract diseases. However, there is currently no technical means for rapid assessment of difficult airways based on vocal features.

[0027] To address the aforementioned technical problems, the first aspect of this invention provides a method for identifying difficult cannulation based on resonant peak frequency changes, specifically including the following steps: Step 1: Obtain the standardized speech signal emitted by the patient to be identified. The standardized speech signal is the speech signal emitted by the patient when uttering a set pinyin or text. Step 2: Preprocess the standardized speech signal obtained in Step 1, including noise reduction, standardization, and segmentation; Step 3: Extract the acoustic features of the preprocessed standardized speech signal; Step 4: Use the trained AI-based classification model to analyze the extracted acoustic features and predict whether the patient has a difficult airway. Preferably, in step 1, when acquiring the standardized speech signal emitted by the patient to be identified, the patient is in a quiet environment.

[0028] Preferably, in step 3, the extracted acoustic features include fundamental frequency, formants, energy distribution, time-frequency characteristics, and spectral envelope.

[0029] Preferably, step 3 includes the following steps: Extract the target parameters from the preprocessed standardized speech signal; Mark the position of each syllable; The mean value of each target parameter corresponding to all syllables is calculated as the acoustic feature.

[0030] Preferably, in step 1, in addition to obtaining the standardized speech signal, demographic indicators and bedside airway assessment indicators are also obtained; in step 4, the demographic indicators and bedside airway assessment indicators, together with the acoustic features, are used as input variables for the artificial intelligence-based classification model.

[0031] Preferably, in step 4, when training the AI-based classification model, standardized speech signals from patients with difficult airways and those with non-difficult airways are used as samples to construct a training set and a validation / test set for model training. In constructing the training and validation sets, missing values ​​in the training set are imputed using multiple imputation and marked with missing values, while the validation / test set is not imputed to prevent data leakage. During training, if the positive rate of the training set is <15%, class weighting is used, and SMOTE oversampling is applied to the training set, while the validation / test set remains unchanged.

[0032] The second aspect of the present invention discloses a system for identifying difficult intubation based on formant frequency changes, used to implement the above-mentioned method for identifying difficult intubation based on formant frequency changes. The system is characterized by comprising a data acquisition module for implementing the standardized speech signal described in step 1, a data preprocessing module for the preprocessing described in step 2, a data feature extraction module for implementing the feature extraction described in step 3, a difficult airway prediction module for predicting whether a patient has a difficult airway described in step 4, and a result output module for outputting the prediction results of the difficult airway prediction module.

[0033] Preferably, the system for identifying difficult intubation based on resonant peak frequency changes is integrated into a portable device or mobile application to be suitable for real-time bedside assessment.

[0034] A third aspect of the present invention is to provide an electronic device comprising: One or more processors; A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute software to implement the method described above.

[0035] A fourth aspect of the present invention is to provide a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to perform the method described above.

[0036] The technical solution disclosed in this invention, when establishing an artificial airway, can indicate differences in normal ventilation based on differences in sound, thereby predicting the existence of certain difficult intubations. It provides a technical solution for rapid, non-invasive identification of difficult airways by analyzing the acoustic characteristics of specific patient speech—specifically, Chinese Pinyin—combined with artificial intelligence algorithms. Compared with existing technical solutions, the specific advantages are as follows: Fast: The entire process can be completed in minutes, significantly improving evaluation efficiency; Non-invasive: No need to rely on invasive examinations, resulting in greater patient comfort; Objective: Reduce human error through data-driven artificial intelligence models; Wide applicability: Suitable for pre-anesthesia assessment, critical and emergency rescue scenarios, and out-of-hospital accident rescue. Attached Figure Description

[0037] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0038] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0039] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0040] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0041] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0042] Detailed background technology may include other technical issues besides those that can be solved independently.

[0043] The human vocal organs are the same passageways used for normal breathing. Sound is formed by airflow pushing the vocal cords to vibrate, which then travels through the airways on the vocal cords to produce different sounds. Changes in any of these three factors can alter the characteristics of the sound. Since sound production and propagation occur through the same airway as breathing, based on the principle of sound production (airflow exhaled from the lower airway impacts the vocal cords and is then emitted through the vocal apparatus), we believe that airway characteristics are related to sound characteristics. For example, each person's voice has its own characteristics because each person's airway characteristics are different. Similarly, difficult intubation corresponds to airway anatomical features. By analyzing the corresponding speech patterns with similar static airway shapes, we can help predict difficult airways. Multiple speech patterns corresponding to multiple features will make it easier to identify difficult intubation cases that conventional methods have not anticipated.

[0044] Based on the above understanding, the first aspect of the present invention discloses a method for identifying difficult cannulation based on resonant peak frequency changes, specifically including the following steps: Step 1: Obtain the standardized speech signal emitted by the patient to be identified. The so-called standardized speech signal refers to the speech signal emitted by the patient when he or she emits a set pinyin or text. For example, it can be the speech signal emitted by the patient when he or she emits the pinyin: a, e, i, o, u.

[0045] In a preferred embodiment of the present invention, the patient is in a quiet environment when the standardized speech signal emitted by the patient is acquired.

[0046] In another preferred embodiment of the present invention, a standardized speech signal emitted by the patient is acquired using a high-sensitivity microphone.

[0047] Step 2: Preprocess the standardized speech signal obtained in Step 1, including but not limited to noise reduction, standardization and segmentation.

[0048] Step 3: Extract the acoustic features of the preprocessed standardized speech signal, including but not limited to fundamental frequency (F0), formants (F1, F2, f3), energy distribution, time-frequency characteristics, and spectral envelope.

[0049] In one preferred embodiment of the invention, the Praat speech analysis software is used to extract key parameters such as the fundamental frequency, formant frequencies (F1-F4), and bandwidth (bw1-bw4) of the preprocessed standardized speech signal. After accurately marking the position of each syllable using automatic speech recognition software (such as Kaldi), the mean value of each parameter is calculated as the basis for analysis. Furthermore, in another preferred embodiment of the invention, after obtaining the key parameters, outliers are truncated according to a preset rule.

[0050] Step 4: Use the trained AI-based classification model to analyze the extracted acoustic features and predict whether the patient has a difficult airway.

[0051] In a preferred embodiment of the present invention, demographic indicators and bedside airway assessment indicators (e.g., Mallampati classification, thyromental distance) can also be obtained and used together with acoustic features such as fundamental frequency, harmonic-to-noise ratio, and MFCC as input variables for an artificial intelligence-based classification model.

[0052] In another preferred embodiment of the present invention, an Elastic Net logistic regression model can be used as the main model to build an AI-based classification model, with a gradient booster (XGBoost / LightGBM) model as a backup model. Alternatively, a deep learning algorithm (such as CNN or LSTM) can be used to build an AI-based classification model, and a deep learning framework (such as TensorFlow or PyTorch) can be used to train the CNN or LSTM model to classify features.

[0053] In another preferred embodiment of the present invention, standardized speech signals from patients with difficult airways and those with non-difficult airways are used as samples to construct training and validation / test sets for model training. When constructing the training and validation sets, the sample signals are processed using the same methods as steps 1, 2, and 3 to obtain sample features. Missing values ​​in the training set are imputed using multiple imputation (MICE) and a missing value marker is added, while no imputation is performed on the validation / test sets to prevent data leakage. Furthermore, in another preferred embodiment of the present invention, if the positive rate of the training set is <15%, class weighting is primarily used during training, and SMOTE oversampling is applied to the training set when necessary. The validation / test sets remain unchanged. Furthermore, in another preferred embodiment of the present invention, before constructing the training and validation sets, feature engineering is performed on the extracted sample features to remove highly correlated and low-variance features, and feature selection is performed within the training set. Continuous predictive variables are all standardized using Z-scores. In another preferred embodiment of the present invention, when validating the classification model, the training set and the reserved test set are randomly divided at 80% / 20%, and nested 5×5-fold cross-validation is used for hyperparameter optimization and internal validation.

[0054] In another preferred embodiment of the present invention, the area under the ROC curve (AUC) is used to evaluate the performance of the established artificial intelligence-based classification model. At the same time, the sensitivity, specificity, F1 score and PR-AUC are reported to evaluate the model calibration (Brier score, calibration curve) and clinical applicability (decision curve analysis).

[0055] In another preferred embodiment of the present invention, the classification model is analyzed using R (tidymodels / caret) or Python (scikit-learn, xgboost), with a fixed random seed to ensure reproducibility of the results.

[0056] A second aspect of this invention discloses a system for identifying difficult intubation based on resonant peak frequency changes, used to implement the aforementioned method for identifying difficult intubation based on resonant peak frequency changes, such as... Figure 1 As shown, it includes a data acquisition module for implementing the standardized speech signal described in step 1 above, a data preprocessing module for implementing the preprocessing described in step 2 above, a data feature extraction module for implementing the feature extraction described in step 3 above, a difficult airway prediction module for predicting whether a patient has a difficult airway described in step 4 above, and a result output module for outputting the prediction results of the difficult airway prediction module.

[0057] In another preferred embodiment of the present invention, the data acquisition module is also used to collect demographic indicators and bedside airway assessment indicators.

[0058] A third aspect of this invention discloses an electronic device including a processor capable of performing various appropriate actions and processes based on a program stored in a read-only memory (ROM) or a program loaded from a storage portion into a random access memory (RAM). The processor may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor may also include onboard memory for caching purposes. The processor may include a single processing unit or multiple processing units for performing different actions of the difficult cannulation identification method based on resonant frequency variations according to embodiments of this disclosure.

[0059] The RAM stores various programs and data required for the operation of the electronic device. The processor, ROM, and RAM are interconnected via a bus. The processor implements the aforementioned method for identifying difficult tube insertion based on resonant frequency variations by executing programs in the ROM and / or RAM. It should be noted that the programs may also be stored in one or more memories other than ROM and RAM. The processor may also perform various operations of the method according to embodiments of this disclosure by executing programs stored in said one or more memories.

[0060] According to embodiments of this disclosure, the electronic device may further include an input / output (I / O) interface, which is also connected to a bus. The electronic device may also include one or more of the following components connected to the I / O interface: an input section including a keyboard, mouse, etc.; an output section including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section including a hard disk, etc.; and a communication section including a network interface card such as a LAN card, modem, etc. The communication section performs communication processing via a network such as the Internet. A drive is also connected to the I / O interface as needed. Removable media, such as a magnetic disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on the drive as needed so that computer programs read from them can be installed into the storage section as needed.

[0061] A fourth aspect of the present invention also provides a computer-readable storage medium, which may be included in the device / apparatus described in the above embodiments; or it may exist independently and not assembled into the device / apparatus. The computer-readable storage medium carries one or more programs, which, when executed, implement the above-described method for identifying difficult cannulation based on resonant frequency changes.

[0062] According to embodiments of this disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In embodiments of the present invention, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this disclosure, the computer-readable storage medium may include ROM 402 and / or RAM 403 and / or one or more memories other than ROM 402 and RAM 403 described above.

[0063] This invention provides an innovative solution for non-invasive identification of difficult airways by combining speech signal analysis with artificial intelligence algorithms, filling a gap in existing technology and having significant clinical and social value.

[0064] Those skilled in the art will understand that the features described in the various embodiments and / or claims of this invention can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in this disclosure. In particular, the features described in the various embodiments and / or claims of this invention can be combined or combined in various ways without departing from the spirit and teachings of this disclosure. All such combinations and / or combinations fall within the scope of this disclosure.

[0065] The embodiments of the present invention have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of this disclosure. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. The scope of the present invention is defined by the appended claims and their equivalents. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of the present invention, and all such substitutions and modifications should fall within the scope of this disclosure.

Claims

1. A method for recognizing difficult intubation based on changes in resonant peak frequency, characterized by, The method comprises the following steps: Step 1: obtaining a standardized voice signal emitted by a patient to be identified, the standardized voice signal being a voice signal emitted by the patient when pronouncing a set pinyin or character; Step 2: pre-processing the standardized voice signal obtained in Step 1, including noise reduction, standardization and segmentation; Step 3: extracting acoustic features of the pre-processed standardized voice signal; Step 4: analyzing the extracted acoustic features using a trained artificial intelligence-based classification model to predict whether the patient has a difficult airway.

2. The method for identifying difficult intubation based on the change in resonant peak frequency according to claim 1, wherein, In Step 1, when obtaining the standardized voice signal emitted by the patient to be identified, the patient is in a quiet environment.

3. The method of claim 1, wherein the method is based on the change in the resonant peak frequency to identify the difficult intubation. In Step 3, the extracted acoustic features include fundamental frequency, formant, energy distribution, time-frequency characteristics and spectral envelope.

4. The method of claim 1, wherein the method is based on the change in the resonant peak frequency to identify the difficult intubation. Step 3 comprises the following steps: extracting target parameters of the pre-processed standardized voice signal; marking the position of each syllable; calculating the mean value of each target parameter corresponding to all syllables as the acoustic feature.

5. The method of claim 1, wherein the method is based on the change in the resonant peak frequency to identify the difficult intubation. In Step 1, in addition to obtaining the standardized voice signal, demographic indicators and bedside airway assessment indicators are also obtained; in Step 4, the demographic indicators and bedside airway assessment indicators are used as input variables of the artificial intelligence-based classification model together with the acoustic features.

6. The method of identifying difficult intubations based on changes in resonant peak frequency of claim 1, wherein, In Step 4, when training the artificial intelligence-based classification model, the standardized voice signals of patients with difficult airways and the standardized voice signals of patients without difficult airways are used as samples to construct a training set and a validation / test set for model training; the training set is constructed by using multiple imputation to fill in missing values and adding a missing flag, while the validation / test set is not imputed to prevent data leakage; during training, if the positive rate of the training set is < 15%, class weighting is used in training, and SMOTE oversampling is applied in the training set, while the validation / test set remains unchanged.

7. A system for identifying difficult intubation based on changes in formant frequencies for implementing the method for identifying difficult intubation based on changes in formant frequencies as claimed in claim 1, characterized in that, The system comprises a data acquisition module for obtaining the standardized voice signal as described in Step 1, a data preprocessing module for pre-processing as described in Step 2, a data feature extraction module for feature extraction as described in Step 3, a difficult airway prediction module for predicting whether the patient has a difficult airway as described in Step 4, and a result output module for outputting the prediction result of the difficult airway prediction module.

8. A system for identifying difficult intubations based on changes in resonant peak frequency as claimed in claim 7, wherein, The difficult intubation recognition system based on formant frequency change is integrated through a portable device or a mobile application to be suitable for bedside real-time assessment.

9. An electronic device comprising: One or more processors; A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute a software implementation of the method according to claim 1.

10. A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to perform the method according to claim 1.

Images