Research on a Machine Learning-Based Predictive Model for Difficult Intubation Using Specific Vocal Characteristics

Research on a Machine Learning-Based Predictive Model for Difficult Intubation Using Specific Vocal Characteristics

1. Study Purpose This research aims to develop a novel, non-invasive, and simple method to predict difficult intubation before surgery. The core idea is that the anatomy of a person's throat and mouth, which determines the ease of intubation, also uniquely shapes their voice.

   By analyzing the acoustic features of specific vowel sounds using machine learning, we seek to identify voice patterns associated with difficult airways. The ultimate goal is to create a tool that allows for a quick, painless pre-operative risk assessment, enhancing patient safety by better preparing anesthesiologists.

2. Study Design This is a prospective, observational, single-center study. It is purely observational and does not involve any changes to standard medical care or anesthesia procedures.

3. Participants We plan to enroll 300 patients. Who can join: Patients aged 15-70 scheduled for elective surgery requiring general anesthesia with tracheal intubation. Who cannot join: Individuals with speech/hearing impairments, significant neurological diseases affecting speech, or conditions contraindicating standard laryngoscopy.

4. Study Procedures For participants, the study involves one key procedure in addition to standard care:Voice Recording: Before surgery, participants will be asked to lie down and pronounce the vowels "a," "e," and "i" steadily for 1-2 seconds. This will be done twice: once with the head in a normal position and once with the head tilted back. A high-quality recorder will capture the sounds. This process is painless and takes only a few minutes. Standard anesthesia and intubation will then proceed as usual. The anesthesiologist will record the laryngeal view obtained during intubation, which will be used to classify the case as "difficult" or "non-difficult" for analysis.

5. Data Analysis The primary goal is to determine if there are statistically significant differences in the key voice resonance frequencies (F1, F2, F3) between the difficult and non-difficult intubation groups. Advanced machine learning models will be built to create the predictive algorithm.

6. Risks and Benefits Benefits: There is no direct medical benefit to participants. The contribution is to future medical knowledge and patient safety.

   Risks: The study involves minimal risk. The voice recording is non-invasive and safe. The main risk is the potential loss of confidentiality, which is mitigated by strict data protection protocols.

7. Confidentiality & Ethics All patient data will be de-identified and stored securely. The study protocol and informed consent form have been approved by the Institutional Ethics Committee of Shanghai Sixth People's Hospital. Participation is voluntary, and participants may withdraw at any time without affecting their medical care. Written informed consent will be obtained from every participant before any study procedures.

Inclusion Criteria:

• Patients scheduled for elective general anesthesia with tracheal intubation.

Exclusion Criteria:

• Patients with speech and pronunciation disorders (vocal cord diseases, cleft palate, craniofacial deformities, extensive tooth defects, cleft lip).
• Hearing impairment (difficulty in communication, inaccurate repetition).
• Patients with central nervous system diseases causing significant pronunciation abnormalities.
• Patients contraindicated for direct laryngoscopy (e.g., post-chemotherapy in the pharynx, diseases prone to mucosal bleeding, etc.) that affect group allocation, as well as those unable to complete the recording or temporarily reassigned to other anesthesia methods.