Prediction of difficult airway management cases using machine learning and image processing techniques.

Airway management poses challenges of oxygenation cannot be maintained by either face mask ventilation, using a supraglottic device, endotracheal intubation or surgical airway (e.g., tracheostomy), when patient looses consciousness or breathing is compromised due to any reason. however, there is no single predictive test to anticipate such difficulties with maximum certainity as the difficult airway is multi factorial. If a system can be developed using artificial intelligence, that is machine learning using image processing, such situations can be instantly diagnosed, anticipated, managed correctly without any damages to the patient.

Hypothesis: Incorporating the airway assessment tools with logic into an app that canaccurately predict difficult airway in unanticipated cases will help inaccurately predicting and managing difficult airway.

AIM

To develop automated system using machine learning to predict the patientswith difficulty in airway management.

OBJECTIVES

Primaryobjective:

To design anautomated system using machine learning algorithms to predict the difficulty inmanaging airway.

Secondaryobjectives:

To predict difiiculty in mask ventilation

To predict difficulty in supraglottic airway insertion

To Predict difficulty in laryngoscopy

To predict difficulty in endotracheal intubation

Compare them with actual clinical outcome

Here we are intending to recruit 250 patients undergoing elective surgeries. Record their airway parameters, take frontal and profile pictures of head and neck, record the outcomes of airway management (mask ventilation, laryngoscopy, use of supraglottic airway, endotracheal intubation etc.) then compare the outcome with the airway indices, form an algorithm to predict difficulty in managing airway, process the images to arrive at predictors of difficulty in airway management. Thus, finally using machine learning launch an app which can predict difficulty in any of the above said parameters, soon after taking photographs in above said views with clinically useful accuracy.