Automatic detection of difficult tracheal intubation

Abstract

Before the routine anesthesia, an airway examination must be performed during the pre-anesthetic examination for all patients who need a surgical operation in order to decide whether the tracheal intubation is easy or hard. In the field of anesthesia and intensive care, many works have been performed in order to reduce as much as possible the anesthetic risks and the mortality rate as well as to provide assistance to the Doctors Specialized in Anesthesia (DSA's).In this work, we suggest a system of Computer Aided Diagnosis for the DSA's during the anesthetic examination. This system was conceived to determine whether the patient's tracheal intubation is easy or hard. For this we tested four different classifiers which are the Multi-Layer Perceptron (MLP), the C4.5 decision tree, the Support Vector Machine (SVM) and the K-Nearest Neighbors algorithm (KNN) applied on a new database which was collected locally.We obtained promising results which are a proof of the reliability and the coherence of our database.

A new decision support model for preanesthetic evaluation

“…For the prediction of the decision of accepting or refusing a patient for surgery, we observe that MR-Sort yields results that are of similar quality, in terms of accuracy, as the five individual machine learning algorithms used in [23]. However, making the decision advocated by the majority of these five algorithms leads to a slightly superior prediction accuracy.…”

“…Several supervised machine learning algorithms were tested in [23] in order to predict the ASA score. Among them, the Support Vector Machine (SVM) algorithm yielded the best predictions.…”

“…In [23,24], for instance, several machine learning algorithms were used to predict the ASA score and other preoperative assessments. A drawback of machine learning methods is that they are not easily interpretable by doctors.…”