An SVM intelligent system for pre-anesthetic examination

Anesthesia is a branch of medical science generally applied to patients who need surgery or painful acts. Research in this field has brought many changes by decreasing the mortality rate that is why in this work, we propose a computer aided diagnosis system based on Support Vector Machines (SVM) aiming to help doctors in the pre-anesthetic examination. For that, a new database has been obtained with the help of the Doctors Specialized in Anesthesia (DSA). The 1275 patients in this database were selected from different private clinics and hospitals of western Algeria to evaluate our classifier. The medical records collected from patients suffering from a variety of diseases ensure the generalization of the performance of the decision system. The proposed method includes four steps, each of which corresponding to a specific classification based framework. The first one is devoted to an automatic detection of ASA (American Society of Anesthesiologists) scores. In the second step, a decision making process is applied in order to accept or refuse the patient for surgery. The aim of the third step is to choose the best anesthetic technique for the patient, either general or local anesthesia. The final stage examines if the patient´s tracheal intubation is easy or hard. We used a majority voting method between the three SVMproposed classifiers for each step in order to insure the best possible results. The classification results obtained by using our system prove the reliability and the coherence of the proposed approach for our database.