Use of artificial intelligence in pulmonary embolismprediction

Introduction and objective: The aim of this study was to predict the risk of PE in patients with suspected pulmonary embolism admitted to the emergency department using artificial intelligence by using physical examination, laboratory and clinical probability prediction scores without computed tomography pulmonary angiography. Materials and methods: A total of 156 patients admitted to the emergency department with PE were prospectively analyzed. In this study, 78 patients were diagnosed with PE based on anamnesis, physical examination, clinical probability prediction scores, investigations and imaging and these patients were enrolled in the PE group. The 78 patients with alternative diagnoses after excluding PE were enrolled as the control group. A patient follow-up form was created for the patients included in the study. Patient information was recorded on this form. In the medical records at the time of admission, admission number, name-surname, gender, age, shock index, vital signs (temperature, pulse rate, systolic and diastolic blood pressures, saturation values), complaints at presentation to the emergency room, comorbidities, medications used, medical history, radiologic examinations, presence of deep vein thrombosis, electrocardiography, echocardiography findings, wells score, geneva score, perc score and laboratory tests performed were analyzed. Results: A total of 156 patients were included in our study. The 156 patients admitted to the emergency department were divided into two groups as PE group and control group. The PE group consisted of 78 patients who were diagnosed with PE by computed tomography pulmonary angiography and the control group consisted of 78 patients who had PE ruled out by BTPA and received alternative diagnoses.The mean age of the patients included in the study was 69.46±15 years.The most common presentation was dyspnea in 88 patients (56.4%). The most common comorbidities were hypertension in 52 patients (33.1%), malignancy in 51 patients (32.7%), and coronary artery disease in 35 patients (22.4%). Wells score, D-dimer, low partial carbon dioxide pressure and tachycardia were found to be significant parameters in the diagnosis of pulmonary embolism. Statistically significant parameters were studied with a multilayer perceptron artificial intelligence model. The diagnosis of pulmonary embolism was correctly predicted with 96% accuracy and 89% selectivity. Conclusion: As a result of our study, it was determined that a good evaluation of anamnesis, physical examination, laboratory and imaging findings of the patients and the use of scores are important in the diagnosis of PE. In addition, it was determined that AI can be used in the diagnosis of PE and before imaging methods are requested. Keywords: Artificial İntelligence, Diagnostic Algorithm, Pulmonary Embolism