Artificial intelligence lung ultrasound interpretation comparison with novice to advanced clinician interpretation

• Grant number: 18/09356-6
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): November 01, 2018
• Effective date (End): July 31, 2019
• Field of knowledge: Health Sciences - Medicine - Medical Clinics
• Principal Investigator: Thiago Martins Santos
• Grantee: Amanda Kaori Ito
• Host Institution: Faculdade de Ciências Médicas (FCM). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil

Abstract

Lung ultrasound is a clinician only application which dates back almost 30 years. However, its importance and utility were clearly established only in the last decade. Although traditional radiology has relatively restricted applications for its use, other areas of medical knowledge have been using lung ultrasound for the evaluation of patients at the bedside. The name point of care ultrasonography is given to this modality of simplified evaluation and based on a specific clinical question. The examination is done at the bedside by the attending physician, and the findings may help in the interpretation of several pathological processes, both pulmonary and extrapulmonary. Because it has no ionizing radiation, hand held portability and greatly lower cost than plane chest x-ray, ultrasound is the only feasible imaging modality at the bedside in most locations around the world. . However, while lung ultrasound is not a complex application, experience is required to accurately interpret findings and integrate them into the clinical decision making process at the patient's bedside. This limitation can be overcome with automation which would allow providers to use ultrasound with minimal (if any) ultrasound or even medical training. Goals: To compare the diagnostic precision of artificial intelligence with the novice to advanced clinician interpretation, from pulmonary ultrasound images obtained in patients attended at the emergency and intensive care units of HC UNICAMP. Methods: Patients will be selected from the Emergency and Intensive Care Unit of HC UNICAMP. Pulmonary ultrasound images obtained will be analyzed by three groups: physiotherapists involved in the pulmonary ultrasound training project ("Training to perform pulmonary ultrasonography and interpretation of images by physiotherapists working in an Intensive Care Unit"); medical professionals and by the algorithm of the artificial intelligence of the ultrasound machine developed by the company EchoNous. A comparison of the analysis of the three groups (from the beginner to the clinician) will be carried out based on the evaluation of parameters established by the BLUE (Bedside Lung Ultrasound in Emergency) and SLESS (Simplified Lung Edema Scoring System) protocols. Expected Outcomes: We expect that the diagnostic accuracy of artificial intelligence is similar to medical professionals, from the novice to the advanced clinician.