Clinical application of artificial intelligence tools to diagnostic aid in simple chest radiography images

• Grant number: 19/22577-4
• Support Opportunities: Research Grants - Young Investigators Grants
• Start date: November 01, 2021
• End date: October 31, 2026
• Field of knowledge: Health Sciences - Medicine - Medical Radiology
• Principal Investigator: Marcel Koenigkam Santos
• Grantee: Marcel Koenigkam Santos
• Host Institution: Faculdade de Medicina de Ribeirão Preto (FMRP). Universidade de São Paulo (USP). Ribeirão Preto , SP, Brazil
• Associated researchers: Elcio dos Santos Oliveira Vianna ; João Paulo Papa ; Julio Cesar Nather Junior ; Lucas Lins de Lima ; Maurício Koenigkam Santos ; Natália Santana Chiari Correia ; Paulo Mazzoncini de Azevedo Marques ; Renato Tinós ; Valdes Roberto Bollela
• Associated scholarship(s): 23/12569-0 - Analysis of plain chest x-ray images of patients with fungal infections and lung cancer using artificial intelligence tools applied to diagnostic aids., BP.IC; 23/09104-5 - Organization of the database of patients with fungal infections and lung cancer to assist in the development of artificial intelligence tools applied to aid diagnosis in simple chest X-ray images, BP.IC

Abstract

Simple chest radiography (X-ray) represents, in most situations, the first imaging exam to assess thoracic and pulmonary diseases. Most bronchopulmonary diseases are present on X-rays as opacities, such as tuberculosis, fungal infections, interstitial lung diseases and lung cancer. These diseases have high morbidity and mortality and require further investigation, with computed tomography (CT) and / or laboratory tests. Despite being considered a simple exam, accurate interpretation of chest X-rays requires experience. In addition, there is often an absence of specialist professionals at the site of the examination, or they are overwhelmed by the large volume of examinations. For this reason, in recent years, computed aided diagnosis / detection (CAD) tools have been created to help detect changes in imaging tests, interpret findings and optimize the workflow at health services. To this end, the development of artificial intelligence (AI), machine learning and deep learning tools has been highlighted. In this way, the objective of this study is to evaluate the clinical application of CAD tools developed in our institution, with AI resources using deep learning and convolutional neural networks, to aid diagnosis in chest X-ray images, with a focus on the detection and classification of the main pulmonary diseases that course with radiographic opacities and require further investigation. These tools are expected to help identify pulmonary opacities, suggest the probable diagnosis of the disease and recognize different patterns or classes of diseases assessed. (AU)