Researchers at Institute of Systems and Computer Engineering, Technology and Science (INESC TEC) led a study on the use of Artificial Intelligence (AI) and Deep Learning (DL), in the complementary diagnosis of COVID-19 by Chest radiography (CXR) through deep learning. The work was recently published in the prestigious scientific journal Nature – Scientific Reports under the scope of CMU Portugal TAMI project.
TAMI proposes to put AI at the service of the health sector, more specifically to support clinical decisions to increase confidence in medical diagnosis. The COVID-19 pandemic has impacted healthcare systems across the world and the speed of transmission made crucial a fast and early diagnosis of the disease. Chest radiography has been one of the main complementary methods for experienced radiologists to diagnose/follow COVID-19 patients, throughout the pandemic. However, the workload of qualified technicians during this period has been compromising the decision process, leading to the use of less experienced clinicians. Automated image analysis through AI techniques and deep learning can play an important role in assessing CXRs, providing a crucial second opinion for radiologists and technicians in the decision process.
Since the pandemic’s beginning, there has been a great effort from the scientific community in these ICT areas to find new approaches to support the medical diagnosis. According to Aurélio Campilho, one of the project PIs, “the goal of this research was exactly to study how Deep Learning can be placed at the service of medical diagnosis”. He adds that “we wanted to evaluate how Deep Learning could help the interpretation/reading of CXRs and support the diagnosis and follow-up of COVID-19 patients. Our study showed that the use of these algorithms in a clinical environment is much more complex than expected”. In close collaboration with ARSN (Regional Health Administration of the North) it was possible to identify the main challenges in applying these Deep Learning tools and develop new techniques that can increase the robustness of these systems.
Deep Learning is a branch of Machine Learning that provides computers with the ability to learn and perform human-like tasks, such as identifying images, recognizing speech, or making prognoses. This study evaluated the performance of a DL system in diagnosing COVID-19 by comparing it to the analysis of radiologists. One of the main conclusions is that distinguishing between COVID-19 and other pathologies on CXRs is a difficult task, even for experienced radiologists. However, it was possible to demonstrate that the performance of DL algorithms in identifying COVID-19 can be significantly improved if they learn directly from radiologists, more clearly identifying the radiological signs of COVID-19 and leading to a better diagnosis.
Although this methodology is still at an early stage, the goal is to apply this research to other pathologies identified by CXR: “Although COVID-19 has been the main focus of our research over the past two years, there are many other pathologies and findings that can be identified on CXRs. Our goal is to develop a system that can identify these automatically. A tool of this sort would be extremely useful to help radiologists, technicians, and less experienced physicians in interpreting CXRs”, concludes Aurélio Campilho.
In a broader scope, the TAMI project that is being led by First Solutions, with INESC TEC, Fraunhofer Portugal, Administração Regional de Saúde do Norte (ARS Norte) and CMU Electrical and Computer Engineering Department will develop tools to support the medical decision, based on artificial intelligence algorithms that will explain to both clinicians and researchers the diagnosis of a specific disease and its causes, focusing on cervical cancer, lung diseases and eye diseases. The project will work on a commercial, scientific, and academic platform that will provide “consumers” access to results and explanations of diagnostic orders, filtered data sets access for investigators or scientists, and a knowledge base for academic purposes.
In the Media: Expresso Online, Público, Observador, SIC Notícias Online, Sapo 24, Sapo Lifestyle, Atlas Saúde, Cidade FM Online, Smooth FM Online, Diário de Notícias da Madeira, HealthNews Online, M80 Online, Notícias ao Minuto, Rádio Comercial Online, S+ Online.
More about the project: https://cmuportugal.org/large-scale-collaborative-research-projects/tami/
