In the continuous fight against COVID-19, artificial intelligence (AI) could be of great assistance.
X-ray. Image Credit: Carnegie Mellon Portugal Program (CMU Portugal).
A study led by a team of Portuguese scientists focused on building a COVID-19 complementary diagnosis tool by using artificial intelligence (AI) and deep learning methods for examining chest radiography (also known as X-Rays).
The goal of this research was to study how Deep Learning can be used to help medical diagnosis.
Aurélio Campilho, Principal Investigator, TAMI Project, Carnegie Mellon Portugal Program
It is said that experience is a vital factor for an accurate diagnosis, but skilled professionals are not available at all times to help deduce examinations. This is exactly where Transparent Artificial Medical Intelligence (TAMI) will be more beneficial.
Nearly two and a half years have gone by since the COVID-19 pandemic first started. The virus SARS-CoV-2 was first detected in the city of Wuhan, China in December 2019. In January 2020, the WHO declared it a “Public Health Emergency of International Concern” and called it a pandemic two months later.
From that moment, the world has grappled against time, and technological and scientific advances have focused on containing the pandemic either by restricting the spread or by preventing critical health problems, such as the anticipated vaccines.
At present, over 5 billion people (65.7% of the world population) have been administered at least one vaccine dose, but COVID-19 has not been eradicated. In reality, some European nations like Portugal, are still facing increasing numbers of infections and deaths associated with the disease.
In Portugal, as in other nations, scientists led a study that focused on developing a COVID-19 complementary diagnosis tool by using AI and Deep Learning methods for examining chest radiography. X-Ray examinations are recommended to help diagnose and follow progress on numerous health problems, including contagions by SARS-CoV-2.
The solution – TAMI – offered by the Portuguese researchers was recently reported in Scientific Reports – Nature, along with the details of the project’s findings.
Our study shows that the application of algorithms in clinical environments was more complex than what we anticipated. In collaboration with our local health administration (ARSN), we successfully identified the main challenges of using Deep Learning tools, which allowed us to develop new techniques and increase the robustness of these systems.
Aurélio Campilho, Principal Investigator, TAMI Project, Carnegie Mellon Portugal Program
Even for skilled experts, it is harder to differentiate between the X-Rays of a patient with COVID-19 and another patient with other complications, which in the long run makes it hard to create an AI model.
However, the researchers also understood that using data directly from radiologists significantly improves the performance of deep learning algorithms for COVID-19 detection.
Although the solution is still not completely available, this study has made way for AI and deep learning to be used for the diagnosis of other complications.
Even though COVID-19 has been very important in the last couple of years, there are many pathologies that are diagnosed using X-rays, and our goal is to develop a system that can identify them automatically, which would be very useful for doctors and radiologists who are less experienced in x-ray interpretation.
Aurélio Campilho, Principal Investigator, TAMI Project, Carnegie Mellon Portugal Program
The TAMI project’s efforts are being promoted within the scope of the Carnegie Mellon Portugal Program (CMU Portugal). TAMI’s team aims to use AI to enhance the health sector by creating decision-supporting tools for the clinical diagnosis of complications such as lung diseases, cervical cancer, and eye diseases.
Furthermore, these solutions will be able to identify what features of the X-Ray helped the diagnosis, which will render the process more accessible and transparent.
TAMI - Transparent Artificial Medical Intelligence
TAMI - Transparent Artificial Medical Intelligence. Video Credit: CMU Portugal
Journal Reference:
Pedrosa, J., et al. (2022) Assessing clinical applicability of COVID-19 detection in chest radiography with deep learning. Scientific Reports. doi.org/10.1038/s41598-022-10568-3.