A new tool is used Artificial intelligence developed at the Radiology Institute of the Clínicas Hospital of the University of São Paulo (USP), helps to diagnose early liver lesions in CT scans of the abdomen, accelerate the treatment in more severe cases.
The platform, called HepatIA, was built from CT scans of 700 patients, including healthy people and cases of cirrhosis due to hepatitis and cancer.
“It is a work that we have been developing since 2018 and it has very detailed and accurate data to develop the algorithm”, says radiologist Bruno Aragão, one of the leaders of the work.
“Tools of this type already exist for other problems, such as pulmonary thrombosis, but we chose the liver because it is an area little studied by artificial intelligence researchers.”
When placing a new exam on the platform, the tool performs an initial screening suggests a possible diagnosis that must be confirmed by the health professional.
If cirrhosis or cancer is suspected, for example, that patient gets priority in the radiologists’ queue for analysis.
Initial results show that the tool is about 94% accurate for automatic identification of contrast phase, 84% for screening for changes in liver shape suggestive of cirrhosis, and 70% for screening for cancer in the scan.
“Our goal is solve flow and process problems especially in a hospital that has a large number of tests to process “, explained the doctor.
According to the researcher, the tool is still evolving, especially to improve the performance of small tumor screening.
A whole line of research has also been created with doctoral students and scientists around the project, which allows for the improvement and development of new functionality.
loss of muscle mass
In addition, as tomography allows assess muscle mass The tool also indicates when there is loss of this tissue – the so-called sarcopenia.
The doctor says that this is essential for cancer patients, because this condition leads to worse results.
“It quickly shows a calculation that ends up not being done regularly. Therefore, this patient can be referred for nutritional monitoring, for example”, explains Aragão.
This helps prevent complications, improve response to treatment and patient survival.
The research was developed in partnership between the Institute of Radiology and the Department of Gastroenterology at Hospital Clínicas, the Center for Technological Innovation in HC (InovaHC), FAPESP (Fundação de Amparo à Pesquisa do Estado de São Paulo) and the startup MaChiron. technology.