Artificial intelligence becomes a resource to streamline stroke exams

by

Artificial intelligence tools have helped to accelerate the diagnosis of stroke (stroke) and optimize the treatment of patients left with sequelae.

This is the case of software, developed by the Israeli company Aidoc, which is able to more accurately identify hemorrhagic-type strokes — when an artery ruptures, causing bleeding in the brain.

The program uses an algorithm to analyze the results of CT scans and highlights areas of bleeding that might go unnoticed by the human eye.

The solution is already offered in seven public hospitals in São Paulo and one in Goiás, through a partnership between Aidoc and Fidi (Foundation Institute for Research and Study of Diagnostic Imaging), a social organization that provides services to the SUS.

Igor Santos, physician and innovation superintendent at Fidi, explains that the result of a CT scan can take up to two hours to reach the doctor who requested it. This is due to the time radiologists take to analyze the image. Through the new solution, the wait can be reduced to up to 30 minutes.

Exam files are stored in the cloud, where images are analyzed. Upon identifying the stroke, the result is triggered to the patient’s doctor, along with information such as date, time, name and age.

Everything is done following the rules of the LGPD (General Data Protection Law), which regulates the use of personal information in digital media.

Since the program began to be tested, in 2018, exams of more than 100,000 patients have already been analyzed.

The technology also makes it possible to detect other types of diseases, such as brain tumors. “It’s an algorithm that specializes in cranial bleeding,” explains Santos.

With headaches and difficulty feeling his legs, Antonio Valentim da Silva, 71, was admitted to the Mandaqui Hospital, in the northern part of São Paulo, on 7 September. It was a stroke. The patient was quickly diagnosed and taken to the ICU thanks to the technology. Today, with no sequels, he says he feels good.

“Rapid diagnosis is super important. Time means nervous system cell viability. [a intervenção] it won’t help,” says José Krieger, a professor at the Department of Cardiopneumology at the USP Medical School (University of São Paulo).

He is part of the C4IA (Artificial Intelligence Center), which is developing an algorithm to improve the performance of tomography in identifying stroke. The project, which began in 2020, is the result of a partnership between USP, FAPESP (Research Support Foundation of the State of São Paulo) and IBM.

The technology wants to make tomography as accurate as MRI—more expensive and time-consuming. According to Krieger, while the first takes about 5 minutes to make, the second takes 30 to 40 minutes.

With the new algorithm, it would be possible to use the simplest exam to accurately locate the lesion and characterize the type of stroke: hemorrhagic or ischemic, when there is an obstruction of an artery.

Marco Antonio Gutierrez, engineer and one of the C4IA researchers, explains that MRI can identify the lesion earlier than CT. The project proposal is to make the second method as effective as the first.

“With the algorithm, we could, with a simpler test, detect an ischemic or hemorrhagic stroke early on,” says Gutierrez.

To build the tool, researchers will use artificial intelligence techniques and compare results from thousands of exams. Each patient participating in the study must have a CT and an MRI. The goal is for the algorithm to recognize patterns between the two diagnostic methods. The project is in its initial phase and is not expected to be finished.

Another initiative, also led by researchers from USP —but from the São Carlos campus (inside the state)— created a robot to help in the recovery of patients who were left with movement restrictions in their lower limbs.

Made of aluminum alloys, the prototype, which is attached to the waist and legs, measures the force a person is using to move. If it is enough to complete the move, the robot remains stationary; if the patient is unable to finish moving, the machine helps to boost him.

“The robot has sensors in the joints, with which we can measure whether the person has started to make the movement and at what intensity he/she did it”, says Adriano Siqueira, a professor of mechanical engineering and a researcher at USP.

As the user performs physiotherapy, the robot also collects movement angle measurements. From this, Siqueira’s team intends to use artificial intelligence to analyze the data and try to identify whether the patient had any evolution.

Although the prototype was completed in 2020, tests have not yet been carried out with a significant number of people. According to Siqueira, the pandemic was one of the factors that hampered the project’s progress.

.

You May Also Like

Recommended for you