New AI tool helps identify dangerous respiratory syndrome

Researchers from Endeavor Health and the Northwestern University have created a tool based on artificial intelligence to help doctors recognize an underdiagnosed and often deadly respiratory syndrome found in seriously sick hospital patients. The tool has already identified historical cases with 93%precision, and it will soon be controlled for patients in processing in Endeavor.

Acute respiratory distress syndrome, or SDRA, occurs when the lungs become seriously inflamed – much more than a typical pulmonary infection. This inflammation leads to significant damage, causing a liquid leak in other parts of the lung in air bags. Due to this accumulation of liquid, patients cannot put oxygen in their blood circulation.

It is often called a kind of “drowning on dry land”, where the body’s immune system fills liquid lungs. Félix Morales, the main scientist of the project data, described it as “leaks on your own circulatory system in the lungs”.

The SDRA has a very high mortality rate, with up to 46% of patients who die from the disease. For those who survive, this often means permanent scars in the lungs or cognitive deficiencies due to a prolonged lack of oxygen. It can be triggered by many different medical conditions, but is often observed in people who are already seriously sick with conditions such as sepsis or pneumonia. It is also a main cause of death in covid-19 patients.

“During the first year of the pandemic, we also had 20 -year -old and 30 -year -old children who died from the SDRA because of severe stuck,” said Dr. Curtis Weiss, a health pulmonologist and a co -director of critical medicine. He has been working on this problem since 2018 and was part of the team that created the automatic learning tool to recognize the signs of SDRA in patients.

It is not a generative AI like Chatgpt – there is no way to create new information and “hallucinate” data that is not real. Instead, he will examine the information already available for patient medical records, such as laboratory results and imagery. The end result, hopes the team, will be an automated revision system ensuring signs of SDRA in patients. The system will not diagnose patients with SDRA – it will be known to doctors that their patients can suffer and suggest that they look at this angle.

Weiss learned at the start of his career that the SDRA is underdiagnosed, both because of its many potential causes and because it is easily confused with other conditions. In order to diagnose the SDRAs, doctors must observe many factors, such as oxygen levels, chest radiographs and if the patient has another condition such as sepsis or pneumonia which is known to trigger the SDRA.

“My hypothesis is that one of the reasons for the sub-authority is that the doctor does not integrate these different parts of the diagnosis,” said Weiss. Doctors in intensive care are under an “overload of information” of dozens of seriously ill patients every day, it is therefore not surprising that the perfect storm of factors for the SDRA can be missed even by the most competent doctors.

Knowing that the SDRA is the reason why the lungs are obstructed, however, can considerably change the way patients are treated.

Weiss used the example of congestive heart failure, which is very similar to the SDRA. Congestive heart failure can also cause an accumulation of liquid in the lungs, but liquid comes from the inability of the heart to effectively pump blood, not pulmonary inflammation causing damage.

If the liquid in a patient’s lungs makes them unable to breathe, a doctor can put them on a fan; The specific fan operation, however, is different for congestive heart failure and the SDRA. In addition, studies have shown that placing a person with SDRA on the belly helps their lungs better clean the liquid. In patients with congestive heart failure, this type of posture could put too much pressure on the heart.

“Sometimes it means that we are undermining something as serious as the SDRA, because it requires choosing the right information and putting everything in the right sequence at the right time to say, OK, this patient has a SDRA,” said Weiss. If there was a computer program that could bring these parts together and alert doctors when they are in place, more TDD cases could be diagnosed and treated.

The next step for the team is to see if it can identify the SDRA in patients who are currently in the hospital, essentially predicting the diagnosis before it is made by a doctor.

Currently, the tool has only been used on medical cases that have already been resolved – patients who have fallen ill and have received a diagnosis of SDRA. The tool has positively identified 93% of cases, with false positives only 17% of the time. The team said they could adjust the tool to have less false alarms, but given the severity of the disease, they prefer that it signals patients as Miss patients who are not.

“I prefer to treat all patients with SDRA, then a few others who may not have SDRA, instead of not treating some of the patients who really have SDRAs,” said Weiss. “We are trying to solve a problem of under-recognition of the SDRA.”