AI tool finds three new signs of coronavirus infection that can cause severe symptoms

Although most cases of neo-coronavirus infection remain in mild or moderate condition, some people do not show any symptoms of COVID-19 until recovery. Those with older or underlying diseases are more likely to need oxygen absorption or use a ventilator. To better screen, the researchers looked at artificial intelligence (AI) tools to find three signs that could accurately predict serious consequences, including routine test parameters in two hospitals.

AI tool finds three new signs of coronavirus infection that can cause severe symptoms

(Research portal: PDF report)

Given that the new coronavirus is still in the global epidemic, this study has a very important reference. At the time of writing, the United States alone accounted for 175,000 of the 818,000 cases worldwide.

Italy, based in Europe, is the hardest hit by the global outbreak, reporting more than 115,000 COVID-19 deaths (11.39%) deaths.

In the face of grim realities, there have been calls for preventive measures such as social alienation and encouragement of frequent hand washing, but we may have to wait a few weeks before we see a marked slowdown in the epidemic curve.

This means that the strain on medical resources has eased, giving patients with severe COVID-19 a better chance of survival, after all, there are no vaccines or special drugs available.

Some of the drugs and vaccines currently being tested have shown potential promise, but it will take a long time to verify before it can be made available to the public.

The good news is that a new artificial intelligence (AI) tool is expected to reveal three clues to COVID-19 complications to doctors after successful development. If extended to more patients, more lives could be saved in the coming months.

It is reported that many people infected with COVID-19 do not have obvious symptoms unless there are signs of fever, cough or shortness of breath. Even so, common problems such as flu, sore throat, and fatigue need to be ruled out in advance.

In addition, doctors observed that some patients claimed to have a barrier to smell and taste. This may be the most special place for COVID-19 than a common flu, but many people still experience only mild discomfort.

In the new study, Chinese and American researchers conducted AI analysis of data on 53 patients with neo-coronavirus infection from two hospitals in Wenzhou.

As a result, machine learning algorithms found three signs of severe illness — including body pain, levels of alanine amino transferase (ALT) enzymes, and abnormalities in hemoglobin levels.

As a liver enzyme, ALT can be used to diagnose diseases such as liver failure, and hemoglobin is part of the entry-level blood testing workflow.

AI believes that these three parameters accurately predict severe COVID-19 cases, and that the algorithm is 70-80% accurate in terms of risk cues for acute respiratory syndrome (ARDS).

ARDS is a complication of COVID-19, which fills the lungs and causes about 50% of patient deaths.

Other highly suspicious signs can be analyzed from specific patterns such as lung imaging, fever, and intense immune response, but poor predictions of whether ease will translate into ARDS critical illness esmos.

The model highlights some clinical data that may be overlooked by doctors, such as mild elevated ALT and hemoglobin, as well as myalgia.

Key characteristics of predictive diagnosis include fever, lymphocyte loss, and chest imaging, but cannot predict severity, as well as epidemiological risks such as age and gender.

It should be noted that although all ARDS patients in this new study were male, the majority of men did not develop ARDS.

Megan Coffee, a physician and professor at New York University’s Grossman School of Medicine, told AFP:

The decision to use machines to help involve large numbers of data points is fascinating, and it may be different from what clinicians usually see.

The research team is still looking for further refinements to prepare for deployment sometime in April.

For more information about this study, please also refer to the journal Computer, Materials and Differences (Computer, Materials and Continua).

Originally published as “Deas an Artificial Intelligence Framework for Data-D Prediction of Coronavirus Clinical Severity.”