Connect with us

Medical

Identifying the Sound of a Covid-19 Cough with Artificial Intelligence

Published

on

It is impossible to pick up with your naked ear the unique sound of a Covid-19 cough. But MIT researchers are now detecting asymptomatic cases of Covid-19 using nothing but the sounds of them coughing. 

We have seen researchers training algorithms to identify conditions like asthma and pneumonia using the sound of coughing. 

A team at MIT was experimenting with detecting the signs of Alzheimer’s using phone recordings when they decided to try the same models on Covid-19.

The team which was based at the Auto-ID Laboratory in MIT, gathered at least 200,000 sample cough recordings. 2,500 of them were confirmed to have tested positive for the virus that causes Covid-19. 

The coughs were crowd sourced from a website they set up allowing people to submit their own recordings of forced coughs.

4,000 of the samples were taken first to train the AI model. Another 1,000 were taken to test the model. 

The researchers only had to slightly tweak the model they had created to make it work for Covid-19 instead of Alzheimer’s and they succeeded in detecting the patterns of vocal cord strength, lung and respiratory performance, sentiment, and muscular degradation. These were the four biomarkers that exhibited patterns unique to Covid-19.

After training and testing, the model was ready for a run. They fed it with new cough recordings and it successfully detected Covid-19 correctly 98.5% of the time. For people who had Covid-19 but were asymptomatic, the accuracy of the model went all the way up to 100%.

According to Professor Brian Subirana who is the director of the Laboratory, these findings show that when you have Covid-19, the way you produce sound will change, even when you are still asymptomatic. 

The team is collaborating with multiple hospitals to gather more cough recordings. They hope to boost the model’s accuracy by working with a greater diversity of cough recordings.

The team is also working to put the model into an app that could allow people who may have asymptomatic Covid-19 to pre-screen themselves in a quick and non-invasive way. 

Anyone who is coughing and thinks that they might have Covid-19 could pre-screen themselves before going in for a test. 

Such a model could be used together with smart speakers and  with other listening devices. 

According to Subirana: “The effective implementation of this group diagnostic tool could diminish the spread of the pandemic is everyone uses it before going to a classroom, a factory, or a restaurant.

Continue Reading
Click to comment
Subscribe
Notify of
0 Comments
Inline Feedbacks
View all comments

Medical

Patients could Start to Monitor their Vision Remotely from Home

Published

on

By

Top UK eye hospital tests smartphone app for monitoring vision.

A unique pilot study by the top eye hospital in the UK will see patients remotely monitor their vision while they are on lockdown at home.

The app is meant for patients whose macula is affected by diseases. These diseases include AMD (age-related macular degeneration), degenerative myopia, and diabetes.

With the Home Vision Monitor, patients can keep on monitoring their vision without having to go to hospital. The 350 plus patients involved in the pilot study are undergoing treatment from Moorfields Eye Hospital in London.

A ‘shape discrimination’ test enables the app to work by displaying many shapes. Patients look at the screen and select shapes.

The results of the self-administered test instantly relay to a Moorfields Eye Hospital clinician, and this saves patients a trip to the hospital. The testing happens twice every week.

The clinician automatically receives an alert when the test reveals that the eye health of the patient has worsened. The clinician will be in a position to choose an intervention. This will enable clinicians to act earlier to stop diseases from progressing and hopefully achieve better outcomes.

After surveying app users, the hospital concluded that the app was easy to use for 93% of patients. 70% felt assured that they were regularly monitoring their vision in the midst of the coronavirus pandemic.

Patients usually test their vision at scheduled in-person eye exams at an interval of 4-12 weeks.

“Exciting new digital technologies, such as Home Vision Monitor, will further empower our patients to actively contribute to the management of their condition, in partnership with their clinicians to achieve the best possible outcomes,” explained consultant ophthalmic surgeon Konstantinos Balaskas. “Placing such tools in the hands of patients will both improve health outcomes for patients and reduce the capacity pressures of hospital-based eye departments.”

Roche global head of ophthalmology Jill Hopkins said: “Supporting such a high-risk group is essential during these extraordinary times. By testing patients’ vision more frequently and at home, the app may eliminate unnecessary hospital visits and escalate urgent cases where needed.”

“We believe that solutions such as these can continue to support patients and healthcare professionals beyond Covid-19 and contribute to the generation of real-world evidence to identify progression trends in AMD,” she continued.

Research has already shown that AI tools accurately detect chronic eye disease in diabetes patients, and these findings show that it is possible to reduce the human labor involved in detecting these eye diseases by as much as 50%.

Continue Reading

Medical

IVF gets Better with AI

Published

on

By

An Israeli startup has found a way to make fertility treatments more successful using Artificial Intelligence.

Embryonics has already seen no less than 6 full term pregnancies as a result of their novel system.

Doctors will often inseminate a female egg in a laboratory – a process known as In Vitro Fertilization. During first few days after the fertilization, the embryos will be housed in special incubators and watched continuously by an embryologist.

This embryologist will then return some of the embryos for in-human incubation. This is where it gets interesting because they have to choose what they feel are the most valuable embryos.

“The decision making is subjective, and is dependent on experience and intuition – a factor that changes between doctors and medical centers. Embryonics wide range of smart solutions simplifies complex processes with data-backed insight, aiming at optimizing quality and fertilizations rate success,” says Embryonics founder and CEO Yael Gold-Zamir.

The company says that the success rate of traditional methods of IVF carries a 30% success rate within the US, and that the success rate is even lower outside the US.

The company developed AI algorithms that help predict the likely success rate of fertility and IVF treatments. This takes a lot of the guesswork out of the selection process.

To train the model, scientists watched embryos developing during IVF treatment. As a result, the embryo-pregnancy success rate shot up by 20% and the model allowed doctors to predict the embryos that would not yield a successful pregnancy.

This difference is significant because it saves IVF patients a significant amount of money by making the process more efficient. By increasing the chances of success, it also saves patients from the emotional toll of unsuccessful IVF attempts.

Test Tube Babies become AI Babies

According to Dr. Gold-Zamir, Embryonics has come up with Ubar, the first commercially viable product that more accurately pinpoints healthy embryos.

It won’t be long before Embryonics’ product is in the market: The product could be CE cleared in a matter of weeks, and FDA approved within the year.

“Once regulatory approvals are in, we will start implementing the product in a number of Israeli medical centers that showed interest.”

“It’s a revolution. Traditional fertility treatments resulted in very low success rate. We are going to replace IVF test-tube babies with AI babies – a transition that will create a process that is shorter, cheaper, and less dangerous for millions of couples around the world.” Adds Zamir.

If we go by the findings of Israeli startups like Embryonics, AIDOC, and Zebra, you may be forgiven for thinking that AI medical imaging analysis may replace specialist like radiologists and embryologists in the near future.

But Dr. Gold-Zamir differs: “Algorithms will take over some of the radiologist’s tasks, opening more time for physicians to take a deeper dice into patient care. But that’s further down the road. Currently, all the available tools aid and improve radiologist and doctors’ workflow and quality, with tools trained in millions of different sample sources – a database too large for humans to process.”

The effect of Artificial Intelligence it appears, will be to make medical experts more accurate in their diagnoses, and not to replace them.

Dr. Yael Gold-Zamir founded Embryonics in 2018 and the company has grown to its current composition of 16 people. Embryonics has raised a $4million round of funding with the Schuctermann Family Investment Office as well as the Israel Innovation Authority.

Continue Reading

Medical

Four-Week Memory Test that could predict the Risk of Alzheimer’s

Published

on

By

Scientists now believe that it may be possible to identify people who have a greater risk of developing Alzheimer’s disease by testing their memory over a four-week period.

Trials have shown that zeroing in on the ability to retain a memory over a longer period could yield potentially more accurate predictors than the traditional memory tests. Traditional memory tests assess people’s memory over a period of half an hour.

University of Bristol researchers led the study which they published in the Alzheimer’s Research and Therapy journal in which they wanted to test using a word list to test people’s memory over four weeks. People were tested for their memory of the list four weeks after hearing it initially.

Those whose memory of the list after four weeks was better off, registered less cognitive decline over the year after. This was true even for those who did not have any problems with cognitive function or memory at the start.

Forty six older people participated in the study. All of them were healthy and their average age was 70.7. The study participants had to perform three memory tasks. Researchers tested t heir delayed recall 30 minutes later and then four weeks later.

They also did the Addenbrooke’s Cognitive Examination III or ACE-III test. ACE-III test is conventionally used to detect cognitive impairments, as well as a brain MRI. One year later, the participants repeated the same ACE-III test to measure their cognitive abilities.

At the end of the study, researchers found that 15 out of 46 participants had experienced a decline in their cognitive abilities, and that the verbal memory tests conducted over the initial period of four weeks would have yielded a better prediction of this decline than the standard memory tests.

After combining the scores from the four-week memory tests together with the results of the MRI brain scan, the researchers found that it was possible to make even more accurate prediction s of cognitive decline.

The MRI brain scans showed that there was a size reduction in the section of the brain that governs memory. This is the part of the brain that gets affected by Alzheimer’s Disease.

By testing long-term memory recall, the researchers found that they could detect Alzheimer’s disease much earlier and achieve better treatment outcomes. Treatments to stop Alzheimer’s disease or slow down its progression are much more effective when given during the early stages of the disease. This is the best time because patients are yet to develop significant problems with their memory.

In the words of Dr. Alfie Wearn, a Research Associate at the Bristol Medical School: Translational Health Sciences: “Our study shows evidence for a low-cost and quick to administer screening tool that could be used to identify the very earliest signs of Alzheimer’s disease. It could also directly speed up the development of effective Alzheimer’s disease therapies, and enable earlier treatment when such therapies are available.”

Dr. Liz Coulthard who is an Associate Professor in Dementia Neurology with the University of Bristol as well as a neurologist with the North Bristol NHS Trust also said this: “It is important to note the participants were healthy older people who did not develop Alzheimer’s during the trial, but some people did show the type of change over the course of a year in memory and thinking that can precede Alzheimer’s disease. Future work will establish whether this test predicts full-blown Alzheimer’s dementia.”

Researchers will now be testing the ability of these tests to detect Alzheimer’s disease in comparison with other cognitive deteriorating diseases. The researchers will be drawing comparisons between long-term memory test results of people who have evidence of Alzheimer’s disease and those without. The evidence is derived from analyzing their cerebrospinal fluid. Even though this method of detecting Alzheimer’s is the most effective, it is also too invasive, and researchers are working on less invasive methods.

The study was conducted with funding from a Wellcome Trust PhD study scholarship award received by Dr. Wearn who is in the Neural Dynamics PhD Programme. It was also funded by the Alzheimer’s Research UK as well as Brace, which is a dementia charity at Southmead Hospital in Bristol.

Continue Reading

Trending

Copyright © 2020 HiTECH