AI stepping up against COVID-19
The technology landscape for Healthcare is flooded with tools, experiments and innovations that are poised to re-shape the domain in a way unimaginable, prior to COVID-19.
In my previous post, I touched upon Tele-medicine as the new key to providing continuity in medical care to patients. This week, the focus is AI and its role in the fight against COVID-19. Patient screening, monitoring and drug repurposing have seen some interesting AI experiments recently.
Screening and Social Control
► Providence St. Joseph Health system in Seattle collaborated with Microsoft to build an online screening and triage tool to help differentiate between Covid-19 patients and those suffering from other ailments.
► Partners Health-care Covid19 screener is an online web tool that throws up questions, based on content from the U.S. Centers for Disease Control and Prevention (CDC) and Partners HealthCare experts, on its chat interface. Basis its assessment, the tool will direct patients to the next steps to be taken.
► Clearstep, Babylon Health’s Covid-19 Care Assistant are other chatbot based online screening and triage tools available.
► AI systems built into cameras equipped with thermal sensors are getting widely deployed to scan for fevers. Chinese company Baidu uses AI and no contact infra red sensors to predict fevers. Currently deployed in Beijing’s Qinghe Railway Station, it can detect forehead temperatures of 200 people a minute.
► A group of scientists at Swiss University have developed an AI based application called “Coughvid” that listens to people cough and determine if they sound like a COVID patient.
► Tampa General Hospital has deployed an AI based facial recognition scanner at the entrances . This performs facial scans to screen visitors to the hospital who may have fever.
► There are also AI software solutions that help interpret lung CT scans. They quickly detect the visual signs of pneumonia linked to Covid patients. Zhongnan Hospital of Wuhan University in Wuhan, China, is experimenting with this software. This will help staff screen patients and prioritise those most likely to have Covid for further testing.
While AI based solutions are being used to facilitate the triage of patients with Covid symptoms, there are also intelligent robotic solutions that help with monitoring the symptoms and automating hospital operations.
Monitoring and Operations
► China’s Wuhan Wuchang Hospital used robots to staff a smart field hospital. They were used to monitor vital signs of patients, deliver medicine and food, reducing physician exposure to the virus and easing the workload of exhausted health care workers.
► At both Brigham and Women’s Hospital and at Massachusetts General Hospital, experiments are underway to use intelligent robots, developed at Boston Dynamics and MIT, to interview Covid patients, to obtain vital signs or deliver medication that would otherwise require human contact.
► MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) has developed Emerald , a WiFI like box, that helps to remotely monitor a COVID-19 patient’s breathing, movement, and sleep patterns using wireless signals .
► British Health-Tech startup, Medopad, is also providing clinicians with a specialized Covid-19 version of its Remote Patient Monitoring platform. The patients have a smart phone version of this app to securely share health data such as their heart rate, respiration rate and body temperature.
The use of AI is not just limited to patient monitoring or robotics, there is a fascinating movement that is using AI to discover drugs. With the global pressure on the discovery of an effective drug against COVID-19 mounting by the day, the use of AI in drug repurposing is the new hope.
What does drug repurposing mean?
It simply means using a drug meant for one disease on another.
This could be based on the expertise and educated guesses of scientists. Using Hydoxychloroquine, an anti-malarial drug, for some advanced Covid-19 patients was one such attempt.
Developing a new drug from scratch to combat Covid-19 is an effort that will take at least a decade. Instead, if you scan existing drugs, that have already been approved by regulatory bodies, to shortlist potential candidates that could work, the time taken to develop a viable could be greatly diminished.
A pre-print paper has recently outlined the use of AI based drug repurposing in the fight against Covid. Deep neural networks could be used to not just scan existing drugs but also scan a list of approved compounds, that have been developed for other ailments, that could work for corona virus symptoms. The pre-print paper is based on a hypothesis using SARS, a virus similar to the Covid-19. Since both these viruses have an 86% similarity in their genome, a drug that works for SARS could be effective in battling Covid-19.
There have been some other interesting experiments on the horizon:
► BenevolentAI, has identified Baricitinib, a drug approved for the treatment of rheumatoid arthritis, as a potential treatment to prevent the virus infecting lung cells
► Exscientia, first to put an AI-discovered drug into human trial, is trawling through 15,000 drugs held by the Scripps research institute, in California. Their idea is to screen every known approved and investigational drug against key Covid-19 drug targets to shortlist compounds that could possibly become viable drugs to treat the coronavirus.
There are many more companies using AI to find a treatment for Covid-19; Healx, Innoplexus, Deargen, Cyclica, Gero, VantAI to name a few.
Some caution may be advised. AI based discoveries need the right data set and mining tools and any solutions thereof need to be backed with controlled trials. The objective intervention of the scientific community is a must.
AI AI everywhere
So there you have it. AI based solutions rising up to the challenge in prevention, diagnosis, prognosis, operations, drug discovery and development, not to rule out its effective use in vaccine deployment too. While experiments are still at an early phase, the tools already in use look very promising to ease the load on the healthcare system. The challenges of unifying data from various resources, scrubbing out the noise and outlying data, getting the right dataset ( CORD-19) to train AI systems, continue. The role of AI in tracking and predicting the spread of this disease is still not accurate and will need exhaustive work. However, joint initiatives involving the government, healthcare, science and tech communities look to produce promising results as one works the learning curve.
Originally published on LinkedIn, by Suzette Sugathan, Director, NextServices