As Coronavirus Storm Accelerates, So Do AI-Driven Innovations

AI technology has slowly become part of the fabric of medical care. This integration has come in uneven dribbles and trickles, often delayed by a natural reticence to adopt new technologies, lack of infrastructure or training, cost, and sluggish FDA approval. Now, as COVID-19 strains the world’s medical infrastructure, so too will AI algorithms and solutions be put to the test.

Currently in use on the frontlines:

  • Chatbots: A variety of automated chatbots have recently been made available to help the public self-diagnose the severity of their symptoms. The CDC recently released one built on the Microsoft Azure platform; some are even programmed for multiple languages, including our own.
  • Facial scanning: Hospitals in major cities have installed cameras in their lobbies and waiting rooms. These cameras screen visitors for temperature, sweating, and other symptoms, which AI algorithms then analyze, identifying people with at-risk characteristics.
  • Data Mining: AI solutions can also help doctors and healthcare experts understand more about the disease and develop more effective treatments by quickly sorting and aggregating data from a variety of sources. On March 16, the White House’s Office of Science and Technology Policy issued an “all hands on deck” call to action for tech community leaders to come together to create a COVID-19 dataset of research studies intended for use with machine learning. The dataset, accessible to the public here, can be searched by topic kernels and notebooks. And however you feel about social media privacy policies, analyses of social media posts have proved useful in identifying epicenters of the outbreak. Dataminr, an AI company focusing on “high-impact events and emerging risks” claims it detected the burgeoning outbreak in Wuhan and delivered the earliest warning about it on December 30, 2019 by combing through social media posts.
  • Image analysis in healthcare settings: Since COVID-19 attacks the lining of the lungs, many AI driven approaches have focused on chest imaging. CAD4COVID is an AI-based tool that generates a heatmap from chest x-rays and scores them for COVID-related abnormalities for use in triage. Since concerns about cross-contamination of imaging equipment are a big concern, small portable scanners have been used to help treat patients. According to John Martin, CMO of Butterfly Network, a company that produces portable ultrasound scanners that plug into a smart phone, images for flu strain A and COVID-induced pneumonia are very similar but have subtle differences that can be detected by AI identification processes. Universities are scrambling to collect a larger data set of chest images on which to train AI systems so they can distinguish COVID-19 pneumonia from other types.

Interesting things in development:

  • Predictive Analytics: A partnership between NYU and Wenzai Central Hospital in China has concluded that that there is a strong correlation between severe COVID-induced lung disease and three specific symptoms: reported myalgia (muscle pain), higher levels of the ALT liver enzyme, and elevated hemoglobin levels. The researchers say that more case studies are needed to augment the data set, but that AI analysis of these three factors is a better predictor than initial lung images seem to be.
  • COVID Voice Detector: Pioneered by Carnegie Mellon University in conjunction with AI tech firms, this project aims to develop a tool that can analyze a person’s cough and speech in order to determine the likelihood of COVID infection. At the end of the assessment, the participant receives a COVID score of 1-10.

For now, the viral storm rages. Hopefully this crisis will deliver some technical silver linings.

About Author

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>