AI Talk: Clinics, collars, a childhood condition and CT scan analysis

May 3, 2019 / By V. “Juggy” Jagannathan, PhD

This week’s AI Talk…

AI-clinics

The future is rapidly creeping up on us in unlikely places: Safeway stores. Safeway stores in Arizona are offering ways to be seen for illness or injury at their virtual health kiosks. The kiosks combine AI, augmented reality and telehealth support all in one place. A variety of sophisticated tools guide the patient through the collection of all their symptoms. Gadgets gauge their blood pressure, temperature, chest sounds, blood oxygen levels, etc. and even provide preliminary diagnosis. The information is sent to a remote physician who reviews the information collected and renders diagnosis and treatment. One interesting aspect of these clinics: The infrastructure for these already existed, built by Safeway for their failed venture with the infamous company Theranos. Apparently, the clinic only needs to see 10-15 patients a day to be profitable. The prognosis for these clinics looks good! A model for the nation?

Childhood asthma detection

I saw a reference to work in early detection of asthma in a communique by the American Medical Informatics Association. First some sobering statistics. 17 percent of children suffer from asthma. Almost two thirds of children with asthma under 18 years had a delay in identification and diagnosis of the condition. Delay implies these children were not afforded the right treatment at the right time resulting in more long-term deleterious effects on their lungs. The solution? Use natural language processing of EHR clinic notes from children to abstract symptoms and conditions relevant for asthma. This research from Mayo Clinic is using more traditional AI—rule and heuristics-based identification of childhood asthma.

Radiation Therapy Target Divining

The Journal of the American Medical Association (JAMA) recently published a paper with a novel way to solve a machine learning (ML) problem. The researchers from Brigham and Women and Dana-Farber Cancer Institute published a ML problem with a bounty: $55,000 in prize money. The problem? How to segment and identify areas in the lungs using CT scans to target radiation therapy. The data? A curated set of CT scans were provided as a training set with a held-out evaluation set. The strategy worked. 564 contestants from 62 countries participated in the contest. After three phases, the winning algorithm was whittled down to a group of 5 algorithms. The resultant solutions were as good as radiologist in identifying target areas of the lung for radiation therapy! An interesting crowd-sourced approach to rapidly converge on a reasonable solution for this particular problem!

Color of your collar

This story on medium.com tickled my fancy. It’s about the coming onslaught of new jobs that pretty much redefine almost every job on the planet! These changes, brought by AI, the Internet of Things (IoT), 5G, blockchain, etc., are being called the fourth industrial revolution. This transformation in the next 10 years alone is valued at $100 trillion. Be that as it may, what do you call the new breed of workers whose jobs involve working with smart-assistants of all kinds? They are calling them “new-collar” jobs. According to Wikipedia: “A new-collar worker is an individual who develops technical and soft skills needed to work in the contemporary technology industry through nontraditional education paths.” What puzzles me is the lack of imagination in naming the new worker class! Have we run out of colors? Blue and white are admittedly taken, but perhaps we could name them “purple collar?” Or red? Some people just lack imagination.

I am always looking for feedback and if you would like me to cover a story, please let me know. “See something, say something”!  Leave me a comment below or ask a question on my blogger profile page.

V. “Juggy” Jagannathan, PhD, is Vice President of Research for M*Modal, with four decades of experience in AI and Computer Science research.