AI Talk: Fair treatment, call clarity and book review

Jan. 29, 2021 / By V. “Juggy” Jagannathan, PhD

This week’s AI Talk…

Fair treatment

MIT Technology Review, my go to magazine for all things tech and AI, linked to an article in Nature this week. A team of researchers from Stanford, Microsoft, University of Chicago and University of California, Berkley focused on a specific problem: how pain caused by knee osteoarthritis is being treated. Typically, an X-ray of a patient’s knee is taken, and a radiologist scores the degree of disfunction of the knee (and the resulting pain) using a specific set of features in the image. It is a fact that the pain assessed based on this approach does not correlate well with pain actually reported by underserved Black and minority populations. The radiologist’s assessment, however, determines which treatment the patient qualifies for, such as drug therapy (opioids) or arthroplasty (knee replacement). Due to the assessment disparity, fewer Black patients become eligible for knee replacement which is associated with significantly better outcomes related to pain relief.

The researchers used thousands of knee X-rays and built a direct model to predict experienced pain from a diverse patient population. That model actually predicted pain very well, both for Black and non-Black patients. If the model is used, it can increase the Black patient population who becomes eligible for arthroplasty by 20 percent. Before the development of this model, physicians were likely to dismiss the extra pain experienced by the minority community as attributable to external to the knee (like stress). Now, they get a clearer picture as the tool helps prove that the pain experienced is indeed related to the internal workings of the knee and not external nebulous factors.

Call clarity

I saw this article in Futurism on how to improve the quality of your Zoom and other video conference calls (3). A new app from Krisp, an AI-based solution, filters out noise from both ends of a call even if only one end uses the app! Like Zoom, a free version is available for occasional users and a premium one for a modest monthly fee. This is a novel solution for a day and age when practically everyone is working from home and has to deal with random dog barks or children crying in the background. The machine learning model was developed using the conversations of 50,000 speakers! An interesting solution for interesting times that I will definitely be trying to see how it actually works.

Range – David Epstein

I am a huge fan of Bill Gates and follow his book recommendations avidly. One recent recommendation in his blog ”5 good books for a lousy year” caught my attention (5). The book: David Epstein – “Range. Why generalists triumph in a specialized world.” It was indeed a fascinating read. The fundamental premise the author explores here is that we need more generalists and not hyper-specialists.

I was hooked from the first examples: Tiger Woods and Roger Federer! Both are my heroes, but little did I know that they had two completely divergent upbringings. Tiger laser focused on golf from the time he was a toddler while Roger experimented with a variety of sports until he settled on tennis fairly late in his teens. Innovation, the author argues, results from people having diverse perspectives and life lessons whereas hyper specialization can lead one astray. For example, an interventional cardiologist may view any blockage in the artery as an opportunity to use their learned skills to intervene with a stent. It has been repeatedly shown in clinical trials that this procedure does not improve actual outcomes.

It was gratifying to note the book featured two 3M inventors! Andy Ouderkirk and Jayshree Seth—for their ability to bring a broad range of skills to innovative solutions (they both have impressive patent portfolios).

Researchers can get stuck in their silos and points-of-view. This is the message that comes across strongly in the book. Sometimes culture also plays an important role in promoting group think and tunnel vision.

The one quibble I have with the book is that the author did not address one organizational antidote for tunnel vision: multi-disciplinary teams. Tumor boards, consisting of a team of experts from different disciplines, treat cancer patients. Concurrent Engineering (CE), a solution that involves bringing together a diverse team of experts to address engineering problems, has been around for decades. I should know, I used to work at a center which focused on CE for more than a decade. In fact, the precursor to CE was called “Tiger team” and that dates back to the late sixties. Nevertheless, as a life lesson for all of us, his overall message resonates strongly with me.

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 Director of Research for 3M M*Modal and is an AI Evangelist with four decades of experience in AI and Computer Science research.