Digital twins of hearts could help diagnose and treat cardiac disease
This piece from The Economist describes efforts by a group, based in Belgium and England, to create a digital version of an individual’s heart that can be used for simulations and problem solving, based on their unique organ. It’s a technique that has been extensively employed with complex machines, like aircraft and vehicles. A detailed virtual model of the object, a “digital twin,” is constructed to be used for testing design modifications and for troubleshooting. The data that are the building blocks of the virtual heartwill be acquired continuously from the patient via a wearable EKG monitor and a portable ultrasound cardiac scanner. Developers dream of incorporating these items into comfortable garments, maybe even washable ones.View Article
The End of Starsky Robotics
Starting in 2015 Starsky Robotics vigorously pursued the dream of driverless trucks. In 2018 they actually put a vehicle on the road (on a closed road witha human driver for backup) and in 2019 on a live highway. The company is dissolving this year. As co-founder of the company, Stefan Seltz-Axmacher, outlines in this article, the technology that supports driving in the “real world” is not ready to perform with the discriminatory skill and rapid decisionmaking that a human driver employs every day in getting from point A to point B. Neither are investors willing to sink endless dollars into a technology whose payoff keeps receding down the road. Information intensive technologies may have replaced well-defined tasks heretofore performed by humans, like welding and connecting customers’ modems to the internet. But we are still plenty far away from having machines take overa wide range of sophisticated tasks that homo sapiens do today. This is especially the case in healthcare, which is at least as data-intensive as driving a truck on a city street and where a mistake can be at least as costly. As things stand now, we ought to be placing most of our efforts into developing AI-driven systems that enhance what we humans do best, not that replace us.View Article
Machine Learning Takes On Antibiotic Resistance
I hope you don’t think I’m against artificial intelligence. I’m just opposed to using it to try and take over the tasks that human beings still do far better than machines, rather than to help us do what we do better. This article from Quanta Magazine gives a fascinating peek at how scientists employed AI to extend their reach in screening for potential antibiotic activity among more than 6000 chemical compounds that had already been cleared to administer to humans as drugs. At least one of the 90% of these compoundsthat had never been used to treat infections has shown real promise against the common human pathogens, e. coli, m. tuberculosis and c. difficile. Reading how the researchers accomplished this task reminds me of the computer programs that routinely cream human competitors at poker by coming up with bluffing strategies that just don’t make sense to human waysof seeing the world. Likewise, the team at MIT used AI to scour big chemical databases for compounds with antibacterial mechanisms that are not yet within the purview of human scientists’ understanding of how things work, possibly uncovering novel anti-bacterial modes of action.View Article
Physicians’ Well-Being Linked To In-Basket Messages Generated ByAlgorithms In Electronic Health Records
The findings in this report from Health Affairs will come as no surprise to any physician who uses an electronic medical record. 934 physicians in a California multispecialty practice gave the authors of the study permission to monitor the volume of their electronic medical record inboxes over the course of 6 weeks. The doctors received on average 243 messages per week: 53 from colleagues, 30 from patients and 114 generated by the EMR itself. Doctors who received more than the average number of system-generated messages were 40% more likely to show signs of burnout and 38% more likely to be planning to reduce their clinical hours in the coming year. Volume of messages generated by human beings (colleagues and patients) did not correlate with burnout. As a rule, systems that automatically churn out clinical reminders are blunt instruments that tend to hammer physicians with way more useless and distracting “to-dos” than with focused and helpful suggestions. For the record, the practice studied used the Epic EMR, the most widely installed electronic medical record in the country, touted to be among the most advanced products on the market at supporting clinical decision making.View Article
Doctor delivers end-of-life news via 'robot,' leaving family frustrated
USA Today carried this story about a 78 year-old patient at a Kaiser Permanente facility in Fremont, California who learned, via video, from a doctor who was covering the intensive care unit, that there were no more effective treatments for his chronic lung disease and that he would not be able to return home for hospice care. What more is there to say about this inappropriate use of telemedicine?View Article
Paper lets scientists play with each other’s results
The journal, Nature, reports here that an online scientific publication, eLife, has taken a step toward opening the papers it publishes in a way that affordsother scientists the opportunity to reuse the code that a study’s authors employed to analyze their data. For example, a scientist could rerun a paper’s data without excluding outliers or with different cutoff points for chunking the data. Others could turn the paper’s statistical crank to analyze their own data. The medical literature is on the way, haltingly, to becoming an interactive resource, rather than a repository for fossilized findings.View Article
Seriously, Stop Trying to Teach Toddlers How to Code
The unifying theme of my book, Digital Healing: People, Information and Healthcare, is that in healthcare we need to relegate to electronic technologywhat it does best so as to free humans to employ those human characteristics--like our ability to communicate, to empathize and to touch--that allow us to do some things uniquely better than machines. That message goes for everybody, including toddlers. When people need to write computer code, and they want to learn it, they do so quickly. Adrienne So, the author of this opinion piece in Wired, argues that it’s human values that children need to master, not the computer skills embedded in “toys” that today’s parents, ambitious to rear the next garage billionaire, purchase for Junior.View Article
