AI Health
Friday Roundup
The AI Health Friday Roundup highlights the week’s news and publications related to artificial intelligence, data science, public health, and clinical research.
June 10, 2022
In today’s Duke AI Health Friday Roundup: new lightning-fast algorithm solves maximum flow; discrimination puts strain on hearts; skeptical views on artificial general intelligence; head-turning cancer trial results from ASCO; using machine learning to reduce cognitive load on healthcare professionals; digital innovations in mental health may not reach everyone; tracking what may be multiple monkeypox outbreaks; much more:
BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH
- “Most of these innovations do little for people suffering from a serious mental illness, such as schizophrenia and bipolar disorder. Many with serious mental illness are unemployed, too poor to pay for online therapy, and, often, on the wrong side of the digital divide. They may not be seeking online therapy and, to be clear, there is no app for the range of social and medical problems they face.” In an essay for Science, former National Institute of Mental Health Director Thomas Insel questions whether recent technological innovations in care delivery for mental health are succeeding in reaching those who need access the most.
- A research article published in the New England Journal of Medicine by Cercek and colleagues describes results from (an admittedly small and preliminary) phase 2 trial of 16 patients with locally advanced rectal cancer, in which administration of an experimental checkpoint inhibitor therapy (dostarlimab) resulted in their tumors vanishing entirely at 6 months, with no reports of recurrence at the time the article went to press. Nor was this the only head-turning result to be presented at the recent American Society of Clinical Oncology (ASCO) meeting: a clinical trial (DESTINY-Breast04) of trastuzumab deruxtecan for so-called “HER2-low” breast cancer also posted noteworthy results, with a 49% reduction in risk of disease progression or death in patients treated with the therapy (versus patients treated with the physician’s choice of standard therapies).
- “The people infected in these three cases contracted the virus over a surprising geographic range of places — one in Nigeria, one elsewhere in West Africa, and the third in either the Middle East or East Africa. This apparent wide dissemination of a related virus — one that differs from the European outbreak strain — suggests monkeypox outbreaks outside of the countries where the virus is considered endemic may have been smoldering for longer than has been appreciated, Damon acknowledged.” STAT News’ Helen Branswell reports on a recent announcement from the Centers for Disease Control and Prevention that warns that the recent outbreak of monkeypox is actually “at least” two genetically distinct outbreaks.
- “The evolving virus and the uncertainty of predicting the trajectory of the pandemic call for strengthened surveillance and continued monitoring of SARS-CoV-2. The TAG-VE will continue to critically appraise state-of-the-art methodologies for predicting further evolution of SARS-CoV-2 and will continue to rapidly determine the threat levels posed by new variants.” A commentary published in Nature Medicine by Subissi and colleagues describes the ongoing work of a World Health Organization Technical Advisory Group charged with developing a framework for global monitoring of new SARS-CoV-2 virus variants.
AI, STATISTICS & DATA SCIENCE
- “Maximum flow has been studied since the 1950s, when it was formulated to study the Soviet railway system. ‘It’s older than maybe the theory of computer science,’ said Edith Cohen of Google Research in Mountain View, California. The problem has many applications: internet data flow, airline scheduling and even matching job applicants to open positions.” Quanta’s Erica Klarreich describes the creation of a new algorithm that solves “maximum network flow” problems faster than some researchers thought possible.
- “The results revealed an unprecedented error rate, with 279 of 281 (99.3%) examined references containing at least one error. Nonacademic documents tended to contain more errors than academic publications….GS data not only fail to be accurate but also potentially expose researchers, who would use these data without verification, to substantial biases in their analyses and results.” An article published in the Journal of Medical Internet Research by Romy Sauvayre describes an unexpectedly high rate of citation errors in Google Scholar entries in a survey of neuroimaging papers.
- “There’s also this underlying belief that the internet reflects society, and if we crawl the entire internet and compress it into a model, it will have diverse views. In a paper I wrote with the linguist Emily Bender and others, we make the point that size doesn’t guarantee diversity. Because of who has access to and uses the internet, you’re going to encode dominant and hegemonic views.” The Markup’s Julia Angwin talks with former Google AI researcher and Black in AI co-founder Timnit Gebru about building ethical AI systems.
- “Machines may someday be as smart as people, and perhaps even smarter, but the game is far from over. There is still an immense amount of work to be done in making machines that truly can comprehend and reason about the world around them. What we really need right now is less posturing and more basic research.” In an article for Scientific American, Gary Marcus brings to bear a skeptical perspective on recent developments in AI applications – including ones that have been heralded as milestones on the way to true artificial general intelligence.
- “ML is capable of decreasing the mental effort required to process immense amounts of biomedical data, yet its ability to do so is rarely, if ever, evaluated. We argue that cognitive load can and should be measured throughout the ML development cycle to maximize the potential for integration of ML into medicine and to improve patient and provider outcomes.” A comment in Nature Medicine by Ehrmann and colleagues argues that a key focus in healthcare AI applications should be the reduction of “cognitive load” on healthcare professionals, and whether that reduction in load can yield better decision-making and improvements in patient outcomes.
- “Our theoretical and quantitative results suggest that the numerical gaps between contrastive and noncontrastive methods in certain regimes can be significantly reduced given better network design choice and hyperparameter tuning.” A preprint by Garrido and colleagues available from arXiv examines similarities between contrastive and non-contrastive self-supervised learning approaches for image analysis.
BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH
- “Most of these innovations do little for people suffering from a serious mental illness, such as schizophrenia and bipolar disorder. Many with serious mental illness are unemployed, too poor to pay for online therapy, and, often, on the wrong side of the digital divide. They may not be seeking online therapy and, to be clear, there is no app for the range of social and medical problems they face.” In an essay for Science, former National Institute of Mental Health Director Thomas Insel questions whether recent technological innovations in care delivery for mental health are succeeding in reaching those who need access the most.
- A research article published in the New England Journal of Medicine by Cercek and colleagues describes results from (an admittedly small and preliminary) phase 2 trial of 16 patients with locally advanced rectal cancer, in which administration of an experimental checkpoint inhibitor therapy (dostarlimab) resulted in their tumors vanishing entirely at 6 months, with no reports of recurrence at the time the article went to press. Nor was this the only head-turning result to be presented at the recent American Society of Clinical Oncology (ASCO) meeting: a clinical trial (DESTINY-Breast04) of trastuzumab deruxtecan for so-called “HER2-low” breast cancer also posted noteworthy results, with a 49% reduction in risk of disease progression or death in patients treated with the therapy (versus patients treated with the physician’s choice of standard therapies).
- “The people infected in these three cases contracted the virus over a surprising geographic range of places — one in Nigeria, one elsewhere in West Africa, and the third in either the Middle East or East Africa. This apparent wide dissemination of a related virus — one that differs from the European outbreak strain — suggests monkeypox outbreaks outside of the countries where the virus is considered endemic may have been smoldering for longer than has been appreciated, Damon acknowledged.” STAT News’ Helen Branswell reports on a recent announcement from the Centers for Disease Control and Prevention that warns that the recent outbreak of monkeypox is actually “at least” two genetically distinct outbreaks.
- “The evolving virus and the uncertainty of predicting the trajectory of the pandemic call for strengthened surveillance and continued monitoring of SARS-CoV-2. The TAG-VE will continue to critically appraise state-of-the-art methodologies for predicting further evolution of SARS-CoV-2 and will continue to rapidly determine the threat levels posed by new variants.” A commentary published in Nature Medicine by Subissi and colleagues describes the ongoing work of a World Health Organization Technical Advisory Group charged with developing a framework for global monitoring of new SARS-CoV-2 virus variants.
COMMUNICATION, Health Equity & Policy
- “As a Black woman, Carter-Williams was at high risk of having a heart attack. Despite that, she is also among the patients most likely to be overlooked in screening tests or have symptoms dismissed as not heart-related. Outdated thinking holds that overeating or a sedentary lifestyle are the main risk factors. But discrimination is also deadly: both within the U.S. and around the world, people who experience gender, race, socioeconomic or other discrimination are far more likely to suffer and die from heart disease.” At Scientific American, Jyoti Madhusoodanan examines how discrimination impacts people seeking care for cardiovascular diseases.
- “Instead, it appears that researchers may have been testing the wrong hypothesis. Quantity of screen time may not be as important as what kids are doing on their screens and what age they are when they are doing it. In other words, it’s quality not quantity that matters.” At the Markup, Julia Angwin talks with researcher Amy Orben about what the science is currently telling us about the effects of screen time on the mental health of younger people.
- “On May 3, the tribal council voted nearly unanimously to banish the Lakota Language Consortium — along with its co-founder Wilhelm Meya and its head linguist, Jan Ullrich — from setting foot on the reservation. What the council took into consideration wasn’t just the organization’s dealings with the Standing Rock Sioux; it turned out at least three other tribes had also raised concerns about Meya, saying he broke agreements over how to use recordings, language materials and historical records, or used them without permission.” NBC’s Graham Lee Brewer reports on a dispute between the Lakota tribe and a nonprofit language consortium over rights to recordings and other teaching materials gathered as part of efforts to preserve the Lakota language.
- “Email alerts to authors and editors are inadequate to correct the impact of retracted publications in citing systematic reviews and guidelines. Changes to bibliographic and referencing systems, and submission processes are needed.” A randomized study published in the journal Accountability in Research by Avenell and colleagues describes results of an experiment aimed at assessing how authors of systematic reviews and clinical guidelines would react to notification of retracted articles cited in their work.
