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.

September 16, 2022

In today’s Duke AI Health Friday Roundup: peering through COVID-induced “brain fog”; critiquing academic culture at computer science conferences; cardiovascular polypill trial results show benefit for secondary prevention; medical racism, radiation, and x-rays; the case for better data on race, ethnicity & language; cases of acute flaccid myelitis increase; AI decodes speech from thought without invasive probes; much more:


Nine small images with schematic representations of differently shaped neural networks, a human hand making a different gesture is placed behind each network. Image credit: Alexa Steinbrück / Better Images of AI / Explainable AI / CC-BY 4.0
Image credit: Alexa Steinbrück / Better Images of AI / Explainable AI / CC-BY 4.0
  • “…ever since conferences adopted a new review process that shields the names of judges, as well as papers’ authors — many made the change in the early 2000s — critics say a new problem has arisen: Rejection notes are often so random, or just factually incorrect, that applicants suspect nobody actually read their paper. “ At Protocol, Anna Kramer reports on accusations levied by a distinguished Berkeley computer science professor that the academic culture of research conferences has gone off the rails in recent years.
  • “Realizing increasingly complex artificial intelligence (AI) functionalities directly on edge devices calls for unprecedented energy efficiency of edge hardware….we present NeuRRAM—a RRAM-based [compute-in-memory] CIM chip that simultaneously delivers versatility in reconfiguring CIM cores for diverse model architectures, energy efficiency that is two-times better than previous state-of-the-art RRAM-CIM chips across various computational bit-precisions, and inference accuracy comparable to software models quantized to four-bit weights across various AI tasks.” A research article by Wan and colleagues published in Nature describes a new kind of computer chip that promises to increase efficiency for computationally intensive ML tasks while reducing power consumption (H/T Stanford News).
  • “The Partnership on AI (PAI) believes that newsrooms using AI must consider AI ethics with the same rigor and sensitivity that they already apply to journalistic practices. Notably, such AI ethics should not be thought of as completely distinct from journalistic ethics, but rather an extension of them.” An article by Adriana Stephan and Claire Leibowicz at the Partnership for AI’s Medium page advocates for a thoughtful approach when journalists adopt AI tools for their profession.
  • “Decoding language from brain activity is a long-awaited goal in both healthcare and neuroscience. Major milestones have recently been reached thanks to intracranial devices: subject-specific pipelines trained on invasive brain responses to basic language tasks now start to efficiently decode interpretable features (e.g. letters, words, spectrograms). However, scaling this approach to natural speech and non-invasive brain recordings remains a major challenge.” In a preprint available from arXiv, a group of researchers from Meta describe the application of AI to parsing and decoding speech from electrical signals in human brains gathered through noninvasive sensors.
  • “We now argue that an advanced agent intervening in the provision of its reward would likely be catastrophic. One good way for an agent to maintain long-term control of its reward is to eliminate potential threats, and use all available energy to secure its computer.” In an article at AI Magazine that has been getting a fair amount of buzz this week, a group of Oxford computer scientists explore – in more sober and rigorous fashion than usual – a set of scenarios in which autonomous goal-seeking AI “agents” could “catastrophically” outcompete humans.


Silhouetted figure standing with back to camera in an open field that’s shrouded with fog. Image credit: Jakub Kriz/Unsplash
Image credit: Jakub Kriz/Unsplash
  • “Long-haulers with brain fog say that it’s like none of the things that people—including many medical professionals—jeeringly compare it to. It is more profound than the clouded thinking that accompanies hangovers, stress, or fatigue. For Davis, it has been distinct from and worse than her experience with ADHD. It is not psychosomatic, and involves real changes to the structure and chemistry of the brain.” The Atlantic’s Ed Yong explores one of COVID’s most mysterious and vexing side-effects – a collection of cognitive challenges known as “brain fog.”
  • “…the mitigation efforts imposed to reduce the spread of the coronavirus that causes Covid-19 also dramatically reduced the transmission of other respiratory viruses like EV-D68. That meant there was no wave in 2020 as would have been expected; there were just 33 AFM cases that year…Experts are watching to see if this EV-D68 wave will turn out to be particularly large.” At STAT News, Andrew Joseph and Helen Branswell report on a recent increase in cases of the polio-like viral illness known as acute flaccid myelitis, which typically affects younger children.
  • “…the lab tests for newborn screenings generally don’t use all of the half-dozen or so drops of blood collected on filter paper cards. So states hold on to the leftover “dried blood spots,” as they’re called, often without parents’ knowledge or consent. In recent years, privacy-related concerns have grown about the sometimes decades-long storage and use of the material.” Kaiser Health News’ Michelle Andrews explores growing concerns around issues of privacy and consent related to leftover blood samples from routine infant heel-sticks done immediately following birth.
  • “A high-fat, high-sugar diet eliminated the helpful bacteria, but giving mice bacterial supplements prevented them from developing obesity and metabolic syndrome, even on the high-fat, high-sugar diet. A closer look at the ingredients in that diet revealed that its high sugar content fuelled the growth of Erysipelotrichaceae bacteria, which killed off the immune-regulating bacteria.” A Nature research highlight reports on a new mouse-model study, published in Cell by Kawano and colleagues, illuminates the potential relationship between the human microbiome, a high-sugar diet, and potential health consequences.
  • “In this phase 3, randomized, controlled clinical trial, we assigned patients with myocardial infarction within the previous 6 months to a polypill-based strategy or usual care. The polypill treatment consisted of aspirin (100 mg), ramipril (2.5, 5, or 10 mg), and atorvastatin (20 or 40 mg)…. Treatment with a polypill containing aspirin, ramipril, and atorvastatin within 6 months after myocardial infarction resulted in a significantly lower risk of major adverse cardiovascular events than usual care.” Results from a randomized phase 3 trial published in the New England Journal of Medicine by Castellano and colleagues suggests that a cardiovascular polypill – a single pill that combines several low-dose medications – helps reduce risk of death and disability following a heart attack.

COMMUNICATION, Health Equity & Policy

X-ray showing wrist and finger bones. Image credit: Cara Shelton/Unsplash
Image credit: Cara Shelton/Unsplash
  • “The history of race adjustment for x-ray dosing reveals how mistaken assumptions can be admitted into medical practices — and how those practices can be ended.” An article in the New England Journal of Medicine by Bavli and Jones explores the not-all-that-distant history of medical racism that led to Black Americans consistent receiving excessive doses of radiation during routine X-ray exams.
  • “We find strong evidence for the status bias: while only 23 percent recommend “reject” when the prominent researcher is the only author shown, 48 percent do so when the paper is anonymized, and 65 percent do so when the little-known author is the only author shown.” A study preprint by Huber and colleagues, available for download at SSRN, finds that perceived author eminence exerts a significant influence on peer-review recommendations.
  • “The move means that bibliometricians, scientometricians and information scientists will be able to reuse citation data in any way they please under the most liberal copyright licence, called CC0. This, in turn, allows other researchers to build on their work.” Nature’s Dalmeet Singh Chawla reports on the opening of the CrossRef publication database to researchers, who are now able to make use of its bibliometric data on relatively unrestricted terms.
  • “The disproportionate impact of the COVID-19 pandemic on historically marginalized communities has elevated and motivated a focus on equity-oriented data—especially race, ethnicity, and language (REL) data—that can be used to identify and intervene on urgent population health priorities.” In an article published by Health Affairs Forefront, Rowen and colleagues argue for better and more widespread collection of “REL” (race, ethnicity, and language) data in order to better identify and understand patterns of disparity in health and healthcare.
  • At STAT News, Sarah Owermohle reports on the appointment of research biologist Renee Wegrzyn to head the newly created Advanced Research Projects Agency – Health (ARPA-H).