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.

November 4, 2022

In today’s Duke AI Health Friday Roundup: graph neural networks to describe galaxy evolution; bias in risk prediction models; CDC issues new opioid guidelines; app for exploring bias in AI-generated images; realizing biology’s potential for this century; how Wikipedia citations can affect impact of journal articles; giving an eel an MRI; the current state of medical malpractice law; much more:

AI, STATISTICS & DATA SCIENCE

Hubble Space Telescope photograph of the swirled and irregular shape of the Antennae Galaxies, two spiral galaxies whose shapes are being distorted by gravity as the two galaxies merge with each other. Image credit: NASA/ESA/Hubble Heritage
Image credit: NASA/ESA/Hubble Heritage
  • “Efficiently mapping baryonic properties onto dark matter is a major challenge in astrophysics. Although semi-analytic models (SAMs) and hydrodynamical simulations have made impressive advances in reproducing galaxy observables across cosmologically significant volumes, these methods still require significant computation times, representing a barrier to many applications. Graph Neural Networks (GNNs) have recently proven to be the natural choice for learning physical relations.” A preprint paper by Jespersen and colleagues, available from arXiv, describes the use of graph neural networks to uncover details of how dark matter affects the formation and evolution of galaxies.
  • “Most models were not reported according to established reporting guidelines or showed methodological flaws during the development and/or validation of the model. Further development of prediction models with thorough quality and validity assessment is an essential first step for future clinical application.” A systematic review by Hueting and colleagues, available as a preprint from the Journal of Clinical Epidemiology, reveals a substantial capacity for bias among hundreds of breast cancer risk prediction models.
  • Hugging Face AI researcher Sasha Luccioni has developed an app for the company’s Stable Diffusion image generator that allows people to explore bias in visual representation by doing side-by-side comparisons with different prompts. A Twitter thread has additional detail.
  • …and in other AI image generation news, the popular Open AI DALL-E image generator now has a publicly available beta release of its API: “Developers can now integrate DALL·E directly into their apps and products through our API. More than 3 million people are already using DALL·E to extend their creativity and speed up their workflows, generating over 4 million images a day. Developers can start building with this same technology in a matter of minutes.”

BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH

Close-up photograph of a bumblebee on a yellow blossom. Image credit: Mandy Henry/Unsplash
Image credit: Mandy Henry/Unsplash
  • “When animals repeatedly engage in behavior that does not provide them with food, shelter or another immediate benefit, researchers consider the behavior play. Play with inanimate objects is widely observed in animals, although most examples come from mammals and birds, with no record of the behavior in insects until now.” Scientific American’s Grace van Deelen reports on recently published research that suggests bumblebees may engage in play behavior – a key indicator of sentience.
  • “…what’s the biological equivalent of the atom bomb? Sure, there are worries that we’ll edit the wrong genes, or create biological inequality to mirror the economic inequality we already have. But the biggest looming problem is that we will simply become lost and confused as to what works and what doesn’t, scuttling our own progress, wasting money, and missing opportunities to save lives. That’s what happens when new technologies in biology outpace our ability to assess them.” A sweeping and truly big-picture summary of the current state of US therapeutic development by STAT News’ Matt Herper includes a calls for fundamental transformations in how we test and evaluate medical products.
  • “The 2022 Clinical Practice Guideline reflects updated evidence and research on the risks and benefits of prescription opioids for acute, subacute, and chronic pain, opioid dosing strategies and dose-response relationships, opioid tapering and discontinuation, comparisons with nonopioid pain treatments, and risk mitigation strategies.” The US Centers for Disease Control today released updated guidelines for prescribing the use of opioids for pain control.
  • “In this phase 2 trial involving participants with treatment-resistant depression, psilocybin at a single dose of 25 mg, but not 10 mg, reduced depression scores significantly more than a 1-mg dose over a period of 3 weeks but was associated with adverse effects. Larger and longer trials, including comparison with existing treatments, are required to determine the efficacy and safety of psilocybin for this disorder.” A paper published in the New England Journal of Medicine by Goodwin and colleagues describes a phase 2 randomized trial comparing 25-mg, 10-mg, and 1-mg control doses of psilocybin for treatment resistant depression.
  • “To determine whether the mass extended into Larry’s nasal passages or bone, veterinarians would need to give him a CT scan. The scans are not invasive, but imaging an eel poses a slippery challenge, requiring zoo employees to ferry Larry to a veterinary hospital across town, safely hoist him out of the water and somehow get him into a fancy piece of medical equipment….They would need to be careful, they would need to be fast, and they would need to remember the baking soda.” If you’ve ever wondered how to give an eel an MRI exam, the New York Times’ Emily Anthes has you covered.

COMMUNICATION, Health Equity & Policy

An ebony, brass-bound judge’s gavel sitting on a desk next to a pair of black notebook binders. Image credit: Sora Shimazaki/Unsplash
Image credit: Sora Shimazaki/Unsplash
  • “The past two decades of relative calm could have been an opportunity for trying innovative reforms aimed at making the malpractice system more accurate, fair, and efficient. For the most part, however, such reform hasn’t happened, and the system in place for more than 150 years grinds on, largely disconnected from wider efforts to improve the quality and safety of patient care.” A perspective article by Studdert and Hall published in the New England Journal of Medicine surveys the current state of medical malpractice law, and the implications of a downward trend in such cases.
  • “Data collected within the EHR are crucial for understanding and addressing inequities that negatively impact health and health care outcomes of marginalized communities, including persons with disabilities. Unfortunately, the lack of standardized collection of patients’ disability status within EHRs has limited progress toward addressing inequities for people with disabilities.” A recent article by Morris and colleagues published in the Health Affairs Forefront blog makes the case for standardizing electronic health record data that capture elements of disability.
  • “Scholarly publishers have reported increased traffic as a result of giving access to their publications to Wikipedia editors, and a controlled experiment on Wikipedia shows that they are right to value Wikipedia citations. Works cited on Wikipedia have an outsized influence on scholarly work — specifically in its literature reviews. Additionally, one research article found that open-access (OA) articles were cited more frequently than non-OA articles on Wikipedia in 2014, an idea supported by the generally increased readership of OA articles compared to paid-access articles…” A guest article by Rachel Helps at the Scholarly Kitchen blog describes how Wikipedia citations can affect a research article’s reach and impact.
  • “…barriers to timely medical care in the US increased for all population groups from 1999 to 2018, with associated increases in disparities among race and ethnicity groups. Interventions beyond those currently implemented are needed to improve access to medical care and to eliminate disparities among race and ethnicity groups.” A cross-sectional study by Caraballo and colleagues, recently published in JAMA Health Forum, describes the impact of racial and ethnic disparities on timely access to health care.