In this week’s Duke AI Health Friday Roundup: toolkit for applying NLP to EHR free-text; AI powers wildlife conservation efforts; addressing racism in medical education; what’s next for AlphaFold; questioning the review process for NSF fellowships; new hydrogel is crushing it, literally; mobile health for reducing health inequities; a new framework for managing medical technologies; AI and a new era of colonialism; much more:
- “Building on TDAI’s open-source platform called Wildbook, which helps wildlife researchers gather and analyze photos, the team is now focusing on generative AI approaches. These programs use existing content to generate meaningful data. In this case, they are attempting to analyze crowdsourced images to make biological traits that humans may naturally miss computable, like the curvature of a fish’s fin — or a leopard’s spots.” The Verge’s Kathleen Wong reports on efforts to equip wildlife conservation with powerful AI tools that can spot fine details that might elude human eyes.
- “For all the challenges, photographers say that hunting bioluminescence is rewarding in part because the phenomenon is endlessly surprising….One clear night, Mr. Bain drove about 40 miles to a beach where he hoped to photograph the Milky Way galaxy. When he arrived, he saw not only a sky full of stars but a glowing shoreline.” A beautiful photo essay narrated by Mike Ives at the New York Times will have you looking up, or down, with equal wonder.
- “This new field, the study of interstellar meteors, certainly has much to tell us about our place in the cosmos. Further investigations of the observed properties of the 2014 meteor could reveal new insights about our local interstellar environment…” In a story for Scientific American, Amir Siraj describes (from his perspective as one of a number of researchers working to unravel the mystery) how data from military spy satellites were key to eventually identifying a fireball meteor as an arrival from interstellar space beyond the edges of the Solar System.
AI, STATISTICS & DATA SCIENCE
- “The Electronic Health Record (EHR) is an essential part of the modern medical system and impacts healthcare delivery, operations, and research. Unstructured text is attracting much attention despite structured information in the EHRs and has become an exciting research field. The success of the recent neural Natural Language Processing (NLP) method has led to a new direction for processing unstructured clinical notes.” In a preprint available from arXiv, Li and colleagues describe the creation of a Python library for applying NLP methods to unstructured text entries in electronic health records (H/T @arXiv_Daily).
- “Crafted precisely and accurately, these models could significantly reduce costs and also keep patients healthier, said Nigam Shah, a biomedical informatics professor at Stanford….But that requires a level of coordination and reliability that so far remains rare in the use of health care algorithms. There’s no guarantee that these models, often homegrown by insurers and health systems, work as they’re intended to.” STAT News’ Mohana Ravindranath reports on the growing interest among health systems in using machine learning applications to anticipate and manage the cost of care for patients, especially the subset of patients known as “high utilizers.”
- “At the US population level, the implementation of AI detection during screening colonoscopy resulted in yearly additional prevention of 7194 colorectal cancer cases and 2089 related deaths, and a yearly saving of US$290 million.” A modelling study published in Lancet Digital Health by Areia and colleagues examines the cost-effectiveness of using artificial intelligence tools as part of colonoscopy screening.
- “…the stories reveal how AI is impoverishing the communities and countries that don’t have a say in its development—the same communities and countries already impoverished by former colonial empires. They also suggest how AI could be so much more—a way for the historically dispossessed to reassert their culture, their voice, and their right to determine their own future.” An article at by Karen Hao at MIT Technology Review, the first in a series, examines the question of whether artificial intelligence is recapitulating old patterns of colonial exploitation worldwide.
- “For summarization, we find that increasing model size is more energy efficient than increasing sequence length for higher accuracy. However, this comes at the cost of a large drop in inference speed. For question answering, we find that smaller models are both more efficient and more accurate due to the larger training batch sizes possible under a fixed resource budget.” A preprint available from arXiv describes a study by Ang and colleagues that compared the tradeoff between accuracy and efficiency for two different “long-context” natural language processing models.
BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH
- “…although the gadgets are a technical achievement, some cardiologists say the information the devices produce isn’t always useful. Notifications from the devices aren’t definitive diagnoses….It’s a conundrum, and a consequential one, for the health care system. Tens of millions of people are armed with these devices, and if even a small fraction of those get a ping, it could mean much more care and costs for the system.” At Kaiser Health News, Darius Tahir plumbs the paradox of wearable devices whose heart-monitoring capabilities offer a sometimes too-rich banquet of individual-level heart data.
- “In some cases, the AI has saved scientists time; in others it has made possible research that was previously inconceivable or wildly impractical. It has limitations, and some scientists are finding its predictions to be too unreliable for their work. But the pace of experimentation is frenetic.” In an article for Nature, Ewen Callaway covers recent developments with AlphaFold, the protein-shape-predicting AI and looks at what’s next for the DeepMind project.
- “The emergency use authorization of the InspectIR Covid-19 Breathalyzer is a meaningful milestone in the yearslong quest to develop more breath-based diagnostics, as well as innovative new tests for Covid, experts said. And it is likely to be the first of many similar breath-based Covid tests, experts said.” The New York Times’ Emily Anthes reports on the recent FDA authorization of a breathalyzer test capable of diagnosing COVID infection.
- “Our strategies enable a soft hydrogel to break a brick and construct underwater structures within a few minutes.” A research article published in Science by Na and colleagues describes the fine-tuning of a hydrogel “soft actuator” that uses a method reminiscent of turgor pressure in plant cells to increase the pressure such gels are able to exert (and withstand).
COMMUNICATIONS & DIGITAL SOCIETY
- “Students and educators, fueled by the same calls for justice that ignited the country’s racial reckoning in 2020, are demanding change from their institutions. They want their education ingrained in antiracism and hope that by acknowledging and teaching about bias and systemic discrimination in the medical field, the next generation of doctors will be better equipped to dismantle racism within health care.” In the latest episode of the STAT News podcast Color Code, host Nicholas St. Fleur talks with medical students and others about efforts to address racism embedded in medical curricula and practice (transcript also available).
- “Retraction of published research is laudable as a post-publication self-correction of science but undesirable as an indicator of grave violations of research and publication ethics. Given its various adverse consequences, retraction has a stigmatizing effect both in and beyond the academic community. However, little theoretical attention has been paid to the stigmatizing nature of retraction.” Nobody wants to have to retract a paper, yet doing so can be not only necessary, but a laudable indicator of scientific honesty and self-correction. A new paper in Minerva by Xu and Hu looks at how the stigma associated with retracting a publication can affect the process.
- “…according to several reviewers, NSF doesn’t provide sufficient concrete guidance on how to weigh different parts of the application, such as a student’s academic and research track record, the quality of their research proposal, and their potential to make a societal impact. Because of that, the resulting scores are sometimes highly variable among reviewers, they say.” An article in Science by Katie Langin distills a gathering controversy over how National Science Foundation fellowships are awarded.
- “Our current laws, regulatory bodies, and other governance structures — both “hard” (such as legally binding laws and regulations) and “soft” (such as voluntary guidelines, standards, and norms) — were largely built for a research, development, and market landscape that has changed substantially over recent years. As a result, our current approach to governance is no longer fit for purpose.” An essay published in the New England Journal of Medicine by Mathews and colleagues from the National Academy of Medicine suggests the necessity of a new framework to provide oversight and governance for the rapidly changing landscape of medical technologies.
- “In the United States, we have a preconceived notion that health care is provided in a doctor’s office with a speckled tile floor and white walls filled with detailed anatomical diagrams. We imagine a blue examination table with itchy white paper and folders on the wall with pamphlets, likely only in English, about “healthy habits.” But health care is already provided in so many other ways. Whether it is in an ambulance, at a school health clinic, or via telehealth platforms, we have the infrastructure to innovate outside of hospital wings and urgent care clinics.” An award-winning essay by Duke Margolis Center Keren Hendel published in Health Affairs Forefront examines ways that innovative deployment of mobile health could chip away at health inequities.
- “Most NCI-designated cancer centers did not publicly disclose payer-specific prices for cancer therapies as required by federal regulation. The findings of this cross-sectional study suggest that, to reduce the financial burden of cancer treatment for patients, institution of public policies to discourage or prevent excessive hospital price markups on parenteral chemotherapeutics might be beneficial.” An analysis by Xiao and colleagues published in JAMA Internal Medicine reveals markups, some quite substantial, on chemotherapy administered to privately-insured patients at the nation’s cancer centers.