Machines don’t have eyes, but you wouldn’t know that if you followed the progression of deep learning models for accurate interpretation of medical images, such as x-rays, computed tomography (CT) ...and magnetic resonance imaging (MRI) scans, pathology slides, and retinal photos. Over the past several years, there has been a torrent of studies that have consistently demonstrated how powerful “machine eyes“ can be, not only compared with medical experts but also for detecting features in medical images that are not readily discernable by humans. For example, a retinal scan is rich with information that people can’t see, but machines can, providing a gateway to multiple aspects of human physiology, including blood pressure; glucose control; risk of Parkinson’s, Alzheimer’s, kidney, and hepatobiliary diseases; and the likelihood of heart attacks and strokes. As a cardiologist, I would not have envisioned that machine interpretation of an electrocardiogram would provide information about the individual’s age, sex, anemia, risk of developing diabetes or arrhythmias, heart function and valve disease, kidney, or thyroid conditions. Likewise, applying deep learning to a pathology slide of tumor tissue can also provide insight about the site of origin, driver mutations, structural genomic variants, and prognosis. Although these machine vision capabilities for medical image interpretation may seem impressive, they foreshadow what is potentially far more expansive terrain for artificial intelligence (AI) to transform medicine. The big shift ahead is the ability to transcend narrow, unimodal tasks, confined to images, and broaden machine capabilities to include text and speech, encompassing all input modes, setting the foundation for multimodal AI.
Asymptomatic infection seems to be a notable feature of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the pathogen that causes coronavirus disease 2019 (COVID-19), but the prevalence ...is uncertain.
To estimate the proportion of persons infected with SARS-CoV-2 who never develop symptoms.
Searches of Google News, Google Scholar, medRxiv, and PubMed using the keywords
,
,
,
,
,
, and
.
Observational, descriptive studies and reports of mass screening for SARS-CoV-2 that were either cross-sectional or longitudinal in design; were published through 17 November 2020; and involved SARS-CoV-2 nucleic acid or antibody testing of a target population, regardless of current symptomatic status, over a defined period.
The authors collaboratively extracted data on the study design, type of testing performed, number of participants, criteria for determining symptomatic status, testing results, and setting.
Sixty-one eligible studies and reports were identified, of which 43 used polymerase chain reaction (PCR) testing of nasopharyngeal swabs to detect current SARS-CoV-2 infection and 18 used antibody testing to detect current or prior infection. In the 14 studies with longitudinal data that reported information on the evolution of symptomatic status, nearly three quarters of persons who tested positive but had no symptoms at the time of testing remained asymptomatic. The highest-quality evidence comes from nationwide, representative serosurveys of England (
= 365 104) and Spain (
= 61 075), which suggest that at least one third of SARS-CoV-2 infections are asymptomatic.
For PCR-based studies, data are limited to distinguish presymptomatic from asymptomatic infection. Heterogeneity precluded formal quantitative syntheses.
Available data suggest that at least one third of SARS-CoV-2 infections are asymptomatic. Longitudinal studies suggest that nearly three quarters of persons who receive a positive PCR test result but have no symptoms at the time of testing will remain asymptomatic. Control strategies for COVID-19 should be altered, taking into account the prevalence and transmission risk of asymptomatic SARS-CoV-2 infection.
National Institutes of Health.
The emerging field of mobile health Steinhubl, Steven R; Muse, Evan D; Topol, Eric J
Science translational medicine,
04/2015, Letnik:
7, Številka:
283
Journal Article
Recenzirano
Odprti dostop
The surge in computing power and mobile connectivity have fashioned a foundation for mobile health (mHealth) technologies that can transform the mode and quality of clinical research and health care ...on a global scale. Unimpeded by geographical boundaries, smartphone-linked wearable sensors, point-of-need diagnostic devices, and medical-grade imaging, all built around real-time data streams and supported by automated clinical decision-support tools, will enable care and enhance our understanding of physiological variability. However, the path to mHealth incorporation into clinical care is fraught with challenges. We currently lack high-quality evidence that supports the adoption of many new technologies and have financial, regulatory, and security hurdles to overcome. Fortunately, sweeping efforts are under way to establish the true capabilities and value of the evolving mHealth field.
The development of medical applications of machine learning has required manual annotation of data, often by medical experts. Yet, the availability of large-scale unannotated data provides ...opportunities for the development of better machine-learning models. In this Review, we highlight self-supervised methods and models for use in medicine and healthcare, and discuss the advantages and limitations of their application to tasks involving electronic health records and datasets of medical images, bioelectrical signals, and sequences and structures of genes and proteins. We also discuss promising applications of self-supervised learning for the development of models leveraging multimodal datasets, and the challenges in collecting unbiased data for their training. Self-supervised learning may accelerate the development of medical artificial intelligence.
Heart rate is routinely measured as part of the clinical examination but is rarely acted upon unless it is well outside a population-based normal range. With wearable sensor technologies, heart rate ...can now be continuously measured, making it possible to accurately identify an individual's "normal" heart rate and potentially important variations in it over time. Our objective is to describe inter- and intra-individual variability in resting heart rate (RHR) collected over the course of two years using a wearable device, studying the variations of resting heart rate as a function of time of year, as well as individuals characteristics like age, sex, average sleep duration, and body mass index (BMI).
Our retrospective, longitudinal cohort study includes 92,457 de-identified individuals from the United States (all 50 states), who consistently-over at least 35 weeks in the period from March 2016 to February 2018, for at least 2 days per week, and at least 20 hours per day-wore a heart rate wrist-worn tracker. In this study, we report daily RHR and its association with age, BMI, sex, and sleep duration, and its variation over time. Individual daily RHR was available for a median of 320 days, providing nearly 33 million daily RHR values. We also explored the range in daily RHR variability between individuals, and the long- and short-term changes in the trajectory of an individual's daily RHR. Mean daily RHR was 65 beats per minute (bpm), with a range of 40 to 109 bpm among all individuals. The mean RHR differed significantly by age, sex, BMI, and average sleep duration. Time of year variations were also noted, with a minimum in July and maximum in January. For most subjects, RHR remained relatively stable over the short term, but 20% experienced at least 1 week in which their RHR fluctuated by 10 bpm or more.
Individuals have a daily RHR that is normal for them but can differ from another individual's normal by as much as 70 bpm. Within individuals, RHR was much more consistent over time, with a small but significant seasonal trend, and detectable discrete and infrequent episodes outside their norms.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
Machines and empathy in medicine Topol, Eric J
The Lancet (British edition),
10/2023, Letnik:
402, Številka:
10411
Journal Article
Recenzirano
...the patient–doctor relationship is affected by the limited time allotted for a clinic visit or bedside rounds and the way clinicians need to update electronic records during a consultation (figure ...A). Large language models (LLMs) of clinic visit conversations could offload many data clerk functions, such as health insurance pre-authorisations, scheduling of follow-up visits, tests, and prescriptions. Even though the ChatGPT compared with Reddit doctor study showed high quality and accurate responses, the questions informing the model were limited. ...there is a danger of mistaken answers and advice from these models, as has been seen with erroneous chatbot responses to diet questions from people with eating disorders.
Telemedicine 2020 and the next decade Dorsey, E Ray; Topol, Eric J
The Lancet (British edition),
03/2020, Letnik:
395, Številka:
10227
Journal Article
Recenzirano
A small randomised controlled trial of acutely ill patients compared hospital versus home care involving audio and video calls with clinicians and remote monitoring of vital signs showed there were ...fewer readmissions, less unnecessary testing and consultations, and lower costs for home care. For chronic conditions, virtual and in-person care might be integrated with a diverse set of health workers (physicians, nurses, dietitians, therapists) who could provide care centred around a patient's home. Faith Hark/Scripps Research Translational Institute ERD is supported by the US National Institutes of Health grant 5P50NS108676-02 and receives consulting fees from American Well (a US telemedicine company), serves as editor for Digital Biomarkers, and has stock ownership in and is on the Medical Advisory Board of Grand Rounds (a remote second opinion service for large employers).
Medical forecasting Topol, Eric J
Science (American Association for the Advancement of Science),
05/2024, Letnik:
384, Številka:
6698
Journal Article
Recenzirano
"AI-Powered Forecasting" was recently on the cover of
, highlighting a new deep learning model for much faster and more accurate weather forecasting. Known as GraphCast, it outperformed the ...gold-standard system and had an accuracy of 99.7% for tropospheric predictions, the most important forecasting region that is closest to Earth's surface. Better warnings for extreme weather events such as hurricanes and cyclones will help save lives. The parallel in medicine is forecasting specific, actionable, high risk for individuals to prevent diseases or severe acute events. But we don't have a gold standard for predicting health outcomes. That is hopefully about to change.
...ejection fraction and many types of valvular heart disease (and severity) were accurately predicted by training an AI model with over 22 000 paired echocardiograms and chest x-rays. Beyond CT ...scans, mammography has information about risk of heart disease via presence of breast artery calcification. ...what cardiologists cannot detect from an electrocardiogram—eg, risk of atrial fibrillation, heart attack, stroke, diabetes, kidney disease, anaemia, and filling pressure of the left heart—have been reported with AI models.
PDF
