Digital Medicine

Here we argue that now is the time to create smarter healthcare systems in which the best treatment decisions are computationally learned from electronic health record data by deep-learning methodologies.

In this Comment, we provide guidelines for reinforcement learning for decisions about patient treatment that we hope will accelerate the rate at which observational cohorts can inform healthcare practice in a safe, risk-conscious manner.

A primer for deep-learning techniques for healthcare, centering on deep learning in computer vision, natural language processing, reinforcement learning, and generalized methods.

To bring the full potential of AI to the clinic, practical and regulatory changes need to be made in health systems globally.

The increased amount of health care data collected brings with it ethical and legal challenges for protecting the patient while optimizing health care and research.

Artificial intelligence is beginning to be applied in the medical setting and has potential to improve workflows and errors, impacting patients and clinicians alike.

As Nature Medicine celebrates its 25th anniversary, we bring you a special Focus on Digital Medicine that highlights the new technologies transforming medicine and healthcare, as well as the related regulatory challenges ahead.

Exploring the evolution, from analysis and endpoints to registration and regulations

Genetics and electronic health records come together to identify heritable traits

Machine learning makes new sense of psychiatric symptoms

Rima Arnaout is an assistant professor of cardiology and a member of the University of California San Francisco Bakar Computational Health Sciences Institute. She has received a Chan Zuckerberg Biohub Intercampus Research Award, as well as funding support from the US National Institutes of Health and the American Heart Association’s Institute for Precision Cardiovascular Medicine.

Kee Yuan Ngiam is the group chief technology officer at National University Health System, Singapore, and assistant professor at the School of Medicine of the National University of Singapore. His research focuses on the effects of using artificial intelligence in healthcare. He is the 2018 recipient of Singapore’s National Health IT Excellence Award, which recognizes individuals who advanced healthcare through innovation.

Deep-learning algorithms can be applied to large datasets of electrocardiograms, are capable of identifying abnormal heart rhythms and mechanical dysfunction, and could aid healthcare decisions.

A deep-learning algorithm, trained on over 17,000 real-world patient facial images, achieves high accuracy in identifying rare genetic disorders.

Analysis of electrocardiograms using an end-to-end deep learning approach can detect and classify cardiac arrhythmia with high accuracy, similar to that of cardiologists.

A deep learning algorithm applied to the electrocardiogram—a test of the heart’s electrical activity—can detect abnormally low contractile function of the heart, opening up the possibility for a simple screening tool for this condition.

A convolutional neural network model using feature extraction and machine-learning techniques provides a tool for classification of lung cancer histopathology images and predicting mutational status of driver oncogenes

A deep-learning algorithm is developed to provide rapid and accurate diagnosis of clinical 3D head CT-scan images to triage and prioritize urgent neurological events, thus potentially accelerating time to diagnosis and care in clinical settings.

A novel deep learning architecture performs device-independent tissue segmentation of clinical 3D retinal images followed by separate diagnostic classification that meets or exceeds human expert clinical diagnoses of retinal disease.

A reinforcement learning agent, the AI Clinician, can assist physicians by providing individualized and clinically interpretable treatment decisions to improve patient outcomes.

AI is used increasingly in medical diagnostics. Here, the authors present a deep learning model that masters medical knowledge, demonstrated by it having passed the written test of the 2017 National Medical Licensing Examination in China, and can provide help with clinical diagnosis based on electronic health care records.

Selection of the right cancer treatment is still a challenge. Here, the authors introduce a framework to analyze treatment benefits, using the idea that patients with similar genetic tumor profiles receiving different treatments can be used to model their responses to the alternative treatment.

Anemia has a global prevalence of over 2 billion people and is diagnosed via blood-based laboratory test. Here the authors describe a smartphone app that can estimate hemoglobin levels and detect anemia by analyzing pictures of fingernail beds taken with a smartphone and without the need of any external equipment.

In this Review, Yacoub and McLeod summarize the rationale for monitoring patients with heart failure or pulmonary arterial hypertension to detect haemodynamic changes that predict the deterioration from subclinical to overt disease, the transition from noninvasive to implantable devices and the current and anticipated clinical use of these devices.

Computational models are increasingly used in cardiology to integrate multiple data sets from individual patients and create virtual-patient simulations. In this Review, Niederer and colleagues discuss how multi-scale models of cardiac electrophysiology and mechanics can support diagnostic assessment and clinical decision-making and pave the way to personalized cardiac care.

Mental health technologies, such as apps, clinical texting, social media platforms and web-based tools, have arrived. Channelling these resources to help people with serious mental illnesses, clinicians in need of support, and people in low-and middle-income countries will have the most impact on the global burden of mental illness.

Just et al. develop a highly accurate biological classification method for identifying suicidal ideators by applying machine learning to neural representations of death- and life-related concepts.

Site-less trials promise to speed up drug development — but obstacles to in-home data collection abound.

Clinical trials for Alzheimer disease drugs have an exceptionally high failure rate, discouraging investment in the field despite the unmet medical need. Drug developers need to more effectively harness existing and emerging data and digital technologies to improve the likelihood of success.

Information Exchange and Data Transformation (INFORMED), a multidisciplinary initiative anchored in the FDA Oncology Center of Excellence, is a decentralized science and technology incubator designed to harness the power of big data and advanced analytics to improve disease outcomes.

Analysis of genetic data and blood lipid measurements from over 300,000 participants in the Million Veteran Program identifies new associations for blood lipid traits.

Genome-wide association analysis using electronic health record data from >94,000 individuals identifies loci associated with plasma lipid concentrations. Longitudinal measurements allow for the calculation of genetic risk scores and increase the variance explained.