As we turn the corner on 2020, we find ourselves in the midst of the worst pandemic in a Century and our health care systems have been pushed to the brink of failure. To some of us it has been clear for several years that the health care system desperately needs disruption, but COVID-19 has made that easier to appreciate.
Artificial intelligence is one of the biggest coming disruptions in healthcare. For some AI conjures images of a malevolent super intelligence that will outperform humanity across the board. For others AI represents a confluence of machine learning algorithms and petabytes of data that allows approximations of human decision making. Much like smart phones and app revolutionized healthcare in the preceding decade, AI is healthcare’s next leap forward
Progress in the field of AI has been powered by petabytes of data and deep neural networks (layered networks inspired by the human cerebral cortex) with tremendous parallel computing capabilities. In some areas like speech recognition, AI has become remarkably capable (hey Siri, sudo make me a sandwich).
Until we have enough annotated training data and human like intelligence that can transfer knowledge and learnings from one domain and transfer them to another, AI in healthcare is very dependent on supervision by human doctors. A collaboration between AI start-ups and health care systems is giving deep learning networks tens of thousands of human labelled examples and the future is bright.
There are several areas where AI has already made a huge impact in healthcare. Let’s look at some of them.
Electronic health records
AI powered platforms take medical voice recording and use natural language processing to make concise and harmonized notes. AI can look at a patients EHRs and predict future health outcomes like heart attacks and prevent drug interactions. Medical learning classifiers can label EHRs, such that deep learning networks can use this data to freak diagnostic AI technologies.
Radiology is a great example an AI’s disruption of healthcare. The average human radiologist has a reference data set of a few thousand images. AI access a reference bank of millions of images. Advances in image and pattern recognition allow AI to see shades of grey that human eyes may never perceive and identify a lesion.
Testing platforms are increasingly using AI to improve their simplicity, accuracy and versatility. Due to COVID-19 many have become familiar with molecular tests like PCR, CRISPR and next generation genomic and protein sequencing. Data generated from these platforms is fed to AI to build models, discover new treatments and to understand the molecular basis of human disease. AI has been used to understand COVID-19, discover new treatments and even in vaccine development!
Much of the focus on AI in healthcare is on how AI can make help doctors, hospitals and labs and thereby provide in detect benefit to patients. However, AI is also powering a massive patient centered revolution in health care – a do it yourself approach to health that empowers patients to take charge of their health.
Diabetes is a great example of patient empowerment positively impacting health. For years empowered diabetic patients have been able to use glucose meters to manage their glycemic control themselves. Empowered diabetic patients who can change their doses in real time have better outcomes than patients dependent on their physicians to modify their insulin doses once every few months. The rise of AI is powering similar DIY movements across healthcare. Not every minor health intervention requires a doctor, but all interventions require patients empowered by evidence, data and knowledge of all available options. AI systems combined with home testing can provide excellent insights for cancer screening, sexual health, blood pressure and blood glucose control, nutrition and weight loss. These systems allow patients to get care and testing their own homes. Although labeled DIY, they are monitored by human doctors, and in fact they can provide doctors with incredibly deep insights about their patients.
Telemedicine, wearables, home devices and the internet of things are revolutionizing the insights patients can share with their doctors. Temperature, heart rate, blood pressure, oximetry, blood glucose and sleep data can be collected all from the comfort of one’s home. Thanks to AI and the internet, doctors can get automated alerts about their patients and can provide immediate advice. AI can help correlate data from wearable devices, sleep trackers, exercise and diet logs with a patient’s complaint, examination findings and lab results.
AI in health care isn’t always for the benefit of patients. Health insurance companies use AI to find increasingly innovative and frustrating ways to reject medical insurance claims. AI also has inherent limitations and biases which leads to ethical concerns. If the data set has biases, then the learnings based on that data will also have similar biases. Furthermore, the system of medicine itself has biases that lead to similar issues in AI systems. Building intelligent systems that can think like humans, but are free from human biases is challenging and most AI systems are years away from the decision making that we expect from human physicians. We need artificial intelligence systems to be artificial philosophers as much as we need them to be artificial data scientists.
Part of the disruption, we need in health care is making it more accessible, affordable and equitable. AI systems should not worsen the problems of inequity and accessibility in health care. Many of healthcare’s problems transcend technology and they need a human touch. AI is not a cure all to fix everything that ails healthcare and part of the challenge lies in correctly identifying those areas of healthcare where AI will provide the right kind of disruption.
The rise of AI in healthcare often becomes about doctors, clinicians, nurses, hospitals and labs. We don’t put the patients at the center of the system. Health data is owned by the patient, the EHR is for the patient and the health outcome impacts the patient. Perhaps the real disruption will be health systems that are powered by artificial intelligence, but have wisdom that makes them human centered. Health systems of the future will be AI powered but they must be human centered.
Disclaimer: The views expressed in the article above are those of the authors’ and do not necessarily represent or reflect the views of this publishing house. Unless otherwise noted, the author is writing in his/her personal capacity. They are not intended and should not be thought to represent official ideas, attitudes, or policies of any agency or institution.