Will artificial intelligence change humanity through healthcare?
Artificial intelligence can take many forms in healthcare and has the potential to augment human activities in pharmaceuticals, cancer diagnostics, medical imaging, medical research, electronic health records, risk management and many more areas.
The application of AI to healthcare also carries the promise of significant monetary savings. According to an Accenture report: Artificial Intelligence: Healthcare’s New Nervous System, by 2026, AI could account for up to $150 billion in annual savings to the U.S. healthcare economy.
In this post, we discuss potential ways in which AI can disrupt healthcare, but not before we list some of the main benefits of AI that make it such a powerful disruptor:
- Optimizing clinical and medical workflows.
- Enabling new insights by the ability to identify subtle patterns that are difficult for humans to spot and by the ability to integrate and process vast amounts of data from multiple sources.
- Augmenting decision making.
With these benefits in mind, we look at six use cases of AI in healthcare, which demonstrate the impact AI will have on our immediate future.
Eradicating infectious diseases
Infectious diseases have plagued humanity since the rise of the first agrarian societies. Even following the fairly recent discovery of antibiotics and other significant medical and scientific advancements, treating and preventing infectious diseases takes a toll on public health systems’ resources.
The impact on public well-being and on health economics is greatest during disease outbreaks. Therefore, detecting and managing outbreaks before they reach epidemic proportions has been a longtime goal of governments and health organizations around the world. For many years, such organizations have been collecting data and analyzing them to detect, or even predict outbreaks as early as possible.
Identifying this as a classic machine learning problem for large amounts of data, Google launched its Google Flu Trends web service in 2008 to provide estimates of influenza activity in more than 30 countries, with the aim of predicting outbreaks based on search query patterns. Although the project was terminated in 2015, follow-up studies demonstrated that search engine and social media data can reduce errors in models generated solely on data collected by the CDC, for example, by more than 50% and the concept has since been developed further by others, such as Osnabrück University and IBM Watson (www.flu-prediction.com project).
Realizing the potential of EHRs for efficiency gains
The introduction of electronic health records (EHRs) had a significant impact on patient care quality. It drastically reduced the use of paper files, increased patient information accessibility and fidelity, removed geographical boundaries and enabled easy sharing. It contributed to improved diagnostics based on patient history and reduced the number of medical errors. It also went beyond the individual patient, allowing for systematic collection of health data for entire populations, providing a birds-eye view of public health trends and risk factors.
A less positive result of EHR popularity is that many healthcare providers now spend more time on documentation than on face-to-face care and assessment of the patient's condition and concerns. As personal connection and engagement between the patient and the provider are critical for effective diagnosis and care, EHR-integrated virtual assistants are being introduced into the healthcare industry. One example is Eva, by eClinicalWork. The virtual assistant is powered by AI-driven voice recognition, reducing the data entry burden on physicians, by allowing them to dictate notes and make more time for direct patient contact.
Another type of AI application, which is also often available on mobile platforms, takes advantage of the availability of vast amounts of data. K Health, for example, can suggest an initial diagnosis or recommend seeing a professional care provider based on symptoms reported by the patient, by comparing them with millions of records and finding cases that match not only the symptoms, but also the patient’s age, gender and other relevant parameters. Such applications have the potential of reducing the burden on entire health systems.
Similarly, AI can be used to analyze data from EHRs for more complex insights, such as risk prediction and stratification, therapy optimization, clinical study matching and so on, holding great promise for financial gain, as well as improved patient care.
Creating more jobs in the healthcare industry
Some voices against the adoption of AI in healthcare are concerned the use of AI may eliminate jobs by replacing doctors, nurses, technicians and more. In all likelihood, the opposite is true.
PwC research has revealed that in the next 20 years, approximately 7 million existing jobs could be displaced due to AI, but around 7.2 million could be created in the UK alone. Healthcare is one of the sectors where PwC expects to see a rise in employment due to AI adoption.
Taking cancer diagnostics to the next level through the enhancement of anatomic pathology
AI is expected to play a critical and growing role in diagnostics. In particular, cancer diagnostics is an area of high interest.
Cancer is primarily diagnosed by examining the morphological and functional properties of human tissue, often through a microscope. The specimen can tell the pathologist whether a malignancy is present and, if so, which stage of cancer it is in. In turn, the prognosis and recommended therapy protocol can be inferred by the treating clinician. Much of this work is performed manually by pathologists.
While cancer incidence keeps increasing every year, the number of pathologists does not grow proportionally, resulting in an unsustainable, ever-growing workload with which pathologists and pathology labs are faced. Without a breakthrough, every specimen will receive less attention, resulting inevitably in more diagnostic errors and, therefore, harming patients.
AI can provide the necessary breakthrough, making cancer diagnostics more accurate and less prone to error, while reducing pathology lab workloads and driving better patient outcomes and better health related quality of life.
The Ibex Second Read™ system is using AI to analyze prostate biopsies in conjunction with human pathologists. It provides a safety net to pathologists, by alerting them when there is suspicion of a diagnostic error. The system is the first to be deployed in routine clinical use. Moreover, it has already diagnosed missed cancers and has led to revised diagnoses and patient care.
Making medical imaging an even more effective diagnostic tool
Medical imaging is a cornerstone of diagnostics. Without X-ray, ultrasound, MRI, CT and other imaging technologies, many diseases and conditions could not be detected early enough to administer effective treatment, not to mention the significant part such technologies play in medical research, advancing our knowledge of how diseases work and how we can better fight them.
The benefits of applying AI to diagnostic imaging are many, including quicker workflows, real-time analytics, less time spent on tedious tasks allowing radiologists to care for more patients and more accurate interpretation of scans, based on objective and reproducible algorithms. But more than making radiologist more effective and efficient at what they already do, AI can help in the detection of conditions that are too early or too small for human pathologists to notice, providing the potential to treat diseases earlier than ever before, thereby improving patient outcomes.
One example of AI being used to better detect tuberculosis (TB) in remote areas demonstrates its enormous potential. Researchers at the Thomas Jefferson University Hospital in Philadelphia are training AI algorithms to identify TB in chest X-ray images. This can be extremely useful to patients in remote areas who tend to be more prone to TB and lack access to trained radiologists. The use of algorithms for this purpose can greatly contribute to early detection of the disease and consequently to better treatment.
Another great example is the way Viz.ai uses AI to analyze CT data and alert physicians in real time when a stroke is identified in one of their patients’ scans. In the event of a stroke, every minute can be crucial to the patient’s survival and subsequent physical and mental condition. Viz.ai can shorten the time it takes from obtaining the CT scan to alerting the physician by up to 91%, greatly increasing patients’ chances for a more favorable outcome.
Reinventing pharmacological and medical research
Research institutes and research divisions of pharmaceutical companies are, by definition, at the forefront of new technology adoption. The path to new medical discoveries, drugs in particular, is long and expensive. That is why pharma and biotech companies are very open to adopting new technologies that can help them make discoveries and develop new therapies faster and more cost-effectively. AI allows researchers to process great amounts of data, both structured and unstructured, and analyze them faster and in innovative ways, in order to obtain valuable insights. In turn, more intelligent choices can be made on research directions and therapies can be easier to develop alongside highly specific tests that can predict a patient’s potential benefit.
A couple of fascinating examples:
- Pfizer is using IBM’s Watson to analyze large quantities of public and private data, helping researchers identify new drug targets and alternative drug indications.
- Cytoreason is a company that applies AI and machine learning to analyze the human immune system to discover complex patterns in biological data. It also uses AI to support or refute various research hypotheses, as outlined by research institutes.
Breakthroughs in healthcare influence humanity on a global scale. AI is expected to have an immensely powerful impact on healthcare, through diagnostics, medical research, disease risk prediction, patient care and virtually every other imaginable area. Propelled by major advancements in Machine Learning and Deep Learning in recent years, it seems like we are on the cusp of a major disruption to healthcare. Medical practitioners will have better decision support tools at their disposal and will be able to make decisions at a higher-than-ever rate, without sacrificing personal interaction or financial performance. As a result, patients everywhere will enjoy improved healthcare services and will live longer, better lives.