Whether we are ready to accept it or not, Artificial Intelligence (AI) is already solving health care problems and revolutionizing the industry.
Valuable time and money are lost daily in the industry because of misdiagnoses, delays, and poor communication. Resources are wasted on something that could have been detected earlier, and, not only does it put a financial strain on the industry, but both the patient and the physician are strained by the results. There is a solution in the horizon: Artificial Intelligence.
The question thus becomes: How can artificial intelligence transform healthcare?
The Solution to Healthcare Complications
Computers can analyse data much faster than humans can. They are able to see patterns and algorithms which can enhance disease diagnosis and overall public health, much faster. With the ability to learn algorithms, machines can be greatly utilized to process the abundance of data that forms part of the healthcare industry. The more data it receives, the better it is able to improve the industry in the future.
Here are 5 ways in which AI is already transforming the healthcare industry.
1) Identifying diseases and providing diagnosis
Heart disease is one of the leading causes of death around the world. Because of this a lot of time has been invested in developing machine learning in such a way to pick up on algorithms which can predict when people might be at risk. The results have been significantly more accurate than those made by other guiding systems. This is the case with Google and its healthcare technology subsidiary, Verily, where scientists created an algorithm that can predict this disease by looking at the back of a person’s eyes, and accurately pinpoint early signs of heart conditions. In a published article citing its findings, these scientists used data from over 300,000 patients, along with a deep learning algorithm to conduct their research.
Other cases in which AI is leading by leaps and bounds is in computer imaging. This has lead to advance algorithms being used to effectively detect melanoma, one of the deadliest forms of skin cancer. This disease has a survival rate between 15% to 65% from early stages, to terminal stages. In some studies, this technology has been able to produce a 98% survival rate after five years.
2) Crowdsourcing Treatment Options and Monitoring Drug Response
In order to reap the rewards of AI, we require data, and lots of it. Past attempts to understand patients and treatment options were extremely limited due to the many obstacles of obtaining data points in a way that was actionable, unobtrusive, and across multiple spectrums of the patient record. Nowadays, because of the advancement of wearable devices, mobile applications, and improved data interoperability thanks to new standards, such as FHIR, data scientists and clinicians are beginning to leverage the power of machine learning in order to gain insights never seen before, and create customizable treatments that can provide better results.
Similarly, the way in which clinical information is gathered and and aggregated with other datasets, is crucial to successful diagnosis and treatment. In this capacity, AI is being used on social media to connect people in sharing different treatment options and to source information about different drug trials that might be available. This allows for the flood gates of data to open up, and the ability for data scientists to continue to track results once a patient is outside a clinical setting.
An interesting study conducted by AstraZeneca showed how the use of a recurrent neural network (RNN), a data algorithm perfect for data that changes over time, helped scientists train generative models for molecular structures, in order to enrich libraries towards a biological target.
An although AI can accelerate the finding of new treatments and drugs, the medical community may be slow to adopting them without some hard street evidence, and the backing of the pertinent medical bodies.
3) Monitoring Health Epidemics
Sifting through mountains of government intelligence, millions of posts on social media sites, and countless news feeds, computers are able to gather a myriad of sparse data, that when combined with ecological, biogeographical, and public health information, it can help authorities pick up on current health threats as they begin happen. One such example is the Ebola outbreak in 2014 in which filoviral hemorrhagic fever .
According to the National Center for Biotechnology Information, “filoviral hemorrhagic fever is associated with multiple hemorrhagic manifestations, marked hepatic involvement, disseminated intravascular coagulation, and shock. Patients who eventually recover have a fever for about five to nine days, while in cases resulting in death, clinical signs develop early, with death occurring between days six and sixteen. Mortality is high and varies between 30 and 90%, depending on the virus.”
And this is what drove a team of scientists and researchers to develop artificial intelligence algorithms to identify species of bats that were more likely to carry the disease, and therefore predict a future outbreak. In the study conducted by Barbara Han, a disease ecologist at the Cary Institute of Ecosystem Studies, they were able to successfully predict outbreaks with an 87% accuracy rate. This is an example that clearly illustrates what AI is extremely effective at accomplishing, classifying and predicting the unknown.
By applying the power of AI globally, we can create one of the most comprehensive disease database encompassing all known parasites and pathogens, in both animals and humans.With this type of information at our disposal, a great many lives can be saved.
4) Virtual assistants help patients and physicians communicate clearly.
Patient engagement in healthcare, just as in commercial marketing efforts to retain and educate consumers, fosters trust between patients and the medical professionals treating them, and is key for maintaining the continuity of care. As a result, this leads to many long-term benefits to all involved parties, including the insurance companies, as well as the physicians.
And one of the biggest challenge in any initiatives has been patient adoption of new technology that was designed to improve communication. This has been the case with some approaches involving mobile technology or patient portals, that although they are supposed to enhance the overall experience, have always placed the burden on the patient of following multiple steps and lengthy questionnaires. And this is where AI can bring a new paradigm into how we interact with patient. Advancements in machine learning have facilitated the creation of software that learns how to best communicate with the patient, synthesizes complex medical terminology and concepts into digestible bits of information. This creates a frictionless approach, and enables physicians to quickly access information or results needed for the patient.
Some samples of how AI is making it easier for patients to communicate with their doctors is by using machine learning algorithms to better craft a message that will elicit a more desired response from the patient, such as in medication adherence, or post-discharge follow ups. Another great example if how ZocDoc uses AI to verify health insurance eligibility before the patient sees their doctor, and ensuring they are better informed about their coverage.
Another AI technology that is gaining popularity is the virtual assistant. Without having to look for lab results, a physician can use a voice command to request a virtual assistant to quickly draw up the information required. This allows the doctor to spend more time with the patient, leaving them more satisfied with the experience.
Other AI examples include the use computer vision technology, in which trained AI algorithms can monitor multiple patients and assist personnel to effectively prevent falls (disclaimer, I worked on this project), and allow staff to remotely communicate with the patient to ease their time while in PACU or ICU.
5) Improving clinical documentation in order to develop better care management.
Computers are really good at looking at thousands, if not, millions of rows of data in a fraction of a second, and be able to discern patterns, classify information, and predict with infalible precision what an outcome will be. As access to high-performance advance AI algorithms, we can quickly process data points that at first glance will appear to be disconnected, such as user written patient notes, Machine Learning is used to develop tools which pick up on irregularities, since they have the ability to compare hundreds of thousands of similar notes, determine what the outcome was, and what provide insights into what delivered positive outcomes. Likewise, doctors can be notified of any missing data or if clarification is needed for any procedures, in order to provide better analysis.
Deep Neural Networks (DNN) are algorithms that can recognize very complex patterns in data with multiple inputs, such as the information contained in written clinical documentation. Deep nets can find concepts in data that are virtually undetectable to humans, due to the numerous nodes that need to be connected in order to synthesize information that can then be classified, and applied towards known medical protocols.
This is what DNN’s are really good at, finding hidden relationships between “metaphors”. Take for example someone mentioning “the king of rock and roll” or “the king of pop”, we could argue that the names Elvis Presley and Michael Jackson automatically come to mind. Our brains are wired to automatically relate these simple metaphors with a subject, and an industry. Now imagine finding concepts amongst hundreds of thousands of clinical notes filled with countless of medical metaphors, the size of the metaphorical spider web of medical conditions and their care becomes to be staggering.
What Comes Next?
As AI continues to gain support amongst medical experts, it is without any doubt that its value will continue to increase as new use cases are discovered and problems solved. And just as with any new disrupting technology, it is imperative that we prepare for its use appropriately. Though there are many advantages in one hand, there are also pitfalls we need to guard against. And, if we can successfully avoid these, we can expect huge medical discoveries and treatment innovations will continue to be made.
What’s been your experience with Artificial Intelligence? Do you see any other areas of healthcare that can greatly benefit from its adoption? Share your thoughts, would love to hear your ideas.
