Top 20 AI Use Cases: Artificial Intelligence in Healthcare
Artificial intelligence and machine learning are proving to be essential for healthcare organizations.
Artificial intelligence (AI) and machine learning (ML) have been instrumental in revolutionizing nearly all industry sectors, but specifically in healthcare—there's a noticeable impact being felt around the world. Real-world AI use cases show AI and ML are essential for many, if not all, healthcare organizations' future.
In fact, even the smallest step forward with AI and ML in medical technology can save hundreds, if not thousands, of human lives.
It comes as no surprise that healthcare providers, life science companies, and health technology vendors are going to spend $54 million on AI projects by 2020 according to a CB Insights 2016 report. In fact, that same report showed that 86% of these organizations were already using some form of AI.
Let’s have a look at the most interesting (and sometimes simply amazing) AI use cases in healthcare.
AI Use Case #1: DynaLIFE and AltaML’s Colon Polyp Project to Begin Pathology Digitization
The digitization of pathology (courtesy of—you guessed it—AI and ML) is seeing massive improvements to applications for diagnostic healthcare with a breakthrough project focusing on colon polyp identification and analysis. Patients and industry pros will experience greater efficiency and scalability in the laboratory environment, plus better patient outcomes.
While efficiency is a huge driver, digitization will also make the work of pathologists less difficult. Currently, colon tissue samples are biopsied by a gastroenterologist and then sent to the lab to be embedded in a paraffin block and sectioned to produce glass slides for interpretation by a pathologist.
Well, because of the small size of polyps, and even smaller biopsy samples, the initial sections of the tissue may not be entirely clear. More time and effort is spent obtaining deeper sections of the tissue block, which causes longer delays in the diagnosis.
But by using ML on digital images of colon biopsies, this offers potential for better triage of such cases, and a much quicker (and more efficient) diagnosis.
AI Use Case #2: Avoiding Unnecessary Surgery
It's not infrequent for patients to undergo surgeries which may later prove to be unnecessary. These unneeded procedures force people to face totally avoidable risks and burdensome surgical expenses.
Health plans often include programs that help patients and their local healthcare teams make better informed decisions by educating them on the latest treatment options available. However, it’s hard to achieve adequate engagement in these programs, especially when patients are in a crisis situation already.
ConsumerMedical developed a new program that leverages AI and predictive analytics to identify patients considering surgery, and evaluates whether these operations are actually necessary or if there are other alternatives with far less risk involved that would better help them in the long run. Earlier identification of people on the pathway to surgeries doubles the average program engagement rate and reduces unnecessary surgeries.
AI Use Case #3: Diagnosing Infectious Diseases
Machine learning and big data may help change the way we diagnose infectious diseases forever. That’s the case of the technology implemented by Aperiomics, a biotech firm in the Washington, DC area that developed an AI-based method to analyze organic samples for diseases at an incredible speed.
Using “deep shotgun metagenomics”, Aperiomics is able to take any genomically sequenced organic sample and identify all 37,000 known and sequenced pathogens. Using a proprietary database and ML algorithms, they are able to relatively quickly run standard genetic samples through their system and create a comprehensive profile of everything in the sample, identifying what should and should not be there and in what quantities they were found.
This work, which would be impossible without the power of ML and AI, means that patients who have difficult to diagnose infectious diseases are finally able to get a diagnosis and treatment.
This technology has been used by sufferers of chronic urinary tract infections and to detect potentially lethal cases of blood sepsis, severe gastrointestinal infections and more.
AI Use Case #4: Enhancing Medical Imaging Analysis
Even the best physician may miss a minuscule detail in a CT scan or echocardiogram. Especially if a doctor is focused on searching information on a medical imaging report relating to a specific condition (i.e. to find data which may confirm a diagnosis), it’s completely normal to overlook signs of other unrelated diseases.
Some companies are harnessing the ability of AI for full algorithm-based analyses of a medical imaging reports which may improve diagnostic accuracy to 90%.
Ultronics’ EchoGo platform, for example is able to analyze nearly 80,000 data points from every echocardiogram image, while by Zebra Medical Vision’s Profound software can detect signs of breast cancer, fatty liver, aneurysms and emphysema.
The AI2 Incubator and Fujifilm SonoSite, instead, deployed deep learning models on portable ultrasound devices. The traditionally low quality of ultrasound images is improved so that this much more affordable diagnostic method can be used on patients without exposing them to unnecessary radiation.
AI Use Case #5: Virtual Nursing Assistants
Healthcare facilities which must deal with high volumes of patients face many challenges to keep care costs low and improve outcomes. Nurses, in particular, must often take care of too many patients at once, so any technology that could help them reduce their workload is always welcome.
Virtual nursing assistants are smart assistants that can perform many tasks, such as monitoring patients’s vitals remotely at any time of day, alerting providers and clinicians when a patient’s symptoms take an alarming dip, checking care plan adherence, or extending care outside of the clinical setting.
Virtual nurses are a futuristic merge of all the latest technologies, including ML, to compare vast databases of patient information, to computer vision, natural language processing (NLP), and even advanced robotics.
Both robotic and completely virtual smart nurses have already been developed and tested by companies such as Sensely and Softbank Robotics. They have a lot of cool names such as Molly, Romeo, and Pepper which coupled with their cute design and nice voices make them look and sound as human as possible.
AI Use Case #6: Helping Teenagers with Health And Sex Lives
AI-powered chatbots have been widely implemented by nearly all verticals where customer service is even remotely important. However, in the healthcare industry, things are often much more complicated when the topics that need to be discussed with the chatbot range from health to personal life and individual medical needs.
Planned Parenthood wanted to create an online tool that would help users — primarily teenagers — ask questions about sex, relationships and growing up in general. This is a challenging task as kids likely aren’t using correct anatomical terms to ask those questions. That’s where LogMeIn and Bold360 came in with its AI’s capability for NLP.
Using its NLP capabilities, Planned Parenthood’s chatbot Roo is able to understand the user’s true meaning generate and an appropriate, conversational response. Incredibly useful to break the ice with teens, Roo can overcome most communication barriers that come with the territory with these highly-sensitive topics.
In less than six months of activity, the bot answered nearly 800,000 questions from a user base that’s 81% teens.
AI Use Case #7: Detecting Dementia
According to a 2018 study, dementia is a serious condition that costs up to $500 billion USD (direct and indirect costs) worldwide every year. With nearly 1 in 3 children born in 2015 and beyond expected to suffer from dementia, early detection and diagnosis of dementia may save a whopping $118,000 USD per patient.
Cognetivity Neurosciences developed a sensitive diagnostic tool which is able to perform a full cognitive assessment test on Apple iOS devices.
During this 5-minute test, several natural images are briefly shown to participants who are asked to respond as quickly and accurately as possible to indicate whether they’ve seen a pre-specified image category.
Their answers are then compared with a dataset by the AI algorithms that will cluster test performance in terms of accuracy, speed and image properties. Depending on the answers, in fact, the performance of large areas of the brain can evaluated, detecting the earliest signs of dementia and other mental impairment which can also be monitored for progression remotely.
AI Use Case #8: Intelligent Robot-Assisted Surgery
Robot surgeons have already been tested for years since they can prevent many of the human errors associated with fatigue and exhaustion. Surgery procedures require, in fact, the utmost patience and precision, and the skill of mechanical surgeons does not falter even when they operate without pause for hours and hours.
AI combined with computer vision software can be used to achieve a new level of precision for even the most minute movements, allowing robot surgeons to perform procedures independently.
Human surgeons can also benefit from this technology since algorithm-assisted instrumentation can be used to perform procedures on a scale too small to be done by hand.
AI Use Case #9: Diagnosing Autism with an HIPAA-Compliant Platform
Cognoa is a digital behavioral health company that developed an AI-based diagnostic tool that is able to spot early signs of autism in children. With a 90% accuracy rate, this software is so precise and efficient that has already received Breakthrough Device Designation from the U.S. Food and Drug Administration (FDA), giving it priority review.
Other than training algorithms to improve outcomes and lower behavioral healthcare costs, however, Cognoa took a step forward to ensure that all sensitive data used to feed these algorithms is collected anonymously.
They hired Immuta, a company that developed a software platform that could help enforce data access roles, permissions, and policies which would meet all compliance requirements. Immuta's platform expedites the process of collecting data required by AI, and allows for the delivery of Health Insurance Portability and Accountability Act (HIPAA) compliant algorithms within the most stringent of regulatory environments.
AI Use Case #10: Improving Security and Reducing Fraud
A wise use of ML can be used to improve the security of patient data by securing it from unwanted accesses and other threats. AI can track all kinds of malicious activity such as potential hacks or suspicious accesses, helping security teams monitor only the most vulnerable activities and relevant leads.
AI may also monitor other types of transactions for potential anomalies in billing, such as signs of kickbacks, upcoding, downcoding or other fraudulent activities. Algorithms can spot and flag hidden patterns that deviate from the enormous amount of data that can be digested from EHRs, insurance claims, and yearly budgets.
Healthcare fraud costs about $68 billion USD every year in the U.S., so detecting any fraudulent activity as it occurs can help organizations save a significant amount of money which can be redirected to medical research, better treatment options and more.
AI Use Case #11: Automating Evaluation & Management (E&M) Scoring
NLP can be used in conjunction with AI to enhance the labor-intensive tasks involved in Facility and Professional evaluation & management (E&M) leveling. E&M is a medical coding process in support of medical billing. This allows medical service providers to document and bill for reimbursement for services provided.
Complex and error-prone processes can be standardized and automated by i10id, a product deployed by Trustedi10 which makes coding much more efficient and compliant, leading to increased revenue for hospitals and physicians that use it.
This product is smart enough to “read” all screens of documents/notes associated with a medical record. i10id converts a 24-minute process into a 16-second process, and automatically creates and stores audit defensible document should there be a request.
AI Use Case #12: Leveraging EHR Data to Improve Health Outcomes
Electronic health records (EHRs) are an often untapped goldmine of precious information about patients. They’re filled with info about a patient’s current status, past medical treatments and procedures, background and history, and much more.
However, this data is often not so readily accessible, and the most interesting insights lie hidden under a mountain of clutter.
AI can analyze EHRs to provide doctors and nurses with vital insights (or insights on their vitals), predict diseases up to one year before they appear, assist in prescribing the most effective medication, and avoid adverse drug interactions and other preventable medical errors which may harm patients.
AI Use Case #13: Reducing EHR-Associated Workloads
Most doctors have an interesting relationship with EHRs. On one hand, they’re a precious source of patient info they can access at the click of a button. On the other hand, the time required to compile these electronics is often seen as an unneeded and burdensome clerical task.
NLP and AI can be used to speed up the record-keeping process so that clinicians can spend much of their time connecting with patients rather than swamped by bureaucracy and administrative tasks.
AI Use Case #14: Streamlining Drugs R&D Process
New drugs and medications require immense investments during the research and development (R&D) phase. Research teams may require years to isolate promising molecules which may later fail when they’re tested on live human subjects. The cost of bringing new drugs to market can be very high.
On top of that, the regulatory challenges of bringing new treatments to market are often extremely hard to be overcome.
AI and ML applications can streamline this process by selecting only the specimens that deserve human attention, as well as simulate trillions of potential drug interactions with their biological targets in a matter of minutes.
Patient data can be also cross-referenced with ease, allowing for safer medications with fewer side effects. Researchers can thus focus their attention only on those molecules that are worth their time.
Newer, safer, and more innovative drugs can reach the market much more quickly, with an estimated save of as much as $2.6 billion USD during the time required to develop a new drug.
AI Use Case #15: Improving Third World Quality of Care
Today, AI that's used to recognize diseases and conditions by looking at patients’ history and medical imaging records is fed and trained with data coming from databases across the entire planet. In this sense, ML is creating global expertise since it’s collecting info obtained by highly trained doctors that already provided a diagnosis to the patient analyzed by the AI.
This will impact rural and less developed countries where access to proper facilities and skilled specialists reduces the overall quality of care provided. By using AI, we can provide patients with the same quality of care and treatment used in the top health facilities of the world since all decisions are based on the opinions of the best doctors and specialists available.
AI Use Case #16: Preventing Heart Diseases
Potentially lethal heart diseases such as aortic stenosis (AS) can be detected only by highly skilled physicians. Minuscule variations in heart beat, such as a heart murmur, must be heard and identified with a stethoscope, and it’s not infrequent that these sounds are so faint and subtle that end up being misdiagnosed.
Smart algorithms such as the one used by Eko's AI are able to detect the sounds associated with AS with a 97.2% precision rate. Other AI-based technologies used to detect heart issues are the wearable Zio Patch developed by iRhythm Technologies that monitor heart rate activity for up to two weeks, and Apple’s ECG app which can detect early signs of atrial fibrillation.
AI Use Case #17: Preventing Pressure Injuries
Pressure injuries (often called bedsores or decubitus ulcers) are a serious issue for patients with limited mobility or who must spend most of their time in a chair or bed. Some of them never heal completely even after treatment, and, in the U.S., they cost up to $11.6 billion every year.
ML-based models have been developed to correctly predict and assess risk for pressure injuries in critical care patients by examining large amounts of clinical data readily available in the patient records. These models can identify those patients who will benefit most from specialty interventions that are too expensive to be used for every patient.
AI Use Case #18: AI-Powered Paediatric Services
Enable My Child is a paediatric therapy platform that provides online, in-person, and hybrid speech, occupational, psychological, and physical therapy to children across the U.S.. Their AI-based platform called the EMC Brain uses machine learning algorithms to track specific intervention strategies that have been successful with students of various ages, abilities, and conditions and uses that data to help formulate effective treatment plans.
Therapists can access the EMC Brain for evidence-based data to assist with therapy planning to make therapy more effective and yield better results for students. The EMC Brain serves as a vehicle for delivering therapy and utilizes artificial intelligence to collect data on treatment plans and make recommendations as the system grows smarter.
AI Use Case #19: Large-Scale AI-Based Insights
Data-driven insights are a fundamental resources for healthcare organizations and governments to establish new health policies which may benefit the population at large. KenSci’s risk prediction platform has been developed to aggregate data coming from admission, discharge, and transfer systems, claims and EHRs and provide insights to get ahead of chronic diseases such as chronic obstructive pulmonary disease (COPD), Chronic Heart Failure and diabetes.
This platform can be used to find new ways to reduce the risk of T2 diabetes through the use of wearables and lifestyle suggestions, identify runaway drug costs and drug variation. Together with Microsoft, it is assisting national governments to identify which COPD patients are getting worse and who needs a higher dose of medication.
It has also been used to identify runaway drug costs and drug variation, revealing that the right data can tell you that administering oral acetaminophen is more effective and less costly than via an IV, for example.
AI Use Case #20: Reducing the Burden of Uncompensated Care
Up to 75% of consumers are not made aware of their financial obligations until after a procedure or service was completed at the hospital, leading to unaccountable bills for medical procedures and more than $38 billion USD in bad debt carried by health systems.
AI is being used for optimizing and automating revenue cycle processes in healthcare, notably self-pay. Some of the accounts are more likely to pay given the opportunity, while others do not qualify for full financial assistance.
Together with robotic process automation, AI’s advanced micro-segmentation capabilities can now drill down into each patient’s ability to manage and pay their bills. An intelligent scoring will then help identifying the categories that are more at risk to incur into debts they can never repay.
From analyzing huge amounts of data in a matter of seconds to provide optimized treatments to patients, to spotting even the most minuscule detail in medical imaging, AI is changing the way humans experience medicine in a lot of ways.
Smart machines can’t help us become smarter (yet), but they’re assisting us with their intelligence to hopefully help us live longer, better and ultimately healthier lives.
More from AltaML
- What are some of the foundational ways that career pros stand out in machine learning?
- What are some of the dangers of using machine learning impulsively without a business plan?
- Why is so much of machine learning behind the scenes - out of sight of the common user?
- What is 'precision and recall' in machine learning?
- Why are GPUs important for deep learning?
- What’s a simple way to describe bias and variance in machine learning?
- What are some of the main benefits of ensemble learning?
- Will machine learning make doctors obsolete?
- What's the difference between machine learning and data mining?
- How does a weighted or probablistic approach help AI to move beyond a purely rules-based or deterministic approach?
- Why is data visualization useful for machine learning algorithms?
- Why are some companies contemplating adding 'human feedback controls' to modern AI systems?
- What does 'connectionism' mean for business AI?