Artificial Intelligence in Healthcare: Machine Learning Use Cases
This article aims to cover all possible use cases of machine learning and artificial intelligence in healthcare that have ever been proposed, predicted, or already implemented.
Since most of these application areas are closely interconnected and it’s hard to draw a sharp line between them, we’ve come up with the following structure to help you navigate through them easily.
1. AI and Machine Learning in Healthcare for Accurate Disease Identification
According to Gurpreet Singh, PwC’s US Health Services Leader, diagnostics is one of the three major “zones of investment” for artificial intelligence in healthcare, with two others being digitization and engagement. AI promises more accurate diagnoses, a benefit that can save and improve millions of lives.
Image analysis powered by AI
How it works
An operator uploads a medical image of an organ to an AI program. The software then checks this image against millions of others featuring the same organ struck by a disease, one by one. Among these images, it looks for patterns that are common for the disease. If the image shows enough similar patterns, it is diagnosed positively.
These AI-powered tools have recently achieved even better accuracy than humans, thanks to using extremely huge amounts of computational power, which was unimaginable just a decade ago.
Use cases
AI can read any type of medical images, and many companies are working to take advantage of it now. You can find notions of radiology scan analysis, retinal imaging, mammogram image analysis, and more.
3D scan analysis
In 2018, a research team led by MIT presented a machine-learning algorithm capable of comparing and analyzing 3D brain scans. It does the job 1,000 times faster than current techniques.
Classification of echocardiograms
Another impressive healthcare AI use case is a medical imaging system trained by an assistant professor and practicing cardiologist at US San Francisco, Rima Arnaout, and her team.
The AI and expert cardiologists were challenged to classify ECGs based on the type of view shown. The software performed the task with 92% accuracy when humans only achieved 79%.
Skin diagnostics
Apart from interpreting images generated by specific medical devices, AI could empower every user of a camera-enabled smartphone. You take a photo of a mole or some other blemish using a specific app.
The AI trained on hundreds of thousands of melanoma pictures analyzes it and gets back to you to either recommend seeing a doctor or secure your peace of mind.
ML-Driven Genome Analysis
Scientists use a technique known as genome sequencing to read genes like a book. This process involves computational power and human expertise, making it slow, expensive and inaccurate.
The introduction of artificial intelligence in healthcare can change that.
How it works
This is where the ability of machine learning algorithms to work with large sets of “if…then…’s” and base its “opinion” on real-world data kicks in.
To create sequences, a researcher feeds a DNA to a special program. The AI software, in its turn, runs these sequences through so-called decision trees and checks them against others in its database.
If similar patterns indicating abnormality are found, the AI program raises a red flag.
Use cases of AI – Cancer diagnostics
Cancer is so extremely difficult to cure due to the dynamic nature of tumor DNA. In 2018, researchers developed Cerebro, a machine learning method, that leverages sequences to the fullest.
A complex set of algorithms evaluates huge decision trees to spot possible DNA mutations based on the real-world and pre-modeled data. Compared to other known methods, the Cerebro’s results proved the most accurate.
Improved Speech Recognition with AI
How it works
Speech recognition is another prominent healthcare AI use case. It works just like image reading, but instead of pictures, an AI program analyzes spoken words.
A patient speaks into a mic-enabled device. This sound is encoded into digits to be further compared with other digits matching particular words and phrases.
The AI program checks the input speech against those taken from mental patients to find similar patterns and diagnose accordingly.
Use cases – Detecting psychotic disorders in development
Scientists at Harvard University and Emory University recently came up with a machine-learning method to detect early signals of developing psychotic disorders.
According to their findings, if you are using too many words associated with sound and your speech is semantically poor, you should start worrying.
The AI-based technique developed by the researchers is very sensitive to these linguistic abnormalities, which human ear fails to identify.
AI-Enabled Big Data Analytics for Better Insights
How it works
Over the years of digital, the healthcare industry has accumulated vast amounts of medical data. Structuring and analyzing these insights is one of the tasks that AI can perform more efficiently than an army of scientists.
By checking millions of diagnoses against each other, machine learning algorithms could tease out patterns and regularities hidden from doctors’ eyes.
Further, developers could feed these regularities to an AI program. A physician would run patient health records, blood and urine tests, tissue-based biomarkers, genomic data and medical images through the program to find matches with illness data.
This is an opportunity to not only diagnose the disease in a particular case but also identify its stage and predict further development.
Use cases
Researches are yet to come up with particular products. However, this technique is already being actively discussed in the scientific community.
Dr. Katarzyna Macura, Professor of Radiology, Urology, and Oncology with Johns Hopkins University in the US suggests that this strategy can help assess tumors and target biopsies with higher accuracy.
2. Using Artificial Intelligence in Medicine to Improve Pharma
Machine learning algorithms can not only effectively detect a negative abnormality but also find a pharmaceutical or high-tech solution to it.
Discovering New Drugs
How it works
The ability of AI to efficiently process loads of multi-format data, pair data segments and narrow down patterns is extremely valuable. For pharma, for example, it is an opportunity to shorten the drug development cycle.
There are plenty of ways to use this potential of AI. Let’s say, an AI program could scan thousands of medical records of patients suffering from the same disease.
The algorithms with the program would compare the efficiency of the same treatment on various people and find out patterns and regularities.
These findings could suggest why this particular medicine works better for some and has little to no effect on others. Based on that, researchers could improve existing drugs and create new ones.
Use cases – Improving flu vaccines
In 2018, one of the major drug manufacturers, Sanofi, launched an AI-based research into heavy loads of data to discover what makes particular flu vaccines unequally effective.
In fact, AI can be a powerful ally for pharma at every stage of the drug lifecycle: target discovery, drug discovery, development and post-approval.
Developing Intelligent Medical Devices with the Help of AI
How it works
Medical appliances is another healthcare AI use case where machine learning can make a tangible difference. Typically, machine learning would be used to empower an appliance with artificial intelligence.
Use cases – More accurate and faster 3D printers
Aether and Procter & Gamble recently joined forces to fuse robotics, machine learning and 3D printing.
The two corporations aim to create a new intelligent bioprinter delivering highly realistic substitutes for human limbs and organs much quicker and easier than available now. And this is just one example of ML for medical devices!
Easier drug manufacturing
In addition to discovering new drugs, AI can give non-pharmaceuticals a leg up in medicine manufacturing.
DARPA’s Make-It program is working on an AI-powered device that can automatically produce particular molecules for drugs. It is thought to enable production of common drugs on demand in remote locations with complicated logistics and shortage of resources.
3. Powering Clinical Research with AI
Scientists can benefit from artificial intelligence in healthcare by streamlining research and improving its accuracy. Here is how.
Accelerating Trials with NLP and Deep Learning
How it works
Machine learning in healthcare is changing how patients are enrolled in clinical trials. The two AI techniques, natural language processing (NLP) and deep learning, can help automate and accelerate the process.
A candidate opens an AI program. They answer a set of questions allowing to determine whether they are a match for a particular trial. It can also find matching trials based on the candidate’s answers.
Use cases – Immunosuppressant dosing calculation
The trial process has been a healthcare AI use case for years now. In 2016, one such trial resulted in the discovery of a mathematical formula calculating the immunosuppressant dosage to be prescribed to patients receiving organ transplants.
Artificial Intelligence for Improved Modeling of Organs
How it works
An AI program can “learn” about the structure and functions of a particular organ. Researchers “teach” it by feeding the software huge amounts of medical images of this organ in various states.
Based on that, the program creates a model of this organ that reacts to every action in particular ways. Researchers can then experiment with this model to learn more about the organ.
Use cases – Brain modeling
Modeling human brain algorithms is one of neural network applications in healthcare. The most famous example of this is Google’s DeepMind.
It is translating intelligence into an algorithmic construct allowing to better understand how the human brain works. This application of deep learning is designed to identify new patterns leading to the development of mental illnesses.
Genetic Research Based on Big Data
How it works
Machine learning in healthcare empowers genetic research thanks to the technology’s ability to work over big data. AI software reads, pairs and compares to find genetic patterns associated with particular diseases.
Use cases – Identifying antibiotic resistance genes
Researchers at the University of California San Diego applied machine learning to analyze tuberculosis-causing bacteria. The algorithm identified 33 known and 24 new antibiotic resistance genes!
The scientists say that this success can be replicated with other infection-causing pathogens.
4. AI Paving the Way for Precision Medicine
How it works
Speaking of the ultimate goal of artificial intelligence in healthcare, many experts believe it to be the rise of precision medicine.
This concept can be described as a comprehensive patient profile unifying all of their health data to allow for highly personalized treatment.
An AI platform can collect electronic health records (EHR) of a person and present them in a clearly structured manner, including predictions and comparative data from external sources.
This will enable clinicians to create highly effective treatment plans and make data-driven decisions.
5. Healthcare AI to Improve Disease Prevention
The level of precision in big data analysis made possible by AI is unprecedented. That said, predictive analytics promises a breakthrough in preventive medicine.
Predicting Diseases Early through AI Symptom Analysis
How it works
As we have already mentioned, AI can effectively check patient health data against external sources to more accurately interpret symptoms. But even if an individual has never reported any disease signs, they might be at risk — which is yet another healthcare AI use case.
Software powered with machine learning algorithms could simply check medical records of a patient against disease records on a regular basis. By spotting any suspicious similarity, it could signalize a physician for further inquiry.
Development of preventive AI solutions in healthcare industry has been underway for years.
Heart attack and stroke risk prevention
In 2017, researchers at the University of Nottingham came up with an intelligent system predicting heart attacks and strokes. It scans routine medical records of patients and identifies the ones at risk within the next 10 years.
This technique beat the conventional methods on accuracy by 355 (7.6%) cases.
Digital patient copies
Just like by modeling organs, an AI program could create a digital copy of a patient. A physician could try various lifestyles and behavior models on this copy to see what and how could harm the patient in the long run.
AI-Powered Prevention of Harmful Behavior
How it works
There are several not-so-obvious threats to human health that clinicians might have no idea of. These include suicide and partner violence.
Based on patient data collected from multiple sources, AI can detect particular patterns associated with suicide, such as search queries, social media posts and clinical records.
The ability of AI to more accurately classify injuries could help detect violence-driven wounds.
Use cases – Analyzing clinical records
In 2017, researchers at Vanderbilt University Medical Center created a machine learning algorithm that analyzes hospital admission data to spot patients who could take their own life soon.
Seeing into social content
In 2017, Facebook rolled out a pattern recognition feature to analyze user content for the risk of self-harm. The Community Relations team reviews the posts flagged by AI and contacts local first responders to get help.
Detecting suspicious injuries
Artificial intelligence in healthcare has the potential to prevent intimate partner violence and domestic abuse from happening again.
Victims tend to conceal the real source of an injury when at a physician’s office. Medical images, in their turn, fail to convey the full story to radiologists.
But AI can identify certain patterns that raise a flag, according to Bharti Khurana, MD, an emergency radiologist at Brigham and Women’s Hospital in Boston.
Palliative care prediction with Machine Learning
Unfortunately, there are still many diseases that gradually drain life from people. The best thing to do in this case is to soothe the pain and relieve the stress. However, an estimated 1 to 1.8 million patients do not get palliative care at the right time.
How it works
The AI’s ability to analyze patient data and check it against known cases of disease development could help predict the palliative stage early enough to provide adequate care.
Use cases
In 2017, researchers trained a machine learning algorithm to analyze patient health data of up to 12 months before they passed away.
The team rolled it out to identify the still-living ones at high risk. Every one of the individuals who the algorithm flagged was later proven eligible for palliative care.
Predicting Epidemic Outbreaks Using Artificial Intelligence
How it works
By analyzing particular non-medical data, machine learning algorithms can predict epidemic outbreaks.
This can be data from satellites (animal migrations), social media posts (complaints about symptoms), queries on search engines (massive search for the same symptoms), and more.
If an AI program sees that a set of current conditions matches what was previously identified as leading to an epidemic outbreaks the operator passes this information to the relevant department for preventive measures.
Use cases
In 2015, researchers at Applied Artificial Intelligence Group published a healthcare AI use case study on predicting malaria outbreaks with machine learning.
They proposed using its algorithms to analyze outdoor temperature, average monthly rainfall, the total number of positive cases and other information to identify the risk of an outbreak in a particular region.
6. Artificial Intelligence for Genuinely Professional Care
Artificial intelligence in healthcare might replace nurses and caregivers for a plethora of tasks. AI-powered systems can ask, listen, answer and watch, providing the necessary guidance and connecting patients with medical staff.
ML Algorithms for Improved Medication Intake Management
How it works
The effectiveness of treatment often depends on a patient’s adherence to medication, while incorrect dosing can do more harm than good. Here, we have another healthcare AI use case.
The ability of machine learning algorithms to “read” and separate the wheat from the tares could help record an intake of drug and analyze its effects.
Use cases – Medication adherence monitoring
The patient launches a special app on your smartphone, looks into the camera and takes a pill. Using facial and image recognition, the AI application identifies who took the medicine, what medicine and how much of it.
The application enters this information onto the physician’s dashboard so they can accurately monitor the treatment progress.
Psychotherapy and Counseling with Chatbots
How it works
Artificial intelligence in healthcare is about smart chatbots. These chatbots use device microphone and speech recognition to analyze what patients say. They are specifically “taught” by expert psychiatrists how to react to particular claims.
Use cases
Intelligent bots could help patients talk through their worries. They launch an app and describe what bothers them. The app then responds by asking them questions and, step by step, helps them better understand their problem.
AI Tools for Caregiver-Patient Communication
Can machine learning in healthcare remove caregivers completely? It is possible, but probably not in the next ten years. Meanwhile, machine learning algorithms might improve how patients and caregivers communicate with each other.
How it works
An AI program could analyze the competencies of a particular caregiver and match them with patients who are in need of such competencies most.
According to Frost & Sullivan, care coordination IT vendors will utilize AI to automate the process of matching patients with caregivers.
Another communication problem artificial intelligence in healthcare can solve is that for stroke patients. Brain-computer interfaces allow them to “speak” with nurses and clinicians soon after the stroke rather than after rehabilitative therapy.
Use cases – Speech synthesizers
Remember the speech synthesizer Stephen Hawking used? It was AI-powered and it should give you an impression of the direction things are going to take.
Home Care and Rehabilitation with Robots and Devices
How it works
When powered with AI, physical robots and chatbots could take over the routine tasks of delivering physical care to immobile patients. It’s practically about combining robots’ strength with AI’s ability to process commands and respond in a meaningful way.
Use cases – Strong robots made intelligent
Take RoBear by RIKEN, a “strong robot with the gentle touch” as they call it. Paired with the conversational and learning abilities of AI, it could make a big difference. A disabled patient would only need to call the robot and explain their needs.
The robot would bring them their pills, cook dinner and even help them move to the bathroom.
Intelligent exoskeletal devices
A probably more elegant healthcare AI use case in rehab is exoskeletal devices using the technology to facilitate faster recovery of motion functions.
An AI-powered device like that could learn motions of a patient and predict the next similar motion to reduce the physical effort.
Chatbots monitoring health and providing guidance
Conversational platforms utilizing artificial intelligence in healthcare enable patients to prepare for surgery and monitor the rehabilitation process, providing the necessary guidance and even detecting their moods.
ML-Driven Intensive Care
How it works
Intensive care units (ICUs) are extremely tech-laden. The machines take upon the vital body functions to keep a patient alive and monitor their state in real time. As a result, data is coming in in huge amounts — not captured, not properly analyzed. A job for AI, isn’t it?
When introduced in an ICU, machine learning algorithms could use “the knowledge” they were taught to accurately read information from multiple devices. Based on these insights, the AI program could carefully calibrate treatment.
AI Programs for Risk Group Supervision
How it works
Patients at risk of stroke, heart attack and seizures require expert supervision around the clock which is often impossible. However, an AI program that reads data from sensors, correctly interprets it and makes informed decisions can do that.
Use cases
In 2016, researchers described the so-called Ambient Intelligence concept. Imagine that your wireless fitness band continuously monitors your heartbeat through electrocardiograms and sends this information to an AI app on your smartphone.
The machine learning algorithms with the app compare this data to the stroke records they have been fed. If they recognize a pattern at some point, they immediately send an SOS to the nearest ambulance. Yeah, like a life-saver but even better.
7. AI Taking over the Healthcare Bureaucracy
Improving administrative workflow and patient-provider relations has been one of the major “zones of investment” for artificial intelligence in healthcare (PwC).
There are several ways how machine learning algorithms can change the game for medical organizations and their patients.
AI-Based Care Navigation
How it works
AI promises to enhance and accelerate care navigation for patients. An AI-powered system with a clinic could ask patients questions based on millions of real-life cases.
It would “understand” their answers to provide guidance for further actions, such as seeing a general practitioner or calling the emergency department.
Health Record Management Streamlined with AI
How it works
Electronic health records have been around for a while. However, some medical data is overlooked or stored somewhere else: audio recordings from multidisciplinary team meetings, data from wearables and implantables, handwritten notes and others.
Artificial intelligence in healthcare can extract data from multiple sources and present it in a unified format thanks to its ability to “hear”, “see” and “read.”
In addition, AI programs could assist physicians in seeing patients. Doctors don’t need to record what they are being told about symptoms, because smart speakers (yes, like Amazon’s Alexa) can capture the story.
A camera-enabled device powered with computer vision and image recognition can analyze the external signs of a malady and provide suggestions to the physician for consider.
AI Chatbots for Time and Resource Management
How it works
AI chatbots will take upon many routine administrative tasks, such as:
- assisting users with appointments
- answering questions about work hours
- medical coding and billing
- triage
Dr. Katherine Andriole, Associate Professor of Radiology at Harvard Medical School, sees particular value in predictive analytics. She proposes utilizing this feature to predict which patients are most likely to skip a visit and plan resources accordingly.
8. Emergency Medicine as a Prospective Healthcare AI Use Case
The swiftness of automated systems and their ability to process multiple requests at a time are invaluable when it comes to emergency medicine. With the introduction of artificial intelligence in healthcare, automation will save many more lives.
Processing Calls with Machine Learning
How it works
The ability of AI to recognize speech could significantly accelerate processing of emergency calls. While a computer listens to an emergency call taken by a human dispatcher, the AI can analyze the words, manner of speaking and background noises to identify heart failures.
Experts believe that machine learning in healthcare will soon assist doctors in urgent decision making. Yan Guang, Deputy Head of Anhui Provincial Hospital, suggests leveraging AI to process big data.
He explains: “Once you call the emergency centre for first aid, your health information, based on all of your hospital records, will be provided to the first-aid personnel in the ambulance and doctors at the hospital.”
Based on this information, intelligent algorithms can also provide advice on particular first-aid measures.
Further, artificial intelligence in healthcare will automate the entire triage process, making it more accurate.
AI for Automated Image Reading
How it works
In case of emergency, automated image reading makes a big difference. UC Irvine Medical Center uses an AI-powered application to identify a stroke from a CT scan in just 20 seconds.
9. Data-Driven Guidance for Surgeons
How it works
Artificial intelligence has the potential to disrupt another hot technology in healthcare — surgical robotics. Image recognition algorithms can check what the robot sees against images from previous similar operations and pre-op records.
These insights enable the AI program to predict the outcome of the operation and advise the surgeon at the workstation.
Use cases
An AI program can read CT scans and create a 3D copy of the organ under surgery. Surgeons model a successful operation on this digital copy. When they take it to the operating room, the robots guides their hand based on the model they created.
10. Training AI with Machine Learning to Train Wannabe Docs
How it works
Medical modeling and simulation are the key AI applications in medical education and training. It’s basically about artificial intelligence learning from myriads of real-world situations to then create a very precise model or simulation.
Use cases – Realistic medical mannequins
The US Defense Health Agency’s Medical Modernization and Simulation Division even married AI with augmented reality (AR).
Students wear an AR helmet to observe mannequins react just like humans: moving, changing in face, crying and shouting — as learned by AI from real humans.
11. AI-Powered Machines to Manage Human Resources
There is yet another area, though not directly associated with medical actions, that can be improved with the help of healthcare AI — human resources. Let’s dive deep into the benefits of AI solutions for medical HR.
How it works in particular cases
Recruitment
Before a human HR manager steps in, an AI-powered chatbot can do the talking with candidates to exclude total mismatches. Further, it can conduct preliminary interviews with prospective employees.
Onboarding
Chatbots using NLP can free new clinicians and other employees from answering newcomer questions regarding the organization’s structure, rules, resources, etc.
Compensation management
Chatbots can even handle benefits enrollment and salary payment, saving healthcare providers a good deal of time and tax money.
Employee satisfaction management
NLP enables chatbots to talk to employees about their general satisfaction level and the problems they are facing at work. The information extracted from these dialogues allows preventing crucial bottlenecks in human resources.
12. How AI in Healthcare will Save the Day for Insurers
Of all healthcare AI use cases, this one may seem least obvious. However, health insurers could save up to $7 billion over 18 months by introducing AI (Accenture).
Use cases
There are three key applications in this area:
Claim processing
McKinsey expects chatbots to cut claim processing headcount by 70-90% in 2030. This alone promises to save insurers a nice chunk of moolah.
Fraud detection
Machine learning algorithms will easily pinpoint anomalies in claims when specifically trained. By feeding the data from provenly fraudulent and non-fraudulent claims to AI, you can teach it to identify medical upcoding.
Another fraud AI will be able to catch is billing for services that have never been rendered. Machine learning algorithms can analyze the electronic health records of a target healthcare provider to find out whether the patient actually received these services or not.
Personalized policies
Predictive analytics will enable insurers to calculate the best health plan in each particular case. By analyzing customer data from accessible sources, machine learning algorithms can forecast costly life events.
For example, Prognos leverages AI to pore over billions of medical records to predict emergency room visits, body part replacements and depression development.
One of the major insurers, John Hancock, made a major leap into behavior-based health insurance in 2018. The company stopped underwriting traditional life insurance and now only sells interactive policies that extract data from wearables and smartphones.
As a result, customers who exercise their way to a healthy lifestyle pay reduced premiums.
13. Better Medical Supplies Thanks to AI
Software has been improving inventory management for a while now. The next stage will be empowering it with predictive analytics and machine learning algorithms.
Predictive Analytics and Medical Inventory Management
How it works
The predictive power of machine learning in healthcare can help hospitals prevent shortage of supplies and reduce waste by optimizing buying lists.
On the other hand, predictive analytics is an opportunity for suppliers and manufacturers to foresee demand and plan business accordingly.
Supplier relations
Artificial intelligence in healthcare enables providers to identify the most competitive suppliers for each service. Machine learning algorithms will analyze their competitors’ spend data and pinpoint benchmarks.
AI-powered analytics will also allow sourcing teams to prioritize renewals of supplier contracts through a customizable set of parameters.
14. Driving Medical Marketing with Artificial Intelligence
If machine learning algorithms enable unimagined levels of precision when diagnosing diseases, imagine how they can bolster medical marketers’ capabilities.
How it works
B2B
The use of AI in the healthcare sector promises two benefits for manufacturers and distributors:
- Analyzing clinical and quality metrics to find matching healthcare providers within a location.
- Using information about potential clients to create accurately calibrated, effective campaigns.
B2C
In this healthcare AI use case, machine learning algorithms can empower both providers and medical retailers to improve their targeting.
The population data is constantly accumulated from multiple sources, such as search engines queries, social media posts, wearable statistics, hospital records, caregiver routes and others.
AI will analyze it and enable marketers to target potential customers with the right message at the right time.
Predictive Analytics for Marketing Budget Optimization
How it works
Predictive analytics is expected to optimize medical marketers’ budget for higher efficiency. Intelligent algorithms will identify the most effective marketing channels for each particular service or product and build accurate ROI forecasts.
Conclusion
This must be it: your comprehensive guide to all applications of artificial intelligence in healthcare.
If you look at intelligent algorithms in a broader context, you might see thousands of ways how AI can improve healthcare, from making professional services more accessible in the developing world to solving the problems of aging population.
What do you think, though? Are there any healthcare AI use cases that might be missing on this list? Let us know in the comments!