My technology-focused colleagues typically give me a big smile when they hear that I‘m a registered nurse with experience in acute care and home health settings. The smiles are not so much about finding a unicorn in their midst; these co-workers are remembering a nurse or doctor or aid who once helped them or a loved one through a healthcare challenge, and that glow is being cast onto me. Like them, I’m grateful for the care I’ve been given over the years — perhaps even more so because I realize the cost to the caregivers.
Healthcare providers face tremendous challenges, from an aging population and rising numbers of chronically ill and critical-care patients to a shortage of clinical specialists. As the cost to provide care to these patients rises, providers are feeling increased pressure to do more with less. It seems there are never enough feet on the floor to deliver optimum patient care.
Challenges like these point to the need to streamline clinical workflows, improve diagnostic capabilities, deliver personalized medicine and reduce the time that patients spend in hospital beds. To achieve these outcomes, many providers are realizing the extraordinary benefits of using artificial intelligence (AI) across a wide range of use cases.
How AI helps
For hospitals, clinics and other providers, AI applications address today’s healthcare challenges with data-driven insights. Derived from stores and streams of existing data, these insights can help providers significantly boost the quality of patient care while combatting rising costs.
Progressive healthcare providers are using AI to integrate and analyze large volumes of data that lead to more effective disease prevention and treatment protocols. AI helps them screen medical images more quickly and with greater accuracy, enabling personalized healthcare. AI also helps streamline the delivery of healthcare services to reduce wait times and costs.
AI use cases for healthcare
To gain a broader and deeper perspective on the ways in which healthcare organizations are using AI, I turned to my colleague Dr. Curtis Breville, DM/IST, a principal analytics specialist for Dell Technologies who specializes in AI solutions for healthcare. When “Dr. Curtis”–as he is known to colleagues and customers–was a child, his scientist father caught an undiagnosed virus that led to complications permanently affecting his quality of life. More than 45 years later, this event still compels Dr. Curtis in his unwavering passion for using technology to improve the diagnosis and treatment of illness.
Here are some of the AI use cases that Dr. Curtis is involved in now.
AI imaging systems
Using deep learning techniques, AI imaging systems can identify abnormalities in radiological images and make predictions about patient issues. In this training process, the AI model, or algorithm, might look at hundreds of thousands of images to determine what is a normal state and what is an abnormal state.
With these capabilities, AI systems can accelerate the diagnostic process and improve its accuracy. The insights from AI can help physicians start down the right therapeutic path in less time. At the same time, AI can help ease the burden on radiologists, who are often in short supply in hospitals, while enabling healthcare providers to see more patients.
Dr. Curtis notes that this “augmented intelligence” doesn’t replace doctors and other clinicians in the diagnostic process. Instead, it helps them to make better decisions in less time.
Quality patient care
AI systems also help healthcare providers improve the quality of patient care on many fronts. One of those is through personalized medicine that tailors therapies based on the patient’s genome, lab samples, healthcare history and other factors. In this way, AI-driven systems help providers avoid the pitfalls of “one-size-fits-all medicine” — because different patients with the same disease don’t necessarily respond to treatments in the same way.
Research indicates that this more precise approach to medicine — “predicting what treatment protocols are likely to succeed on a patient based on various patient attributes and the treatment context” — is the most common application of machine learning in healthcare.
AI-driven systems can also facilitate remote monitoring and diagnostics. For example, patients can wear connected intelligent devices that track their vital signs, blood oxygen saturation, blood sugar levels and more, and issue alerts to their doctors when the numbers fall outside of acceptable ranges.
Infection prevention and control
Avoiding outbreaks in hospitals is a matter of life and death. AI-driven systems can assist by tracking compliance with hygiene protocols, such as handwashing and sterilization, and issuing reminders and alerts in the event of non-compliance. AI can also be used to detect infection outbreaks and predict at-risk patients for proactive infection prevention and control measures.
Translation and communication
In recent years, we have seen dramatic advances in natural language processing (NLP), a form of AI that is used for voice-to-text, text-to-voice and language-to-language translation. In a clinical setting, NLP systems can be used to break down communication barriers and extend quality healthcare to more people.
NLP techniques also can drive efficiencies in day-to-day clinical operations. For example, hospitals can use voice-to-text transcription to ease the burdens that come with entering data in the hospital’s electronic medical records (EMR) system and submitting orders for prescriptions and tests.
For healthcare providers working to contain the rising cost of care, staff efficiency is paramount. AI-driven processes can help providers make more effective use of their limited staff resources, in part by streamlining workflows and scheduling processes and procedures to ensure that the right people are in the right place at the right time.
While optimized processes can help cut the costs of delivering care, they might also help reduce the risk of costly malpractice suits by identifying problems with processes that could result in substandard patient care.
How do healthcare providers take advantage of AI to further their goals? Dr. Curtis offers these considerations for capitalizing on AI:
Establish a master data management system.
In many hospital and healthcare systems, each medical discipline has its own IT systems housed in dedicated infrastructure. Data isn’t easily shared among these disparate systems. This isn’t how the body works, Dr. Curtis points out.
“Doctors today know that the 11 different organ systems of our bodies affect one another,” he says. “They are also continually finding new correlations, and even some causations, between activity in one part of the body and a reaction in another. To use data analytics and AI to facilitate these cross-discipline findings, doctors need a data management system that brings together datasets from disparate systems.”
Prepare to train and scale
Dr. Curtis notes that these master data management (MDM) systems must be able to scale processing power for faster data analysis, as well to accommodate greater amounts of data. In addition, they require protection from ransomware attacks and need plenty of network bandwidth to replicate and ingest data quickly. They also need to be flexible, adaptable and scalable to accommodate new requirements.
Training the AI system requires a lot of data, Dr. Curtis notes. He offers the example of training a diagnostic-imaging system to recognize the differences between scans of a healthy lung versus scans of the unhealthy lungs of a long-time smoker.
“The process is actually quite complex because the AI model needs to identify and accurately diagnose someone who may have a history of emphysema, asthma, bronchitis and pneumonia with perhaps an additional illness that may require a specific medication,” Dr. Curtis says. “The AI model must be trained on tens of thousands to millions of scanned images so that when a new image is scanned, a quick and accurate diagnosis can be made.”
Plan for retraining
While AI can help screen and flag anomalies (or “outliers”) for attention, doctors and trained clinicians still need to verify the results. If they find an error has been made by the AI model, that information needs to be fed back during additional model training to continue improving accuracy.
“Training never ends,” Dr Curtis says. “New health risks pop up, and those affected are desperately seeking a cure. Step one is to accurately diagnose what it is and what it is not.”
Continuously-trained AI systems can identify non-conforming anomalies, and begin to address those as distinctly different from other diagnoses that the system is familiar with. Doing this allows faster, targeted and personalized medicine to be administered.
The power of AI is complemented by edge computing–any point where the digital and physical world intersect and data is securely generated, collected, processed and used to create new value without sacrificing quality.
“In the healthcare world, the edge is any place outside of the hospital or clinic where insights can be gleaned from data — for example, mobile remote medical systems that can run tests and diagnose patients where there may not be any trained medical staff,” Dr. Curtis says. “Like AI models, edge systems must continue to be trained to improve the accuracy of diagnosis and treatment. That means we need fast connectivity with networks to upload and download data. With 5G networks becoming more available, the future looks bright for bringing advanced technology to every corner of the globe.”