Adoption of AI in Healthcare, insights from HIMSS19

Leave a comment
Artificial Intelligence / Healthcare

 Last month, we’ve attended HIMSS19. If you are not familiar with the event, Healthcare information and Management Systems Society (HIMSS) is an annual healthcare information conference, which brings together more than 45,000 healthcare and information technology professionals from all over the world. 

Artificial Intelligence was a big topic at HIMSS this year. On Monday, we attended “Machine Learning and AI for Healthcare” pre-conference event where health systems, technology companies and clinicians discussed their efforts to utilize AI and Machine Learning to improve healthcare system, boost medical outcomes and reduce costs.

During the pre-conference event, we’ve heard from experts who represented an array of leading healthcare organizations – Duke, CHOC Children’s, Hackensack Meridian, Optium, WMC Health and others. Many presenters agreed that AI is finally being embraced by the healthcare industry, but noted that the application of AI is still in the early stages. According to survey results shared by Optum, healthcare systems plan to spend around $34.2 million dollars on AI over the next five years.

As data scientists, we were interested in real world applications of AI in healthcare. Current applications mostly focus on automating repetitive tasks and many are designed for specific use cases. Most solutions that we saw at HIMSS19 focused on the following areas:

  • Better clinical outcomes
  • Operational Improved analytical capabilities such as (patient identification and personalized treatment and management)
  • Lower cost of care
  • Expanded patient reach (serving broader population through digital tools)
  • Better access to care
  • Patient satisfaction
  • Increased cybersecurity

“The first and most impactful applications of AI will most likely involve workflow improvement and subsequent better patient care. For example, AI can pre-screen medical images and identify high risk cases and bump them up to the top of the queue, or, rule out any disease and put them on the bottom.” – Herman Oosterwijk, Trainer and consultant of DICOM, HL7, PACS, FHIR. President of OTech, training and consulting company in Healthcare IT space.  

Unleashing Healthcare Data with AI and Machine Learning

Healthcare creates an enormous amount of data from patient care, academic research, and commercial activities like drug testing and trials. According to IDC, healthcare data will grow to 2,314 Exabytes by 2020 with an annual growth rate of 45%. Now, the question is, how do we put this data to work?

“We have spent the last seven or eight years building a foundation of data and we are living on a mountain of it, we digitized the healthcare experience for everybody in the country and now we are able to use that data into Machine Learning and AI tools that will lower the cognitive burden not only for the caregivers and the clinical team, but for patients as they are making choices in the healthcare system.” – Dr. Karen DeSalvo, Former National Coordinator for Health IT, U.S. Federal Government, now a professor at the University of Texas at Austin Dell Medical School

Most of the healthcare data we collect is unstructured, requiring extensive work to structure, code, and de-identify it to make it actionable. That’s where Machine Automation can help. Digamma worked with several technology companies to develop and AI algorithm and training it to perform at its best to help physicians gain insights into patient’s medical history. There is a growing number of companies, including Medal, Inovalon, Telegenisys that process raw, unstructured and previously unusable healthcare data and preset it to physicians in a timely and relevant way.

Your Algorithm is Only as Good as its Data

Poor quality of healthcare data is the main enemy to the widespread, profitable use of Machine Learning. As AI-powered systems become the next frontier in healthcare, the data used to train these systems is crucial to ensure accuracy.

“If the data is flawed, biased, incomplete, or doesn’t accurately represent a population of patients, then any algorithm relying on the data is at a higher risk of making a mistake,” said Dmitriy Tochilnik, President and CTO of Dicom Systems“My team is focusing on securely de-identifying the clinical images that would serve as the raw, usable materials to develop neural networks. By “usable”, we mean that the raw materials fed into the algorithms can actually teach it. De-identifying images is crucial in and of itself to doing Machine Learning, but it’s the right thing to do to protect patient privacy.”

It’s important to think strategically about data and reserve enough time to label, annotate and prepare it for analysis. Recognizing the potential for built-in biases and errors in the data and designing adaptable, dynamic AI algorithms data scientists can build better defenses against common AI shortcomings.

Conclusion

Even though functional Artificial Intelligence, has yet to be achieved in healthcare we are starting to see some concrete use cases. Digamma.ai was fortunate to be involved in several interesting Healthcare AI projects, most recently with MedNition, Inc., healthcare technology company that applies Machine Learning to provide physicians and clinical practitioners with real-time clinical decision support to improve patient safety and operational efficiency of hospitals. Our team analyzed patient’s data and played a key role in setting up MedNition’s Machine Learning framework. With new use cases emerging every day, AI continues to be a focus across the digital health ecosystem.

Most HIMSS19 attendees agreed that that the best opportunities for Artificial Intelligence in healthcare in the upcoming years are hybrid models, where physicians are supported in identifying risk factors, diagnosis and treatment planning. This approach will result in faster adoption of AI by healthcare providers and will start to deliver operational efficiency and improved patient outcomes.

Leave a Reply

Your email address will not be published. Required fields are marked *