• Rebecca Anderson

Blog: The Current State of Machine Learning in Radiology: What It Can, Can't, and Shouldn't Do

Updated: Nov 17

(Written for Change Healthcare)

Machine learning in healthcare is no longer a “might-be.” Even within radiology, there are already numerous systems that are learning to make important decisions with little or no human input. As machine learning becomes more sophisticated, experts anticipate even more applications in the future.


But that doesn’t mean radiologists should start brushing up their resumes. Read on to learn about machine learning tech that’s available now, what’s on the horizon and what work humans will always do better than machines.

Machine Learning Tech Available Now

Technology advancements have been always been key to driving radiology and imaging forward. So, it should be no surprise that radiology is at the forefront of machine learning within the healthcare space. Here’s how radiology departments can benefit from machine learning today:


Radiology Reporting

Machine learning in radiology is now able to expedite the reporting process through aggregate analysis of unstructured data, such as free-text or narrative content in the patient record. The software can uncover key findings and track recommendations made by the radiologist to the referring physician.


Analysis of this type of data has been historically difficult, or even impossible, without manual intervention. The benefits of machine learning in reporting are therefore significant. Radiologists who use intelligent reporting solutions spend less time gathering essential data for a comprehensive radiology report.


Order Scheduling and Patient Screening

With patient no-shows rates rising, missed appointments can have a serious impact on departmental revenue. Machine learning can identify patients at risk of skipping radiology appointments and help by providing preemptive reminders. Machine learning can also improve patient screening by analyzing data pulled from the electronic health record (EHR) to identify potentially relevant cases.


A current example of this technology is Smart Connect and Smart Appointment Scheduling from Change Healthcare. This system reminds patients of appointments via multi-channel, multi-language communication options including IVR, email, SMS text, web portal, and live advocates. In addition, patients identified through screening are sent targeted education materials with dynamic, integrated analytics. The automation of key communications to current and potential patients helps ensure that people receive the treatment they need.

Workflow Management

A workflow management system rooted in machine learning can take radiology to the next level of efficiency. Based on what the system learns, it will be able to identify patterns and balance workloads by automatically prioritizing cases based on medical and operational factors, transmitting radiology results, and keeping a permanent record of findings and communications.


This type of intelligent workflow is available as part of some enterprise medical imaging solutions, helping healthcare systems save time and money, improve quality of care, and meet regulatory requirements.


Machine Learning Tech on The Horizon

While machine learning has already revolutionized radiology, there are even more advancements on the horizon. Here’s what radiologists can expect in the future:


Image Acquisition

An ongoing area of research focuses on making imaging systems intelligent. Machine learning could soon decrease imaging time, reduce unnecessary imaging, improve positioning, and help characterize the findings. A key benefit of improved accuracy and efficiency? Quicker exam times and potentially reduced dosage.


Detection

Machine learning may also help detect critical findings and flag them for human interpretation. This has huge implications in the identification and management of breast cancer, pulmonary nodules, and in-bone age analysis. It also holds the promise of improved patient care, particularly in pediatric radiology and endocrinology.


Radiation Dose Management

Experts believe that machine learning algorithms could help radiologists determine the optimal radiation dose before exams. This guidance could be invaluable as radiologists strive to deliver the lowest dose that will achieve the desired effect.


Image Quality Analysis

Ensuring medical image quality has long been the realm of trained human observers. As the need for radiology imagery increases so does the need for quality checking by human eyes. There is the possibility that machine learning will shoulder some of the responsibility for image quality analysis in the future.


What Machine Learning Can’t or Shouldn’t Do

The advent of machine learning isn’t the first time that disruptive technology transformed patient care, and it won’t be the last. While machine learning comes with many promises, there will be an ongoing need for radiologists.


While machine-like precision can be valuable for an accurate patient diagnosis, a human connection is critical, especially since a visit to radiology is typically not something patients look forward to. By shifting the responsibility for repetitive tasks such as reading images and writing reports to automated systems, radiologists might be able to spend more time away from their computer screens and be more visible participants in the care continuum.


Learn How Machine Learning Can Transform Your Radiology Department Today

Someday, machine learning may produce a fully AI “radiology assistant.” But that day is a long way off. In the meantime, there are still plenty of opportunities to use AI and machine learning to increase efficiency and help improve quality of life for radiologists and patients alike.


Enterprise Medical Imaging Solutions offered by Change Healthcare are at the leading edge of machine learning. Contact us to learn how machine learning can improve efficiency and patient experience within your radiology group.


 

©2018 by Rebecca Ann Anderson