Artificial intelligence, machine learning find role in radiology
Artificial intelligence and machine learning capabilities are beginning to make an impact within radiology, as vendors start rolling out initiatives to assist professionals in making diagnoses.
The radiology profession is ripe for technology—as radiologists deal with an increasing number of images and bear more responsibility in the clinical process.
For example, radiologists now are being called upon to determine what course of treatment might be less invasive, thus reducing cost, patient recovery times and the risk of complications. Or they typically are asked to assess the rate of progression of a disease such as cancer, to determine what course of treatment is most appropriate.
The use of advanced technology could be considered disruptive and perhaps threatening to some radiologists, but it will become essential for professionals to do their jobs effectively, says Leo Wolansky, neuroradiologist and professor of Radiology at University Hospitals, Cleveland.
“The role of the radiologist has changed so much,” said Wolansky at the recent annual meeting of the Radiological Society of North America in Chicago. “It used to be that we were just asked to distinguish black from white. Now, we’re asked to tell referring physicians what is the percentage of white and black and gray, or how much that percentage has changed over the years. There is criteria for what to do and when to intervene based on what we tell them.”
Because of the growing need for accuracy in assessments, applications that can assess images becomes important, Wolansky adds.
“With software, we can see changes in diseases such as multiple sclerosis over time,” he says. “”If a patient comes back with more lesions, that can impact how the patient is treated. That’s something software can do—it can compare images and find a new lesion, or indicate where a lesion might be. So if a followup scan shows an increase in lesions, that could have an important implication for treatment.
“There’s an explosion of information confronting radiologists, and there’s a need for speed in processing it for patient care,” he adds. “With stroke, the window of opportunity to effectively treat the patient means an assessment has to be done in minutes, and radiologists need to look at images and be able to make a diagnosis instantly.” That’s where technology can assist, he says.
Vendors are beginning to factor in artificial intelligence and other capabilities to assist imaging professionals in their efforts.
For example, Royal Philips has developed IntelliSpace Portal 9.0, the latest edition of its comprehensive, advanced visual analysis and quantification platform. The product, introduced at RSNA, helps radiologists detect, diagnose and followup on treatment of diseases, while using new machine learning capabilities to support the physician. The solution addresses radiologists’ needs for tools to better support the growing group of patients with brain injuries and neurological disorders such as dementia, strokes, amyotrophic lateral sclerosis (ALS) and multiple sclerosis (MS), says Jeroen Tas, business leader for connected care and health informatics for Philips.
Also at RSNA, Philips unveiled Philips Illumeo, an imaging and informatics technology which uses adaptive intelligence to redefine and enhance how radiologists work with medical images. Its built-in intelligence records the radiologists’ preferences and adapts the user interface to assists the clinician by offering tool sets and measurements driven by the understanding of the clinical context, Tas says.
In another demonstration at the conference, IBM Watson Health demonstrated capabilities that it is developing with Merge Healthcare, another IBM company. It’s working in areas designed to help healthcare professionals pursue personalized approaches to patient diagnosis, treatment and monitoring.
Artificial intelligence and machine learning are ideal for imaging work because diagnostic images are the largest and fastest growing data source in the healthcare industry, with IBM researchers estimating that they account for at least 90 percent of all medical data today.
“Tools to help clinicians extract insights from medical images remain limited, requiring most analysis to be done manually,” says Anne LeGrand, vice president of imaging for IBM Watson Health. “This has created an opportunity to analyze and cross-reference medical images against a deep trove of lab results, electronic health records, genomic tests, clinical studies and other health-related data sources.”
That kind of analysis will help providers compare new medical images with a patient’s image history, as well as populations of similar patients, to detect changes and anomalies, LeGrand added.
IBM Watson Health demonstrated a variety of capabilities at RSNA that are under development but show promise, including:
- A cognitive peer review tool intended to help healthcare professionals reconcile differences between a patient’s clinical evidence and data in that patient’s electronic health record.
- A cognitive data summarization tool intended to provide radiologists, cardiologists and other physicians with patient-specific clinical information to use when interpreting imaging studies or when diagnosing and treating patients.
- A cognitive physician support tool intended to help doctors personalize healthcare decisions based on integrating imaging data with other types of patient data.
- The MedyMatch “Brain Bleed” App, a cognitive image review tool intended to help emergency room physicians
diagnose a stroke or brain bleed in a trauma patient by identifying relevant evidence in a patient record.