No two words struck more fear into the hearts of radiologists in 2016 than “artificial” and “intelligence,” unless you added “deep learning” to the mix. Some radiologists have expressed concern that big imaging data could continue feeding the deep-learning beast to the point where computing power becomes robust enough to replace the radiologist entirely. We at Sg2 hedge our bets on, “Probably not.” However, if radiologists are willing to embrace and collaborate with artificial intelligence (AI) technology, they may position themselves to lead all of health care deep into the promising new world of precision medicine (the emerging approach to treatment and prevention that takes into account variability in genes, environment and lifestyle by individual).
In December, Chicago hosted the 102nd annual meeting of the Radiological Society of North America (RSNA). Major vendors highlighted their evolving programs to harness big data, incorporate artificial intelligence, and deliver dashboards that provide contextual data on operations at a more granular level than ever before, including all the basic features of an asset management product; e.g. usage, dose management, performance, turnaround times.
From GE’s Health Cloud to Philips Illumeo “adaptive intelligence” product to Siemens Healthineers recent partnership with IBM Watson Health, vendors attempted to ride the wave of the AI theme and demonstrate a growth platform for the future, even though many were only in the very early stages of development. There were, however, more focused applications of AI and deep learning that showed early promise in areas such as stroke assessment, helping clinicians better assess parenchymal patterns for breast cancer risk and reducing unnecessary biopsies of microcalcifications found on mammography.
From a technology perspective, this conference was more evolutionary than revolutionary. Vendors continue to refine their product configurations by adding new value offerings across all modalities, with products that provide more comprehensive features at lower price points. This was in line with Sg2’s take-home message on radiology services: “The 20th century was all about the hardware, while the 21st century is all about the software and services.”
Vendors emphasized new approaches to harness big data, AI
RSNA received a record number of submissions for the scientific sessions this year, with over 1,700 presentations throughout the week. Summaries of the key themes by modality follow.
- Ultrasound Imaging: Sessions covered the ways in which ultrasound can help clinicians improve their staging and treatment plans, distinguish between benign and malignant lesions, and track healing. Attendees also heard about research regarding the use of high-intensity focused ultrasound to treat fibroids, how musculoskeletal shoulder ultrasound can help evaluate rotator cuff injuries, the benefits of using weight-bearing ultrasound to triage patients with medial knee pain, and how contrast-enhanced ultrasound can characterize malignant versus benign focal liver lesions.
- Magnetic Resonance Imaging (MRI): Attendees heard updates from a number of studies that featured multiparametric MRI for applications ranging from cardiology to neurology. MRI also allows researchers to delve deep into the brains of Alzheimer’s disease patients and patients with mild cognitive impairment to see where these conditions manifest. Also discussed was the crucial information provided by MRI on the potential link between concussions and traumatic brain injuries and degenerative neurological disorders, such as chronic traumatic encephalopathy (CTE). MRI even holds the promise of diagnosing CTE while a patient is still alive, instead of the current accepted practice in which diagnosis is made at autopsy.
- Computed Tomography (CT) Imaging: Physicians are being aided in their search by radiomics—the science of extracting quantitative information from imaging data. Through radiomics, they have mined CT’s fertile ground to the point where they can now acquire information that might have previously been attained through MRI or nuclear medicine applications, or simply may not have been available at all. Spectral imaging, iodine imaging, K-edge imaging, photon-counting CT, iterative reconstruction and other quantitative imaging techniques were on display in this year’s sessions, offering new ways to see data hidden inside images. To take just one example, metastatic epidural spinal cord compression is a critical finding impacting survival, and quantitative CT can now find it.
- Molecular Imaging: Key themes included the development of novel radiotracers to explore neurodegenerative conditions like the research being done at the Washington University School of Medicine in St. Louis, MO, where researchers have developed a positron emission tomography (PET) tracer known as 18F-AV-1451 to track neurofibrillary tau pathology in vivo and its relationship with atrophy in the brain and Alzheimer’s disease.
We believe the future of medical imaging is a bright one, with enhanced capabilities to provide information that goes beyond the simple anatomic and morphologic interpretation of images. The technological advancements in AI and deep learning coupled with the professional skills of radiologists will enable imaging to play an essential role in both value-based care and precision medicine for the foreseeable future.
To find out how Sg2 can help you define your path forward in imaging and health care in 2017, click here.
About the author. In his role as a vice president at Sg2, Henry Soch analyzes the impact of medical technologies on care delivery, especially in cardiovascular services, neurosciences, oncology, orthopedics, surgery and physician practices. As a digital health evangelist and a passionate promoter of digital health technologies, e-learning and disruptive innovation, he is a leading voice in the convergence of technology and health, helping to define, dissect and deliberate global trends.
Sources: Radiological Society of North America. RSNA 2016. November 27–December 2, Chicago, IL; Sg2 Analysis, 2016.