At last, AI is coming to a PACS near you—and it’s all to the benefit of your radiologists, rad administrators, and collaborating clinicians.
Carestream and Zebra.
Carestream is partnering with Zebra Medical Vision, an expert in deep learning—teaching computers to read and analyze medical images. The name, Zebra, comes from their ability to help radiologists recognize the “zebras” and distinguish them from the more common “horses” they see every day. Immediately, Radiology Assistant provides insight into four key areas where the population is at risk: Osteoporosis, Fatty Liver, Lung Emphysema, and Coronary Calcium. By providing real-time diagnostic insights to the radiologist when risk factors in these four areas are present, outcomes for affected patients can be significantly improved. It is this type of enrichment that expands the role of radiology in the healthcare environment, and adds to its value.
How does Zebra do it?
Zebra has developed algorithms based on deep learning in computer image recognition that rely on a proprietary neural net analysis to generate reliable findings. These algorithms have been approved by the FDA to provide a normal/abnormal indication to a physician who can then review the automated findings and decide on further actions to be taken.
Here’s an example of how Radiology Assistant would alert the radiologist to an abnormal finding based on its analysis of the images shown here—note the alert box at the bottom right.
It’s up to the radiologist to decide how (or if) to include a reference to this finding in the final report.
The larger picture: the role of analytics in the enterprise image platform.
It is Carestream’s vision that analytics can help clinicians and administrators understand and improve patient care if information is available on every aspect of the workflow. This begins with a macro view of data (Population Health) and provides algorithm results for patient studies in a histogram fashion. These results can then be filtered for system, site, date range etc. In addition, the suite includes clinical analytics comprised of imaging, reports and dose—as well as business and operational analytics.
For the Radiology Administrator/Director, this analytic solution is essentially a risk stratification tool with which customers can determine the level of risk that is present (for the conditions checked by the algorithms) in their patient population. This can be the key to significant insight into future direction, while the analytics can offer immediate benefits for determining risks which can lead to improving outcomes and reducing hospital admissions for at-risk patients.
The business case for advanced analytics.
In the US, the move from Fee For Service to Value-Based Care, including Risk Adjusted Care factors is well underway and most hospitals and imaging centers are learning to find their way around the issues. For all, a reduction in hospital admissions from high-risk patients provides a benefit over and above its worthiness as a clinical goal. For example, identifying patients at risk for osteoporosis will lead to a treatment plan that may include additional imaging, while it reduces hospital admissions for this frequent condition and provides the basis to adjust risk profiles to increase risk-adjusted reimbursement. Because the risk factors are so prevalent, many facilities can justify the cost of the analytics in savings and/or incremental revenue. (Ask your Carestream representative to help with the business case for your facility.)
PACS Administrators need to know.
PACS administrators need to understand the power of the improved AI that Zebra has developed to provide imaging analysis and real-time indications for physicians. Radiology Assistant is available as an immediate option for Carestream PACS 220.127.116.11 and higher. (Population health is also available for any Carestream PACS version.) But the installation and training of these advanced systems will be lead by Carestream and Zebra—it is not intended to add to the workload of busy PACS administrators who have more than enough to keep them occupied. Long term, these analytics are designed to improve outcomes and help radiologists become more efficient and effective in the use of their systems—goals to which we can all subscribe.
This is just the beginning. We are working to secure regulatory approvals for at least four more high-risk condition analyses to add to the Radiology Assistant module, as well as on new quality control software that will speed up and improve quality review of radiology reporting. Stay tuned to VIBE for progress as we add capabilities to our Imaging Analytics System.
[Released in the United States. Please contact us for availability in your geographic location]