Finding cancer in a mammogram is difficult, especially with women who have dense breast tissue that can mask cancer on mammograms. Over 45% of women globally have dense breasts accounting for 71% of all breast cancers.
Healthcare quality in general and Mammography quality in particular, are hot topics (170k hits for mammography quality on scholar.google.com). The American College of Radiology (ACR) and the Mammography Quality Standards Act (MQSA) have established quality standards1. These standards have provided fertile soil for a micro-industry of “quality improvement” tools, databases, and analytical software. But few quality tools provide in-workflow information that could impact the interpretation of mammograms, affecting quality before it gets to the retrospective report.
In a national study, 37% of mammographers had recall rates outside the acceptable range. Simply asking for better statistics has not led to improved scores. AI cancer detection software and improved image acquisition and display technologies continue to improve the sensitivity of mammography but that increases follow-ups.
DL Precise™ is a quality improvement tool that uses visual enhancement -- color, shape-recognition, density mapping – to delineate tissue boundaries and improve the performance of breast imagers.
Unclear visualization in mammography is associated with higher and lower call-back rates, resulting in supplemental imaging or missed cancers, respectively.
It is widely recognized that low Call Back rates result in missed cancer cases;
conversely, high Call Back rates lead to unnecessary supplemental imaging and
contribute to the problem of unwarranted biopsies. The American College of Radiology (ACR) and the FDA have established national standards for Call Back rates with the target rate between 8-9 percent. The current national acceptable range is 5-12 percent.
The annual Call Back rates of the six radiologists in the urban metropolitan New York hospital breast imaging department showed the same wide disparity reflected in national performance rates (5% to 13%). Breast imaging mammography is analyzed with the standard scoring system, the
Breast Imaging Reporting and Data System (BI-RADS.
In this study, each Reader scored the study 0, 1 or 2 and provided written comments and recommendations on a paper survey. Readers focused on the ROI when “scoring” the study. The ROI was automatically generated using an Emory University database of over 500,000 images; the ROI – designated by a green box - determined by readers enrolled in the Emory study and when applicable were confirmed with pathology reports.
• 6 radiologists in metropolitan radiology department
• 17 anonymized screening mammograms (1 ca+, 16 ca-)
• Region of interest pre-identified for reader’s evaluation
• First read without DL PreciseTM, second read with DL PreciseTM
• Evaluated changes in BI-RADS with vs without DL PreciseTM
Improved Utilization rate by 12% using DL Precise™
With the DL Precise™ intervention, the cohort made a net 12 assessment changes that moved the cohort closer to the National Performance Benchmark. Contrary to post hoc quality tools that report unassisted performance, DL Precise™ provides in-workflow image enhancement that may favorably affect Recall Rate performance at the time of reading.
The dominant change to lower utilization occurred from BI-RADS 0(Requiring Follow-Up) to BI-RADS 2 (Benign Finding). The dominant change to higher utilization was in the reverse direction. The difference between the sums of those two change types, divided by the 102 case-reader pairs, accounts for a 12% favorable change in utilization rate.
Conclusion
With only 6 readers and only 17 cases each, this simulated use testing was not powered for sweeping generalizations. However, taken as a low-power study plus two notable case reports, the data points to an important impact of DL Precise™.
Reimbursement and capitation motivate management to press for their radiologists to perform statistically within the ACR guidelines. When radiology groups consistently perform outside of the “Acceptable Range”, CMS can discount their reimbursement until they return to compliance, from 2-4% per patient. This translates into management mandates to fix the departmental performance for the adherence of standards and avoidance of reimbursement reductions anywhere from $50k - $200k per year per institution.
Furthermore, in at least one case, the intervention of DL PreciseTM was correlated with a new cancer detection. The application of DL PreciseTM is an intervention at the point of care in the normal breast imaging workflow that may improve clinical decisions toward improved quality outcomes. A properly powered, prospective study will help to solidify these initial
impressions.
View full report here:
Comments