Mammography Reporting: Where Weve Been, Where Were Going
Practice reporting has had a tumultuous history. Specifically in the U.S., mammography reporting has suffered from a lack of uniformity and standardization, which changed with the implementation of the American College of Radiology’s (ACR) Breast Imaging and Reporting and Data System (BI-RADS).  While limitations still remain for breast imaging, emerging technologies and software applications are finessing the mammography reporting technique, and paving the way for further homogeny and an overall improved patient experience.

A rocky start

Free-text reporting has been the standard method used by radiologists, despite increasing use of speech recognition and structured reporting, says Bruce I. Reiner, MD, of Baltimore Veterans Affairs Medical Center in Baltimore. “At the end of the day, radiology as a whole has undergone tremendous transformation and innovation, and reporting really hasn’t,” says Reiner. 

One exception is mammography which has successfully developed standardization to the ACR BI-RADs reporting methods.

BI-RADs arose in the early 1990s out of the need for simplicity and uniformity. “The clinical community was up in arms, saying that you’re giving us all this descriptive information in these free-text reports, but we don’t know what to do with it and we don’t know what the take-home message is,” Reiner recalls. Because of the “heterogeneous nature” of the reporting, the clinical community sought to better understand the clinical significance, as well as follow-up recommendations.

The heterogeneous nature of the reporting stemmed from the fact that there was a tremendous amount of variability with regard to the radiologist, explains Reiner. “Depending on the time of the day and the day of the week, as well as the stress level of the physician, the contents and the organization of the report could be all different,” he says. Essentially, BI-RADS created a more uniform reporting method that has been “very well received by the clinical community.”

But BI-RADS also presents other challenges, namely, tthat it is scored according to the single most significant finding, oftentimes “camouflaging” the other findings that are also significant, explains Reiner. 

To remedy this, a more itemized report needs to be created for the clinical community, to spell out all significant findings of the study. According to Reiner, these reports should give clinical significance, anatomic location, modifiers regarding size, morphology, follow-up recommendations and if possible, a diagnosis or differential diagnosis.

“But in order to do that,” he warns, “which the clinical community would really welcome, there would really need to be a certain amount of constraints on reporting,” he says, referring to the free text report as a historic “security blanket” within the community. The implementation of itemized reporting could potentially lead to data mining, the most important feature of the mammography report, as well as the future of evidence-based medicine, Reiner offers. “You cannot mine radiology reports unless you have standardized data, so it becomes circular,” he says, explaining that unless there is a standardized lexicon and structure within the reports, evidence-based medicine cannot be achieved. “This is where we need to go,” he notes.

BI-RADS smoothes out the bumps

Wendie Berg, MD, PhD, FACR, of American Radiology Services, Johns Hopkins Medical Institutions, in Lutherville, Md., agrees with Reiner, saying that not only is mammography reporting the most standardized area of radiology, but possibly of all areas of medicine as well.
“It is extremely standardized,” she offers. The report always follows the same model: breast density is first quantified into one of four categories, followed by a description of the findings for which there is a standard terminology, or lexicon, followed by standardized wording for final assessments and recommendations.
“Two people looking at the same lesions generally, with a high degree of consistency, will describe [the finding] the same way. This has been in place since the first BI-RADS lexicon was produced in the early 1990s,” she says.
But despite its organization and consistency, “It’s still an art, and it’s still difficult,” Berg says. Focus can be a challenge for the radiologist who is not a dedicated mammography reader, and may be in the middle of reading studies from other areas of radiology.
“The [reader] may be an excellent radiologist, it’s just very difficult to be an expert or have the same level of performance if you are being interrupted or reading other studies at the same time. I think that to some extent in this country, we have shied away from having a dedicated radiologist doing the breast imaging,” she says. In most other countries, she notes, mammography is offered in specialty centers for women with a dedicated breast imager.

Currently, Berg observes that more mistakes are being made in the characterization of lesions than in detection. “I think that people are getting better. Maybe as more people are using [CAD] to help mark things in the first place, they are seeing more things.”
With standardization in place that creates a more uniform report, what lies ahead for mammography reporting is the consistent collection of outcomes, Berg forecasts. However, data are not collected at a national level, she says. “While there are some programs that do allow [for data collection], it is much less systematic.”

There is still room for improvement, says Berg. With regard to interpretive skills, “there is a lot of potential variability, and we don’t control or test that at all right now,” she explains. However, training in BI-RADS can make radiologists more accurate in their interpretations, Berg believes.

Accreditation takesmammography up a notch 

Constance Lehman, MD, PhD, professor and vice chair of radiology and director of breast imaging at the University of Washington Medical Center in Seattle, believes that breast imaging reporting has served as a model for other areas of radiology due to the development of standard definitions and approach to the exam. “We hope to use the model created by breast imaging to also create standardized reporting for other imaging exams—whether a chest CT or a head MRI.” she says.
Without having a standardized method of reporting, it is difficult to track outcomes and results of imaging exams. “We need to know how was that exam assessed? And was it assessed as being a normal exam or an abnormal exam. If you have a long report with several paragraphs describing various findings, but you don’t know at the end of the day if that exam was read as abnormal or normal, it’s really hard to access the outcomes of the program.”

However, mammography reporting isn’t the only portion of the practice that has come under review. Decades ago, an assessment found that the quality of mammography varied widely across the U.S., leading to the ACR to develop a program to accredit centers offering the practice to women, explains Lehman. “These events led to the Mammography Quality Standards Act which requires accreditation of all mammography facilities in the U.S.,” she says.

Next to follow suit in accreditation were breast ultrasound programs, followed by ultrasound-guided biopsy, stereotactic biopsy and accreditation components for digital mammography. “Now we finally have the final component of the breast imaging center which is breast MRI,” she says, with regard to the Breast MRI Accereditation program by the ACR launched in May.

Next step: Image integration

“I think it would be fair to characterize that BI-RADS have been a complete and definitive answer and a great start, but there is still a long way to go,” says Reiner, noting that the structured concepts behind BI-RADS should be extended throughout radiology.

In terms of technology for the future of reporting, Reiner says the goal needs to be more dynamic reports that link the report with images they describe. He calls the disassociation between the reports and images a “disservice to clinical management.”
“Right now, the current model of having a standalone report just distinct and separate from the imaging data makes no sense,” he explains. “The imaging and reporting data should be integrated into one type of display so that the individual reader, whether that is a radiologist, clinician or technologist can access the imaging data.”