Q&A: Automated Breast US Screening for Dense Tissue

A study recently published in the American Journal of Roentgenology concluded that combining mammography with automated breast ultrasound (ABUS) improved readers’ detection of breast cancer in women with dense breast tissue, compared with mammography alone.

According to the authors, this study also found that use of ABUS allowed for standardization and reduced operator dependence required with the use of hand-held ultrasound. In addition, imaging with ABUS was quicker (1.5 minutes) compared with the handheld examination (20 minutes). The time difference could make it easier to integrate ABUS into the screening environment, they wrote.

Diagnostic Imaging interviewed corresponding author Maryellen L. Giger, PhD, A.N. Pritzker Professor of Radiology, the Committee on Medical Physics, and the College at the University of Chicago, to learn more about the study and its significance.

What drew your interest to ABUS in this particular technique?

I have a 30-year history of working in breast cancer imaging, having investigated various multimodalities, including mammography, ultrasound, and MRI. My long-term history is in developing computer-AED diagnosis and automatic computer vision methods to extract information from breast images to help in the detection, diagnosis, and assessment of therapeutic response of breast cancer.

In addition, as an imaging physicist, I have an insight into new modalities, and a history of conducting reader studies, which are statistically rigorous methods of comparing or assessing the role of new technology in terms of their benefit to radiologists.

How would the ABUS system fit in with the goal of personalized medicine?

Think of this scenario: A woman comes in for her screening mammogram. She has the mammogram and it’s noted that she has dense breasts, and nothing was found on the mammogram; that is, it’s a mammography negative read. In that case, she would go to another imaging modality, such as ABUS for improving detection of lesions in the dense breast.  This is how the reader study was structured: Mammography alone versus mammography plus ABUS.[[{"type":"media","view_mode":"media_crop","fid":"49669","attributes":{"alt":"Maryellen L. Giger, PhD","class":"media-image media-image-right","id":"media_crop_6178111963688","media_crop_h":"0","media_crop_image_style":"-1","media_crop_instance":"6019","media_crop_rotate":"0","media_crop_scale_h":"0","media_crop_scale_w":"0","media_crop_w":"0","media_crop_x":"0","media_crop_y":"0","style":"height: 255px; width: 170px; border-width: 0px; border-style: solid; margin: 1px; float: right;","title":"Maryellen L. Giger, PhD","typeof":"foaf:Image"}}]]

Currently women with dense breast tissue may go on to have a hand-held ultrasound screening following mammography. Do you see ABUS as a replacement for this?

To me, and this is my opinion and I’m not a radiologist, if a woman comes in and has a mammogram and there’s no cancer seen, no lesion seen, instead of going to hand-held ultrasound, she would go for ABUS, because ABUS is going to cover the entire volume of the breast. You don’t have to worry about missing a region of the breast as you might with the hand-held device.

It can be difficult to get some women screened for breast cancer in the first place because of the back and forth controversy about its utility and the risk of false positives. Some studies say women should be screened often, other studies say not as often. How can we educate women about knowing that this system may be available to them, that this is something they may want to discuss with their physician?

There are laws now in some states that when a woman goes for screening, she has to be told if she has dense breasts. There are various websites and public initiatives that let people know about the importance of mammography, plus the importance of subsequent screening of dense breasts. So I think women are becoming more aware of the issue.

The physicians could explain how the dense portions in a dense breast interact similarly with the X-rays used to produce a mammogram, and thus the dense portion could potentially hide a cancer.  The cancer is camouflaged by the dense parts of the breast.  Also, since a mammogram is a 2D image from a 3D volume (that is, the breast), density parts elsewhere in the breast could also “hide” the cancer.

So by going through regular PR routes, and us talking about it-sometimes we’re asked to give talks to women’s clubs, for example-that’s another route. Women are very interested in this kind of introduction to new screening protocols.

What would you like radiologists to take away from the study?

Some people may argue that this study was a reader study and not real life. The reason we did a reader study is that if we did this in real life, we would have thousands and thousands of women having to participate in a clinical study for multiple years – potentially tens of thousands of women because the probability for cancer in the general public is low. But in a reader study, we can evaluate the new imaging system efficiently, and include multiple radiologists as readers. In a clinical study, you might have one reader per image, and here we had 17 readers per case. 

We worked very hard on developing a reader study that would mimic the clinical scenario. And keep in mind that in our reader study, we had both mammographically-negative and mammographically-positive cases.

In the study, it mentions the need for extensive training to use ABUS effectively. In a separate email interview, co-author Marc Inciardi, MD, explained the training provided:

The readers, who are radiologists, came for a full day of training at the reader study site, after receiving online training. They then had two days to read all the cases in the reader study. By the end of the reader study, it appeared that each radiologist was quite comfortable with reading the 3D ABUS images.

Reader training is a combination of orientation to the unique workstation displaying 3D data sets including “knobology,” direction on evaluation of the 3D datasets in a step-wise logical progression, and finally an approach to breast cancer screening using ABUS, with the goal of improving sensitivity without negatively affecting specificity (call-back rate). Recorded acquisition of the 3D volume ABUS datasets helps the radiologist appreciate the unique nature of the large format transducer and contrast the size difference between hand held transducer and the ABUS transducer. Understanding redundant information provided in overlapping 3D datasets improves sensitivity and specificity. Quality issues that may compromise sensitivity and specificity are addressed with discussion of potential pitfalls of positioning and compression help the reader gain a better foundation in interpretation of ABUS images.

Training is divided into four modules:

• Module 1: A one-hour peer-to-peer webinar with an initial orientation to ABUS.

• Module 2: A three-hour web-based session of self-paced tutorials of ABUS cases.

• Modules 3 and 4: The “nuts and bolts” of ABUS interpretation, consisting of live peer-to-peer interactive instruction with hands-on ABUS interpretation and assessment of image quality.

Modules 3 and 4 may be delivered in-person by a qualified Peer-Educator, or remotely through a virtual classroom that utilizes bilateral remote desktop sharing between the Peer-Educator and ABUS-Learner with live video/voice conferencing. After training is complete, each ABUS-Learner completes a final Self-Assessment of 10 cases at his/her own pace to gauge individual performance. The cases consist of cancer cases, benign findings, and normal cases. Individual results are provided to the ABUS-Learner and benchmarked against the performance of his or her peers.