After a two-year hiatus because of the Covid pandemic, the 16th International Workshop on Breast Imaging took place in Leuven, Belgium last week. There was a full programme of 80 oral presentations and 22 posters, with keynote addresses by luminaries such as Martin Yaffe from Toronto and Etta Pisano from Harvard. An award for one of the best scientific presentations, sponsored by Hologic, was given to Melissa Hill for her paper entitled “Mammographic compression pressure as a predictor of interval cancer” which will shortly be published by SPIE.
Hill is a consultant imaging scientist at Volpara Health and collaborated with Mark Halling-Brown, who is head of scientific computing at the Royal Surrey County Hospital in Guildford, England, and curator of the OPTIMAM database that was used in the study. Their purpose was to identify mammographic image quality indicators predictive of interval breast cancers as opposed to screen-detected cancers. As we highlighted two years ago, 15% of breast cancers are diagnosed after the patient is screened mammographically with negative results but during the interval before the next recommended screening examination.
One of the key parameters measured was pressure – force divided by area – applied by the compression paddle (seen above right) to the breast by the radiographer. The major finding was that compression pressure (CP) was found to be a significant predictor of interval cancer (IC) versus screen-detected cancer (SDC), with higher CP being associated with a lower risk of IC. Their results demonstrated the lower limit of CP to be 7 kPa, with pressures below that increasing the odds of having an IC. The upper limit is about 13 kPa, with a group from Amsterdam arguing that the optimum CP is 10 kPa – or 75 mmHg – which is the diastolic blood pressure.
The figure above left (credit: OPTIMAM database under licence from Cancer Research Technology) shows cranio-caudal mammograms at the prior screening exam and then at the time the interval cancer was detected 1.7 years later. Interestingly, the CP at the prior screening exam was just 3 kPa, which may have played a role in the false negative finding.
There are some important take-home messages and one of these was shared by Hill (seen right): “Taken together, our results and those from the literature suggest that radiographers should account for the breast size when they compress. The idea is that rather than having a target compression force – which some guidelines suggest – radiographers should aim for a target pressure, specifically between 7 and 13 kPa. They must understand the relationship between compression force, contact area with the paddle, and pressure.”
A few problems:
1. If you look at enough variables, one is likely to be “statistically significant” simply by chance.
2. On the images provided, it was not the compression that was the important factor, but rather the poor positioning of the breast. The cancer was likely just off the image. There are several small densities at the edge of the image where the cancer was ultimately diagnosed one being the edge of the cancer. Clearly there is breast tissue that is not imaged beyond the edge of the mammogram. The image should have been repeated (or the patient recalled for additional imaging based on these asymmetries which have no counterparts in the opposite breast).
3. Insufficient pressure could result in motion of the breast blurring the image that might lead to a cancer being missed, or the fact that the tissues are not spread sufficiently and a cancer hidden in the normal tissues, but it is likely that pressure is a secondary indicator.
4. True-true and unrelated?? The fact that Young’s modulus may converge between normal tissue and breast cancer with increasing pressure has little to do with the conspicuity of the cancer. The question is whether or not the compression can spread the structures around the cancer so that normal tissues do not superimpose and obscure the cancer. Also pushing the lesion closer to the detector reduces geometric unsharpness.
There are three ways that cancer can be evident.
a. Its margins are visible due to adjacent fat highlighting its (usually) irregular margins. With insufficient compression the superimposition of normal tissues can hide the cancer.
b. The tissue architecture is distorted by the presence of the cancer or cicatrization due to tissue response to the cancer.
c. Calcifications in the in situ portion of the cancer
5. In theory the more compression the more the structures can be separated, but at some point the breast cannot be compressed any further. This is taught to the technologists as when the skin is “taut”. Additional compression only causes pain with no benefit for the imaging.
Dear Dr. Kopans,
Many thanks for your astute comments. These are excellent points and I agree that there are multiple additional factors that need to be considered related to lesion conspicuity. There are likely some important relationships buried in these ‘population level’ statistics.
Regarding the images, this figure was included in the paper (with all 4 views) as an illustration of both breast positioning and compression pressure deficiencies. Agreed that the CC view positioning likely affected the reading, although the MLO view did appear to include the relevant tissue.
Regards,
Melissa