Several studies have already shown [7; 8; 10] that ARFI imaging with VTIQ can aid in the differentiation of malignant from benign breast lesions.
Our study assessed the added value of ARFI imaging to B-mode breast US and demonstrated a tendency towards higher diagnostic accuracy for most of the examiners.
Moreover,
all examiners reported a higher diagnostic confidence through the combined B-mode and ARFI reading.
Several SWV cut-off values have been proposed in literature.
In concordance to previous work from our group [10],
ROC curve analysis in our present study suggested an optimal threshold of 3.23 m/s with a sensitivity of 85.71% (95% CI 76,4-92,4%) and a specificity of 76,79% (95% CI 67,9-84,2%).
However,
there is significant overlap between benign and malignant lesions,
something that limits the value of these cut-off values in everyday clinical practice,
although they represent an optimal mathematical trade-off between sensitivity and specificity.
It is desirable that a diagnostic test can rule in or rule out malignancy.
In our study,
by applying a SWV cut-off value of 6.5 m/s,
the probability of malignancy of a lesion was shown to be higher than 95%,
in concordance with the BI-RADS category 5 (highly suggestive of malignancy).
Thus,
it is reasonable to upgrade a lesion that looks otherwise less suspicious in B-mode US,
if it shows a SWV higher than the aforementioned cut-off value.
On the other hand,
a SWV threshold of 1.9 m/s leads to a sensitivity of almost 99%.
This allows for a downgrade of otherwise suspicious (i.e.
BI-RADS 4) lesions to the BI-RADS 3 category,
without missing any significant number of cancers,
while at the same time reducing the false positive cases of B-mode US and the number of false positive biopsies.
In fact,
the only case that would be missed in our study was a 7mm Grade 1 DCIS,
most likely due to the lack of any significant desmoplastic reaction in the surrounding parenchyma.
It has to be noted,
that these cut-off values have been calculated in examinations with a very low precompression.
There is an unavoidable degree of inter-examiner variability,
although ARFI imaging (similarly to SWE) shows a high reproducibility among different examiners [10; 11].
One limitation of our study is its retrospective nature,
which allows for a possible selection bias.
The cases included were found by searching the local PACS and there was no way of controlling,
in which cases the examiner had chosen to apply ARFI imaging.
It has to be noted though,
that ARFI imaging is routinely used in our department.
A further limitation is that when performing the combined reading,
the readers did not strictly adhere to any quantitative cut-off values; the BI-RADS category was assigned by an intuitive evaluation of the color-coded map or the SWV measurement or both.
Finally,
our study was performed in a single institution,
with a relatively small cohort.
Larger,
multicentric trials,
with more examiners and a larger number of participants are necessary to validate our results and more specifically the “rule-in” and “rule-out” thresholds proposed.
In conclusion,
our study demonstrates that ARFI imaging with VTIQ can aid in the differentiation of malignant from benign breast lesions and that the application of “rule-in” and “rule-out” thresholds is feasible.