Patients
Institutional Review Board approval was obtained for this retrospective,
cross-sectional study.
All diagnostic US examinations of the breast,
which included ARFI imaging with VTIQ,
performed at our institution between June 2013 and June 2015,
were retrospectively reviewed by searching the local PACS system.
In order to be eligible for inclusion in the study,
an examination had to meet following criteria: an abnormality visible at breast B-mode US with at least 4 mm in size (BI-RADS 2-5); no previous biopsy or neoadjuvant chemotherapy; color coded ARFI image of the lesion with quantitative measurement of its SWV; histologic confirmation of the lesions or stability for at least 18 months of follow-up.
Finally 189 patients with 196 sonographically evident lesions were identified,
who fulfilled the above criteria and were included in the study.
Data acquisition
All examinations were performed on a Siemens Acuson S3000 US device (Siemens Medical Solutions Inc.,
Mountain View,
CA,
USA).
A linear 18L6HD transducer was used for routine B-mode imaging whereas a linear 9L4 transducer was used for ARFI imaging with VTIQ.
On the color-coded ARFI image of the lesion,
the examiner placed a quantification ROI at the stiffest area of the lesion,
in order to measure the SWV.
In cases when the system could not achieve a valid measurement and only indicated “High” (that is higher than the maximum measurable velocity of 10 m/s) it was replaced by a value of 10 m/s.
The ARFI image was obtained by applying minimal compression; that is the least possible compression,
in order to achieve a reasonable,
artefact-free image,
since it has been shown that a high degree of precompression when performing SWV measurements leads to an artificial stiffening of tissue and thus measurement of higher velocities [9].
Data analysis
4 radiologists,
other than the ones who had performed the examinations or searched the local PACS,
independently reviewed all B-mode images,
blinded to clinical,
mammographic or histopathologic results.
Using the descriptors published in the BI-RADS lexicon,
each one assigned a BI-RADS score for each lesion.
Subsequently the ARFI image of each lesion was reviewed combined with the B-mode image and a new BI-RADS score for each lesion was assigned.
Receiver operating characteristics (ROC) curve analysis was used to calculate the respective diagnostic performances of sole B-mode and combined B-mode and ARFI imaging.
Additionally ROC curve analysis was performed,
in order to specify “rule-in” and “rule-out” thresholds for the probability of malignancy.
To do this,
the diagnostic SWV threshold to differentiate between benign and malignant lesions was varied.
At appropriate cut-off values the sensitivity,
specificity,
positive and negative predictive values were calculated.
Statistical calculations were performed using the software SPSS 20 (IBM Corp,
NY,
USA) and MedCalc 12.7.5.0 (Mariakerke,
Belgium).
The standard of reference was either histopathology or follow-up stability for >18 months.