This poster is published under an
open license. Please read the
disclaimer for further details.
Keywords:
Image verification, Image registration, Cancer, Screening, Computer Applications-General, Ultrasound, Computer applications, Breast
Authors:
S. S. Andrei1, J. Gao1, J. Chiu2, J. Turian3, C. Caluser4; 1Villa Park, IL/US, 2Evanston, IL/US, 3Chicago, IL/US, 4Glen Ellyn, IL/US
DOI:
10.1594/ecr2014/C-1382
Methods and materials
A realistic breast ultrasound phantom (Computerized Imaging Reference Systems Inc.,
Norfolk,
VA,
USA) was used with the FHUMS (Metritrack LLC,
Hilside,
IL,
USA) and an ultrasound machine with a linear 12 MHz probe to generate the SBC and SAM.
The FHUMS uses multiple magnetic sensors to perform the 3D mapping of the ultrasound probe and the patient's body and breast.
An experienced ultrasound operator mapped the SBC and SAM with the phantom aligned with the exam table as shown in Fig.
1.
The ultrasound cumulated probe head positions were recorded and displayed over the breast 3D diagram.
The ultrasound operator mapped once the phantom SBC in the initial position aligned with the exam table.
The SBC area and the SBC position in relation with a reference point outside the SBC,
simulating a chest wall reference which moves with the SBC movements,
were automatically recorded by FHUMS,
with the phantom in the initial position (Fig.2) and then rotated 30 degrees and 20 degrees in the horizontal and vertical planes (Fig.3),
for a total of 20 measurements,
respectively. The SBC area and the displacement of 2 fixed points at the SBC from the reference point outside the SBC were measured with a calibrated caliper and served as gold standards.
The SBC area difference in mm2 (AD) was calculated as the difference between each FHUMS area to the gold standard and the percent area displacement (PAD) was calculated as the surface mismatching between the first FHUMS area in the initial position and subsequent FHUMS SBC areas.
The linear displacement error (LD) with the breast phantom still and moved was calculated and compared with the gold standard values.
The average error (AE) and standard deviation (SD) were calculated for LD,
AD and LAD.
The breast phantom surface was scanned by the first operator following the tracks on the surface and an area of 2x2 mm2 at the intersection of the tracks was excluded (Fig.1,
2 and 3).
Five measurements were performed with the phantom in the initial position
followed by 5 measurements with the breast phantom tilted as shown in Fig 3.
The detection rate (DR) for the 2x2 mm2 area excluded in SAM was calculated by a second operator.