|ECR 2018 / C-1623|
|Estimation of breast density using radio wave radar imaging techniques|
Methods and materials
Patients attending symptomatic clinics at three sites were identified by clinicians as having a palpable lump. Following informed consent, eligible patients underwent a MARIA® scan. The bilateral reconstructed 3D images were used in this study. The density classification process was accessed by comparison with BIRAD density scores estimated from mammograms.
MARIA® is a CE-marked radio-frequency (RF) medical imaging system  comprising a patient bed, a Scanning and Data Processing (SDP) unit which is located under the bed and a touch-screen console featuring a software user interface that controls both acquisition and review - see Fig.1. The SDP unit contains a hemispherical array of 60 RF antenna  which encircle the breast. The breast lies pendant in the array through a hole in the bed and one of a set of conformal ceramic inserts, placed into the array are provided to fit a range of breast sizes without the need for compression (Fig. 1). By serially energising each antenna, scattering parameters are collected from the signals received at all remaining antennas. As signals propagate across the breast, their magnitudes and phases are modified according to the dielectric properties of the tissue traversed.
Description of the procedure:
Patients were scanned before any surgical or biopsy intervention. Data collected were BIRADS score, age, menopausal status, and breast size. The subject was required to lie prone with the breast inserted into the ceramic cup. A coupling fluid , similar to hand moisturiser, is applied between the breast and insert to maintain good RF contact. The scan consisted of checks for goodness of fit of the breast (lack of air gap) and then at least two scans of about 30 seconds each. The total scan time with patient present was typically less than 5 minutes.
A flow diagram of the evaluation process can be seen in Fig. 2. A 3D image of the patient is created from the scattering parameters (Fig. 3) using a modified version of the classical delay-and-sum (DAS) beamforming algorithm . A median filter is applied to reduce image noise and a histogram of pixel intensity is produced (Fig. 4). The area under histogram for the highest 35% of pixel magnitudes is calculated and used as a metric in a linear classification procedure. Breasts are classified as being either "lucent" or "dense". The position of the threshold is obtained from a training set of cases with known density characteristics, for example, BIRAD scores estimated from mammograms.
Thematically related posters
ECR 2018 / C-2105
Breast Diseases Involving the Whole Breast : Clinicopathologic and Imaging Features
ECR 2018 / C-1838
Multiparametric MRI in prostate cancer: a radiomic study on different diffusion and perfusion models
ECR 2018 / C-2194
The role of breast magnetic resonance imaging in the preoperative management of patients with invasive lobular carcinoma