On-Site Validation of a Microwave Breast Imaging System, before First Patient Study

Abstract

This paper presents the Wavelia microwave breast imaging system that has been recently installed at the Galway University Hospital, Ireland, for a first-in-human pilot clinical test. Microwave breast imaging has been extensively investigated over the last two decades as an alternative imaging modality that could potentially bring complementary information to state-of-the-art modalities such as X-ray mammography. Following an overview of the main working principles of this technology, the Wavelia imaging system architecture is presented, as are the radar signal processing algorithms that are used in forming the microwave images in which small tumors could be detectable for disease diagnosis. The methodology and specific quality metrics that have been developed to properly evaluate and validate the performance of the imaging system using complex breast phantoms that are scanned at controlled measurement conditions are also presented in the paper. Indicative results from the application of this methodology to the on-site validation of the imaging system after its installation at the hospital for pilot clinical testing are thoroughly presented and discussed. Given that the imaging system is still at the prototype level of development, a rigorous quality assessment and system validation at nominal operating conditions is very important in order to ensure high-quality clinical data collection.

Keywords: breast cancer diagnosis; breast phantoms; medical radar; microwave imaging; on-site validation.

Figure 1

The Wavelia breast imaging system, which was recently installed in Galway University Hospital, for a first-in-human clinical test.

Figure 2

Wavelia microwave breast imaging system: (a) Top view of the examination table; (b) Zoomed view on the transition liquid in which the breast is immersed during the scan.

Figure 3

Microwave breast imaging examination: the principle.

Figure 4

The breast molds: (a) Bottom view of the outer breast surface mold (on the left), covered with the black 2-mm thick skin phantom, and the outer fibroglandular tissue mold; (b) Original geometry of the breast, segmented from an MRI breast image.

Figure 5

The tumor phantom: microlobulated shape, average radius of 14 mm, dielectric properties matching the measured dielectric properties of excised malignant tissue.

Figure 6

Preparation of the breast phantom for the microwave breast imaging test: (a) Immersion of the breast in the circular opening of the examination table, filled with transition liquid; (b) Inclusion of the tumor phantom in the fibroglandular tissue-mimicking liquid.

Figure 7

Realistic modeling of the near-field, non-planar, multi-layer, high permittivity transmission medium for electromagnetic wave penetration.

Figure 8

“Average” equivalent dielectric constant along a bistatic path through the breast, under various assumptions on the percentage of fibroglandular tissue $\mathrm{p}{\mathrm{c}}_{\mathrm{f}\mathrm{i}\mathrm{b}}$. Applicability to the breast phantoms used for the design validation of the microwave breast imaging system.

Figure 9

Example of focusing evaluation on a coronal breast slice. Parametric images generated for five assumptions in terms of percentage of fibroglandular tissue in the breast. Optimal $\mathrm{p}{\mathrm{c}}_{\mathrm{f}\mathrm{i}\mathrm{b}}$ = 45%, automatically selected based on maximization of the focusing metric (FM). Single tumor (dominant scatterer) detected on the specific breast coronal slice.

Figure 10

Reconstructed outer surface of the breast phantom #1 (a) Screen capture from the Wavelia Optical Breast Contour Detection subsystem, as provided to the user, (b) Zoomed bottom view of the reconstructed outer surface of the breast.

Figure 11

Reconstructed outer surface of a second breast phantom #2, of significantly bigger size, (a) Screen capture from the Wavelia Optical Breast Contour Detection subsystem, as provided to the user, (b) Zoomed bottom view of the reconstructed outer surface of the breast.

Figure 12

Breast phantom ready for the microwave imaging test: examination table top view (a) Breast phantom rotational position #1; (b) Breast phantom rotational position #2.

Figure 13

Breast rotational position #1: (a) Test Date 1, estimated outer breast surface at a given coronal slice; (b) Test Date 1, breast centering assessment map; (c) Test Date 2, estimated outer breast surface at a given coronal slice; (d) Test Date 2, breast centering assessment map.

Figure 14

Breast rotational position #2: (a) Test Date 1, estimated outer breast surface at a given coronal slice; (b) Test Date 1, breast centering assessment map; (c) Test Date 2, estimated outer breast surface at a given coronal slice; (d) Test Date 2, breast centering assessment map.

Figure 15

Experimental setup for the breast rotational position #1: (a) Photo—Top view of the breast phantom, installed on the examination table; (b) Schematic definition of the tumor location (red sphere with a 14-mm diameter, equal to the average diameter of the microlobulated tumor phantom), in the fibroglandular tissue of the breast (the outer surface of both the fibroglandular mold and the outer breast surface mold are depicted with orange color)—Top View; (c) Schematic definition of the tumor location in the fibroglandular tissue of the breast—Side View.

Figure 16

Breast rotational position #1—image formed at a single vertical position of the sensor network, in front of the tumor; top XY view of the full imaging domain (the antennae center positions are depicted with purple dots): (a) Test Date 1; (b) Test Date 2.

Figure 17

Breast rotational position #1—image in the interior of the breast only, formed at a single vertical position of the sensor network, in front of the tumor: (a) Test Date 1, top XY view, superposition of the fibroglandular and outer breast surface molds (blue color), red sphere indicating the tumor location; (b) Test Date 2, top XY view, superposition of the fibroglandular and outer breast surface molds (blue color), red sphere indicating the tumor location; (c) Test Date 1, 3D view, superposition of the outer breast surface mold (blue color), red sphere indicating the tumor location; (d) Test Date 2, 3D view, superposition of the outer breast surface mold (blue color), red sphere indicating the tumor location.

Figure 18

Breast rotational position #1: (a) Test Date 1, image quality assessment (QA) metric 1; (b) Test Date 2, image QA metric 1; (c) Test Date 1, image QA metric 2; (d) Test Date 2, image QA metric 2.

Figure 19

Experimental setup for the breast rotational position #2: (a) Photo—Top view of the breast phantom, installed on the examination table; (b) Schematic definition of the tumor location (red sphere of diameter 14 mm, equal to the average diameter of the microlobulated tumor phantom) in the fibroglandular tissue of the breast (the outer surface of both the fibroglandular mold and the outer breast surface mold are depicted with orange color)—Top View; (c) Schematic definition of the tumor location in the fibroglandular tissue of the breast—Side View.

Figure 20

Breast rotational position #2—image formed at a single vertical position of the sensor network, in front of the tumor, top XY view of the full imaging domain (the antennae center positions are depicted with purple spots): (a) Test Date 1; (b) Test Date 2.

Figure 21

Breast rotational position #2—image in the interior of the breast only, formed at a single vertical position of the sensor network, in front of the tumor: (a) Test Date 1, top XY view, superposition of the fibroglandular and outer breast surface molds (blue color), red sphere indicating the tumor location; (b) Test Date 2, top XY view, superposition of the fibroglandular and outer breast surface molds (blue color), red sphere indicating the tumor location; (c) Test Date 1, 3D view, superposition of the outer breast surface mold (blue color), red sphere indicating the tumor location; (d) Test Date 2, 3D view, superposition of the outer breast surface mold (blue color), red sphere indicating the tumor location.

Figure 22

Breast rotational position #2: (a) Test Date 1, image QA metric 1; (b) Test Date 2, image QA metric 1; (c) Test Date 1, image QA metric 2; (d) Test Date 2, image QA metric 2.

Figure 23

Imaging focusing analysis on 3D multi-H data: (a) Rotational position #1—Test Date 1; (b) Rotational position #1—Test Date 2; (c) Rotational position #2—Test Date 1; (d) Rotational position #2—Test Date 2.

Figure 24

Breast rotational position #1—Test Date 1—image formed using six vertical positions of the sensor network in the vicinity of the tumor—image in the interior of the breast superimposed with the image of the full imaging domain: (a) top XY view; (b) side XZ view.

Figure 25

Breast rotational position #1—Test Date 2—image formed using six vertical positions of the sensor network, in the vicinity of the tumor—image in the interior of the breast superimposed with the image of the full imaging domain: (a) top XY view; (b) side XZ view.

Figure 26

Breast rotational position #2—Test Date 1—image formed using six vertical positions of the sensor network, in the vicinity of the tumor—image in the interior of the breast superimposed with the image of the full imaging domain: (a) top XY view; (b) side XZ view.

Figure 27

Breast rotational position #2—Test Date 1—image formed using six vertical positions of the sensor network, in the vicinity of the tumor—image in the interior of the breast superimposed with the image of the full imaging domain: (a) top XY view; (b) side XZ view.