Dedicated 3D photoacoustic breast imaging

Abstract

Purpose: To report the design and imaging methodology of a photoacoustic scanner dedicated to imaging hemoglobin distribution throughout a human breast.

Methods: The authors developed a dedicated breast photoacoustic mammography (PAM) system using a spherical detector aperture based on our previous photoacoustic tomography scanner. The system uses 512 detectors with rectilinear scanning. The scan shape is a spiral pattern whose radius varies from 24 to 96 mm, thereby allowing a field of view that accommodates a wide range of breast sizes. The authors measured the contrast-to-noise ratio (CNR) using a target comprised of 1-mm dots printed on clear plastic. Each dot absorption coefficient was approximately the same as a 1-mm thickness of whole blood at 756 nm, the output wavelength of the Alexandrite laser used by this imaging system. The target was immersed in varying depths of an 8% solution of stock Liposyn II-20%, which mimics the attenuation of breast tissue (1.1 cm(-1)). The spatial resolution was measured using a 6 μm-diameter carbon fiber embedded in agar. The breasts of four healthy female volunteers, spanning a range of breast size from a brassiere C cup to a DD cup, were imaged using a 96-mm spiral protocol.

Results: The CNR target was clearly visualized to a depth of 53 mm. Spatial resolution, which was estimated from the full width at half-maximum of a profile across the PAM image of a carbon fiber, was 0.42 mm. In the four human volunteers, the vasculature was well visualized throughout the breast tissue, including to the chest wall.

Conclusions: CNR, lateral field-of-view and penetration depth of our dedicated PAM scanning system is sufficient to image breasts as large as 1335 mL, which should accommodate up to 90% of the women in the United States.

Figure 1

Photograph of PAM scanner showing the exam table (T) and the breast positioning cup (C), below which is located the hemispherical detector array.

Figure 2

Drawing showing the hemispherical array (A) mounted on a two-axis translational stage (XY). The hemispherical array and an extension (E) are filled with degassed RO water. Laser light is fed from the bottom of the array via an articulating arm (not shown) through a negative lens that diverges the laser light (L) to a diameter of ∼60 mm at the breast surface.

Figure 3

Cut-away view of imaging geometry showing the relationships among the tabletop, breast positioner (spherical cup of .020″ thick PETG), and the detector array.

Figure 4

Two of the spiral scan patterns used in this report: (a) our smallest spiral (120 locations, radius = 24 mm), and (b) a larger spiral (480 locations, radius = 48 mm). Note that the spacing between discrete locations is approximately the same for both spirals.

Figure 5

A contrast phantom was fabricated by printing 1 mm red dots on a transparent film and attaching it to a 1-cm-thick disk of PVCp. The dots are spaced 5 mm apart radially and enclosed by an 80 mm diameter circle. The absorption of the red dots was equivalent to the absorption of 1 mm blood thickness at 756 nm.

Figure 6

Imaging geometry for measuring contrast and noise using the phantom pictured in Fig. 4 as a function of depth of breast-tissue-mimicking liquid (8% LiposynII-20%).

Figure 7

PAM images (24 mm spiral scan) of the contrast phantom through 10, 30, and 53 mm of 8% LiposynII-20%. The contrasts of the three images have been normalized to one another.

Figure 8

(a) Photograph of carbon fiber (6 μm diameter) placed in agar mold. (b) MIP of PAM image of carbon-fiber phantom. In making the photoacoustic image, we placed the dot phantom (Figure 5) atop the agar to keep it from floating, so it is projected on the photoacoustic image of the graphite filament phantom. [Image of 1 mm dot phantom (Figure 5) provides scale.]

Figure 9

(a) Contrast as a function of depth in Liposyn II solution. The data, when fit to an exponential, revealed an effective attenuation coefficient of 1.1 cm⁻¹. (b) Contrast-to-noise ratio (CNR) was calculated from images such as those displayed in Fig. 7. The CNR was approximately constant (∼200) for depths < 30 mm and was limited by streak noise, which is proportional to contrast. For greater depths, the system electronic noise [∼1 unit on Fig. 9(a)] began to degrade the CNR.

Figure 10

PAM image of contrast phantom displaced 80 mm laterally from the center of the breast cup. To capture this image, the radius of the spiral scan was increased to 96 mm (1920 locations). The phantom was visualized clearly at the periphery of the image volume.

Figure 11

Maximum intensity projections in the medial-lateral (ML) projection of bilateral PAM exams of four healthy volunteers with known mammographic breast density and brassiere cup size (back-to-back images: left breast on right, right breast on left as is normally presented clinically for x-ray mammograms): (1) heterogeneously dense, D cup; (2) scattered fibroglandular densities, DD cup; (3) scattered fibroglandular densities, C cup; and (4) scattered fibroglandular densities, DD cup. Please note that the front surface of the breasts, being pliable, assume the smooth shape of the breast cup.

Figure 12

Maximum intensity projections in the coronal projection of bilateral PAM exams of the four healthy volunteers.