Photoacoustic and ultrasound system for breast cancer imaging

Abstract

An automated dual‑modality breast imaging platform captures co‑registered photoacoustic and ultrasound volumes while a patient stands and gently contacts an imaging window. In some embodiments, pulsed near‑infrared illumination is delivered through an optical‑acoustic combiner aligned with an ultrasound transducer array. A positioning subsystem translates the imaging head along a scan path that covers a large frontal field of view, with opposed longitudinal passes separated by a lateral offset and an overlap region. At successive scan positions, a controller triggers optical pulses, acquires light‑generated acoustic signals, and reconstructs photoacoustic cross‑section images, then transmits ultrasound pulses and reconstructs corresponding ultrasound images. Cross‑section images are assembled into three‑dimensional representations and output as overlays or fused displays. Neural‑network processing may denoise photoacoustic channel data before reconstruction and enhance volumetric vasculature while reducing skin and background artifacts.

Description

Attorney Docket No.: 011520.02022PHOTOACOUSTIC AND ULTRASOUND SYSTEM FOR BREAST CANCER IMAGINGCross-Reference to Related Applications

[0001] This application claims the benefit of and priority to U.S. Provisional Patent Application No. 63 / 753,930, filed February 4, 2025, now pending, the entire disclosure of which is incorporated herein by reference.Statement Regarding Federally Sponsored Research

[0002] This invention was made with government support under grant number EB029596 awarded by National Institutes of Health. The government has certain rights in the invention.Field of the Disclosure

[0003] The present disclosure relates generally to patient imaging, and more specifically to a combined ultrasound and photoacoustic tomography breast imaging system.Background of the Disclosure

[0004] Breast cancer is a major global health concern, with 2.26 million new cases diagnosed in 2020. It is a leading cause of death among women, with approximately 684,996 women losing their lives to the disease around the world in 2020. As the evidence shows, early detection has significantly reduced mortality rates, with the female breast cancer death rate falling by 41% in the US since 1989. While current diagnostic methods, such as X-ray mammography, ultrasound imaging, and magnetic resonance imaging (MRI). have vastly improved over the past decades, they still have limitations that compromise accuracy or availability. For example. X-ray mammograms involve painful breast compression and have lower sensitivity in dense breast tissue. While digital tomosynthesis and contrast-enhanced mammography improve detection sensitivity for patients with high breast density, they are either associated with a high radiation dosage or need contrast injection. MRI, on the other hand, is expensive, time-consuming, and not universally available. Finally, handheld ultrasound has high false-positive rates and is subject to operator variability. To address these limitations, there is a need for more accurate breast cancer screening methods, particularly for radiographically dense breasts.Attorney Docket No.: 011520.02022Brief Summary of the Disclosure

[0005] An automated photoacoustic and ultrasound breast tomography system that images the patient in the standing pose is disclosed. An embodiment of the system, named OneTouch-PAT, utilizes linear transducer arrays with optical-acoustic combiners for effective dual-modal imaging. During scanning, subjects only need to gently attach their breasts to the imaging window, and co-registered three-dimensional ultrasonic and photoacoustic images of the breast can be obtained within one minute. Our system has a large field of view of 17 cm by 15 cm and achieves an imaging depth of 3.5 cm with sub-millimeter resolution. Three-dimensional deep-learning networks were also developed to further improve the image quality by improving the 3D resolution, enhancing vasculature, eliminating skin signals, and reducing noise. The performance of the example system was tested on four healthy subjects and 61 patients with breast cancer. The results indicate that the ultrasound structural information can be combined with the photoacoustic vascular information for better tissue characterization.Representative cases from different molecular subtypes have indicated different photoacoustic and ultrasound features that may be used for imaging-based cancer classification. Statistical analysis among all patients indicates that the regional photoacoustic intensity and vessel intensity are indicators of breast malignancy. These promising results show that embodiments of the presently disclosed system could significantly enhance breast cancer diagnosis and classification.Description of the Drawings

[0006] For a fuller understanding of the nature and objects of the disclosure, reference should be made to the following detailed description taken in conjunction with the accompanying drawings.

[0007] Figure 1: Top row : a, schematic drawing of the OneTouch-PAT system, b, zoom-in image of the double-reflector design for coplanar illumination and acoustic detection. After accounting for acoustic reflection, the distance from the transducer to plastic film ranges from 30 to 35 mm. The transducer focuses at 40 mm. Bottom row: c, PA image from the upw ard scan. The transducer scanned from the bottom left comer of the water tank up over 17 cm. d, PA image from the downward scan. After the upward scan, the transducer laterally moved 7 cm and then scanned down over 17cm. The purple dashed box marked the overlapped region between the two scans, e, the combined PA image from two scans. The breast images are presented in frontal view.Attorney Docket No.: 011520.02022

[0008] Figure 2: The flowchart of the breast data processing. Top row: a 2DFD U-net model was used to remove the experimental noise from the raw data. The raw data frames were then reconstructed and stacked to form a max amplitude projection (MAP) image. This method preserves the original signal intensity for quantitative analysis. Bottom row: the raw data frames were reconstructed and stacked to form a 3D volume. The trained 3DFD U-net model was then applied to the volume to enhance the 3D vasculature. This method provides better visualization of blood vasculature.

[0009] Figure 3. Depth-encoded MAP images of breast vasculature from four healthy volunteers with cup sizes from A-D. The subjects also have different skin colors. Each row represents images of left and right breast from the same subject. The imaging field of view is 156 mm x 170 mm. The depth was quantified in reference to a planar surface close to the plastic membrane. Vessels near 3 cm deep are seen in red colors and marked by white arrows.

[0010] Figure 4: OneTouch-PAT imaging results of two LUMA cases, a&f represent the clinical US images, b&g represent the one-touch US (gray) images, c&h represent the PA (color) images overlaid on top of grayscale US images, d&i represent the 3D projection PA overlaid on top of US images (frontal view of the breast), e&j represent the depth-encoded 3D projection PA images (frontal view of the breast). For the overlay images, the PA intensity was linearly mapped to a transparency matrix, with the strongest PA signals rendered completely opaque and the w eakest signals fully transparent.

[0011] Figure 5: OneTouch-PAT imaging results of tw o LUMB cases, a&f represent the clinical ultrasound image, b&g represent the one-touch ultrasound (gray) images, c, h represent the photoacoustic (color) images overlaid on top of US images, d&i represent the 3D projection PA overlaid on top of US images (frontal view of the breast), e&j represent the depth-encoded 3D projection PA images (frontal view' of the breast).

[0012] Figure 6: OneTouch-PAT imaging result of a TNBC case, a represents the clinical ultrasound image, b represents the one-touch ultrasound (gray) image, c represents the photoacoustic (color) image overlaid on top of the US image, d represents the 3D projection PA overlaid on top of the US image (frontal view of the breast), e represents the depth-encoded 3D projection PA image (frontal view of the breast).Attorney Docket No.: 011520.02022

[0013] Figure 7: Statistical analysis of PA signal intensity in malignant and contralateral healthy breast, a. Violin plot of relative (ratio of malignant and healthy) PA intensity in 61 patients on the regional intensity, background. STD background, and the AVE vessels. (***: p < 0.001, **: p < 0.01, *: p < 0.05 for upper-tailed t-test). The white dots represent the mean value, b. The relative PA intensity comparison between heterogeneously dense breast (n = 23) and scatter fibroglandular breast (n = 35). c. The relative PA intensity comparison between fair color (n = 49) and dark color breasts (n = 8).

[0014] Figure 8 depicts a high-level diagram of a system according to another embodiment of the present disclosure.

[0015] Figure 9: Statistical analysis of PA signal intensity in malignant and contralateral healthy breast, a. Comparison of relative PA intensity between malignant and healthy breast tissue, averaged across 61 patients. The tumor-bearing breast result has been normalized by the contralateral healthy breast, b. Distribution of relative PA intensity in 61 patients on the regional intensity, background, STD background, and the AVE vessels. (***: p<0.0005, **: p<0.005, *: p< 0.05). c. The relative PA intensity comparison between heterogeneously dense breast (n = 23) and scatter fibroglandular breast (n = 36).

[0016] Figure 10: Statistical analysis of vessel branching points, a. Relative vessel branching point density ratio (the malignant side density has been normalized by the contralateral heathy side) among different tumor subtypes (*: p< 0.05). b. Relative vessel branching point density ratio among different Nottingham Prognostic Index (NPI) scores (**p<0.005, *: p< 0.05).

[0017] Figure 11 : A flowchart showing another embodiment of breast data processing. Top row: use the trained FD U-net model to remove the EMI noise from the raw data and reconstruct the breast image. Middle row: reconstruct the breast image and use the trained 3DFD U-net model to enhance the 3D vasculature. Bottom row: the conventional reconstruction method to use wavelet filter (WF) to eliminate the EMI noise and reconstruct the breast image. Compared with WF image, Denoised image demonstrates clear and smooth vessels and 3DFD image demonstrates resolution-refined and background clean vessels.

[0018] Figure 12: Breast reconstructed images in wavelet filter and 3DFD U-net (in an experimental embodiment). Top row: a. fair skin reconstructed breast images in WF. b. fair skinAttorney Docket No.: 011520.02022color reconstructed breast images in 3DFD. Bottom row: c. dark skin reconstructed breast images in WF. d. dark skin reconstructed breast images in 3DFD. The deep-leaming processed image shows cleaner vasculature structure with little skin signals. All images are depth encoded with color and presented in frontal view.

[0019] Figure 13: Breast reconstructed images from healthy subjects of different cup sizes. The yellow box (left group of 8 images) marks the left breast reconstructed images. a,e,i,m are the breast imaged reconstructed in WF at the breast cup size of A-D, respectively. b,fj,n are the breast imaged reconstructed in 3DFD at the breast cup size of A-D, respectively. The green box (right group of 8 images) marks the right breast reconstructed images. c,g,k,o are the breast imaged reconstructed in WF at the breast cup size of A-D, respectively. d,h,l,p are the breast imaged reconstructed in 3DFD at the breast cup size of A-D, respectively. All images are depth encoded with color and presented in frontal view.

[0020] Figure 14: OneTouch-PAT imaging results of two LUMA cases in an experimental embodiment. The left column represents the clinical ultrasound image, while the middle column represents the one-touch ultrasound (gray) and photoacoustic (color) images. The right column represents the 3D projection image (frontal view of the breast).

[0021] Figure 15: OneTouch-PAT imaging results of two LUMB cases in an experimental embodiment. The left column represents the clinical ultrasound image, while the middle column represents the one-touch ultrasound (gray) and photoacoustic (color) images. The right column represents the 3D projection image (frontal view of the breast).

[0022] Figure 16: OneTouch-PAT imaging result of a TNBC case in an experimental embodiment. The left image represents the clinical ultrasound image, while the middle images represents the one-touch ultrasound (gray) and photoacoustic (color) images. The right image represents the 3D projection image (frontal view of the breast).Detailed Description of the Disclosure

[0023] The present disclosure relates generally to patient imaging and, more specifically, to a dual-modality breast imaging system that uses photoacoustic imaging and ultrasound imaging to obtain co-registered information about breast tissue while a patient is in a standing pose. In some embodiments, the disclosed system supports a workflow7in which a subject standsAttorney Docket No.: 011520.02022at an imaging window and gently places a breast against the imaging window so that data acquisition can be performed with minimal compression while maintaining consistent coupling conditions.

[0024] With reference to Figure 8, in some embodiments, the system 10 includes a pulsed light source 20 configured to irradiate a region of interest in breast tissue 90. The pulsed light source 20 may be configured to emit light in a near-infrared wavelength range, since near-infrared illumination can provide useful penetration in tissue while enabling photoacoustic contrast associated with optical absorbers such as hemoglobin. In some embodiments, the pulsed light source has a wavelength in a range from about 680 nm to about 1100 nm, or in a range from about 680 nm to about 1300 nm, and may include example wavelengths such as about 750 nm. about 800 nm, about 850 nm, or about 1064 nm. In some embodiments, the pulsed light source is a laser such as a Nd:YAG laser, a Ti: Sapphire laser, one or more laser diodes, or other pulsed sources suitable for photoacoustic excitation. In some embodiments, the pulsed light source produces pulses having a pulse width in a range from about 1 ns to about 200 ns, and in some embodiments a pulse width in a range from about 5 ns to about 10 ns. In some embodiments, the pulsed light source is or includes a fiber optic (e.g., fiber optic bundle), for example, receiving light from a laser, laser diode, etc.

[0025] In some embodiments, the pulsed light source is configured to emit optical pulses at a single wavelength, while in other embodiments the pulsed light source is configured to emit optical pulses at a plurality of wavelengths to enable multi-spectral photoacoustic imaging. The plurality of wavelengths may include at least two wavelengths selected from a range such as about 680 nm to about 1300 nm, including wavelengths in the near-infrared region. In some embodiments, a first wavelength and a second wavelength are selected to provide different absorption contrast for oxygenated hemoglobin and deoxygenated hemoglobin, enabling estimation of an oxygenation parameter in tissue. The wavelengths may be emitted sequentially, interleaved, or in repeated cycles at a given scan position, and the controller may record wavelength identifiers for each photoacoustic acquisition. In some embodiments, the system uses a wavelength near 1064 nm for deeper penetration and uses one or more additional wavelengths to provide oxygenation sensitivity.

[0026] In some embodiments, the system 10 includes an ultrasound transducer 30 configured to transmit ultrasound pulses into, and receive acoustic signals from, the region ofAttorney Docket No.: 011520.02022interest. The ultrasound transducer 30 may receive light-generated acoustic signals produced by photoacoustic effect in response to irradiation by the pulsed light source and may also transmit ultrasound pulses and receive ultrasound echo signals resulting from reflections of the transmitted ultrasound pulses. In some embodiments, the ultrasound transducer is an ultrasound transducer array, such as a linear array transducer. In some embodiments, the ultrasound transducer is implemented as an array transducer, such as a linear array that may have, for example, about 128 elements, and the array may be used for both photoacoustic receive operations and ultrasound pulse-echo operations. In some embodiments, the array may be a linear, curved, or two-dimensional array having at least 16 elements, such as 32-512 elements (e.g., 64, 96, 128, 192, 256, or 512). The element pitch and center frequency may be selected based on target depth and resolution requirements.

[0027] In some embodiments, the system 10 includes one or more optical components 24 configured to align an optical path of the pulsed light source 20 and an ultrasound path of the ultrasound transducer 30 such that the optical path and the ultrasound path are at least one of co-planar or co-axial. In a representative embodiment, the one or more optical components form an optical-acoustic combiner that aligns illumination and acoustic detection in a compact geometry. In some embodiments, the optical-acoustic combiner includes a first reflector configured to reflect pulsed optical energy toward a shared propagation region and a second reflector configured to transmit at least a portion of the pulsed optical energy toward tissue while reflecting acoustic signals toward the ultrasound transducer. In some embodiments, the reflectors are dichroic mirrors arranged at an angle such as about 45 degrees (with respect to surface of the ultrasound transducer) to support coplanar illumination and acoustic detection in a compact head.

[0028] In some embodiments, the system 10 includes a vessel 50 configured to contain an acoustic coupling medium 54 and having an imaging window 52 positioned to receive the breast 90 of the standing patient. The imaging window 52 may be configured to transmit pulsed optical energy and acoustic energy. In some embodiments, the imaging window is sealed by a film configured to contact tissue and permit acoustic and optical transmission. For example, the vessel is a water tank with an imaging window sealed with a clear film having a thickness of about 0.002 inches that allows transmission of light pulses through the window to the region of interest. In some embodiments, the acoustic coupling medium includes de-ionized water or other coupling fluids suitable for acoustic transmission. In some embodiments, the vessel is used with ultrasound gel applied at least at the tissue-window interface to promote consistent coupling.Attorney Docket No.: 011520.02022

[0029] In some embodiments, the vessel contains a coupling medium selected to provide acoustic coupling while exhibiting relatively low optical attenuation at one or more near-infrared wavelengths used for photoacoustic excitation. In some embodiments, the coupling medium is mineral oil, which can provide reduced optical attenuation relative to water at wavelengths in a near-infrared region, including wavelengths near 1064 nm, thereby increasing the optical fluence delivered to tissue at the imaging window for a given source output. In other embodiments, the coupling medium includes heavy water (D2O) or another liquid selected to reduce optical absorption and / or scattering at the excitation wavelength. The coupling medium may be selected based on optical attenuation coefficient, chemical compatibility with the vessel and window film, viscosity, ease of handling, patient comfort, or more than one of these or other considerations.

[0030] In some embodiments, image reconstruction accounts for different acoustic propagation velocities in different regions along an acoustic path. For example, when the coupling medium has a speed of sound that differs from the speed of sound in tissue, reconstruction may apply a dual speed-of-sound model. In some embodiments, reconstruction uses a back-projection approach in which a first region corresponding to the coupling medium is defined based on system geometry, including a known window location and a known transducer location, and propagation times within the first region are calculated using a first speed of sound associated with the coupling medium. For a second region corresponding to tissue, propagation times are calculated using a second speed of sound associated with biological tissue. In some embodiments, the first speed of sound is in a range from about 1.3 to about 1.5 mm / ps and the second speed of sound is in a range from about 1.45 to about 1.60 mm / ps, and these values may be refined based on calibration targets or tissue-specific estimates. In some embodiments, using region-specific propagation velocities improves focus and sharpness of reconstructed vasculature relative to reconstruction based on a single speed of sound.

[0031] In some embodiments, defining the coupling-medium region includes determining a coupling-path length corresponding to a distance between an acoustic surface of the transducer and the imaging window / film (e. , after accounting for any acoustic reflection), and using the coupling-path length to determine propagation time contributions within the coupling-medium region.

[0032] In some embodiments, the system includes a support platform, such as a height- adjustable support platform configured to position the standing patient relative to theAttorney Docket No.: 011520.02022imaging window. For example, a lift table, support structure, handhold, or similar fixture may be used so that the patient can maintain a comfortable standing pose and stable breast placement at the imaging window during data acquisition. In some embodiments, a raised bar or other support element is provided to assist the patient in maintaining posture and reducing motion during scanning.

[0033] In some embodiments, the system includes a positioning subsystem 40 configured to translate at least one of the ultrasound transducer or the one or more optical components along a scan trajectory to implement a plurality of scan positions. In some embodiments, a positioning subsystem includes one or more scanning stages that translate an imaging head and / or one or more optical components along one or more axes to steer or reposition an optical path for scanning a region of interest. In some embodiments, the positioning subsystem includes one or more translation stages, and in some embodiments includes first and second translation stages that enable a scan trajectory with both longitudinal and lateral movement.

[0034] In some embodiments, the positioning subsystem supports a scan trajectory that includes a first longitudinal scan along the breast, a lateral translation, and a second longitudinal scan along the breast in a direction opposite the first longitudinal scan. In some embodiments, a positioning subsystem executes a round-trip scan in which an imaging head performs a first longitudinal pass across the breast, then undergoes a lateral offset, and then performs a second longitudinal pass in an opposite direction, with data from the two passes having an overlap region that is used for combination into a composite output; by way of example, each longitudinal pass may span about 10-25 cm (e.g., about 17 cm), the lateral offset may be about 3-12 cm (e.g, about 7 cm), and the overlap region may be selected to cover at least 0.5-10 cm of the longitudinal extent (or about 5-50% of the lateral coverage), depending on desired field of view and sampling density. The scanning results may be combined to form final imaging data presented in a frontal view of the breast. In some embodiments, the field of view is on the order of 17 cm by 15 cm, and in some embodiments the system provides such coverage in approximately one minute, with a representative round-trip scan described as taking approximately 52 seconds.

[0035] In some embodiments, the system includes a controller in electronic communication with the pulsed light source and the ultrasound transducer. The controller may be configured to coordinate photoacoustic acquisition and ultrasound acquisition during scanning.Attorney Docket No.: 011520.02022In some embodiments, a controller is configured to coordinate dual-modality imaging by triggering one or more optical excitation pulses, acquiring corresponding light-generated acoustic signals for reconstruction of a photoacoustic image, and by transmitting one or more ultrasound transmit pulses, acquiring corresponding echo signals, and reconstructing an ultrasound image, with the photoacoustic and ultrasound images associated with a common region of interest and optionally acquired at a plurality of scan positions. In some embodiments, the controller coordinates acquisition with motion control such that data are acquired at a plurality of scan positions along a scan trajectory.

[0036] In some embodiments, the controller is configured to acquire photoacoustic data and ultrasound data in a coordinated acquisition sequence at each scan position corresponding to a cross-section of the region of interest. In some embodiments, for a given scan position, the controller triggers one or more light pulses from the pulsed light source to irradiate an illumination region at that scan position, receives light-generated acoustic signals corresponding to the one or more light pulses, and reconstructs a photoacoustic image of the cross-section at that scan position. After receiving the light-generated acoustic signals, the controller transmits one or more ultrasound pulse signals associated with that cross-section, receives ultrasound echo signals corresponding to those ultrasound pulse signals, and reconstructs an ultrasound image of that cross-section. In such embodiments, these operations collectively define a paired acquisition for the cross-section at the scan position, and the controller repeats the paired acquisition at different scan positions to scan the region of interest.

[0037] In some embodiments, the controller generates a multi-wavelength photoacoustic dataset by reconstructing a photoacoustic image volume for each of the plurality of wavelengths. The controller then estimates an oxygenation parameter based on relative photoacoustic signal amplitudes across wavelengths. For example, the controller may estimate concentrations of oxygenated and deoxygenated hemoglobin using spectral unmixing, least-squares fitting, or another model-based approach that relates measured photoacoustic amplitude to optical absorption at the selected wavelengths. The controller may output an oxygenation map (e. ., a blood oxygen saturation representation) co-registered with an ultrasound image volume or with a structural ultrasound frame.

[0038] Some embodiments refer to the coordinated acquisition sequence as an interleaved acquisition sequence. In various embodiments, interleaving is carried out byAttorney Docket No.: 011520.02022alternating photoacoustic acquisition and ultrasound acquisition at each scan position, while in other embodiments interleaving is carried out by performing photoacoustic acquisition and ultrasound acquisition sequentially at each scan position while the breast remains positioned against the imaging window. In some embodiments, photoacoustic acquisition and ultrasound acquisition are performed in separate time intervals while the patient maintains a stable pose at the imaging window, and co-registration is achieved as described below. This disclosure supports these variants by describing coordinated control of light pulses, photoacoustic reception, ultrasound pulsing, and echo reception to obtain photoacoustic and ultrasound images associated with the region of interest.

[0039] In some embodiments, illumination regions corresponding to successive light pulses overlap. Overlap may occur due to the spatial extent of the illumination delivered through the light delivery optics and the spacing between successive scan positions or successive pulse events. For example, when a fiber bundle produces a line output and the scan step size or scan speed results in successive pulse events occurring with less spatial separation than a width of the illumination region, adjacent illumination regions overlap in the tissue. In some embodiments, overlap is selected to support continuous coverage of the region of interest and to reduce the likelihood of unirradiated gaps between adjacent acquisition points. In some embodiments, overlap is expressed as anon-zero intersection area between adjacent illumination footprints at the tissue plane, and the overlap may be selected based on the optical output profile of the light delivery, the standoff geometry, and the scanning trajectory.

[0040] In some embodiments, the controller reconstructs, from cross-sectional frames acquired at the plurality of scan positions, a three-dimensional photoacoustic representation and a three-dimensional ultrasound representation. For example, photoacoustic frames acquired at successive scan positions may be stacked or otherwise assembled into a three-dimensional photoacoustic image volume, and ultrasound frames acquired at successive scan positions may be stacked or otherwise assembled into a three-dimensional ultrasound image volume. In some embodiments, a three-dimensional volumetric ultrasound image may be constructed by stacking cross-sectional B-mode images along the scanning direction. Scan passes may be combined to form composite breast imaging data.

[0041] In some embodiments, the controller generates co-registered output that spatially aligns the three-dimensional photoacoustic representation with the three-dimensional ultrasoundAttorney Docket No.: 011520.02022representation. In some embodiments, co-registration is supported by using a shared ultrasound transducer array for both modalities so that photoacoustic reception and ultrasound echo reception are inherently tied to the same transducer coordinate system. In some embodiments, co-registration is verified or refined using anatomical landmarks or geometric features observable in both modalities. Landmarks such as breast outline and nipple location, may be recorded, and such landmarks may be used to ensure accurate alignment between photoacoustic and ultrasound images.

[0042] In some embodiments, the controller generates a display output that overlays photoacoustic intensity on a grayscale ultrasound image. For example, photoacoustic intensity values may be mapped to a color scale and rendered as an overlay , while ultrasound image data are rendered in grayscale to depict morphological structure. In some embodiments, output images include an overlay in which photoacoustic vascular information is rendered over an ultrasound image to provide combined functional and structural visualization, and the system further generates three-dimensional renderings and one or more projection images that depict vasculature in a frontal view of the breast; for example, the projection images may include maximum-amplitude or maximum-intensity projections and may be depth-encoded, and the rendered field of view may be on the order of about 10-25 cm by about 10-25 cm (e.g., about 15-17 cm by about 15-17 cm), with an imaging depth on the order of about 2-5 cm depending on wavelength, transducer bandwidth, and reconstruction settings.

[0043] In some embodiments, the ultrasound imaging portion uses a multi-angle wide-beam imaging strategy to obtain high-quality' ultrasound images. For example, ultrasound waves may be transmitted at multiple steering angles such as about -15°, 0°, and +15°. with multiple transmissions per angle across a lateral length of the ultrasound transducer array. The received ultrasound echo data may be spatially compounded and averaged to form a single ultrasound frame for a cross-section. In some embodiments, ultrasound imaging is performed using a multi-angle wide-beam transmit sequence that transmits ultrasound energy at a plurality of steering angles and receives corresponding echo data, and the echo data are spatially compounded and optionally averaged to form a single ultrasound frame and to reduce side-lobe artifacts while maintaining acquisition speed. In some embodiments, the plurality' of steering angles includes at least three angles, such as angles within about -30° to +30°, for example 15°.0°, and +15°, and the transmit sequence includes multiple transmissions per steering angle, suchAttorney Docket No.: 011520.02022as at least 2, 5-100, or 10-60 transmissions per angle (e.g., about 42 transmissions per angle) distributed across a lateral extent of an ultrasound transducer array.

[0044] In some embodiments, photoacoustic data processing includes operations to improve signal-to-noise ratio and visualization of vasculature. Photoacoustic signals may include contributions from thermal noise and electromagnetic interference associated with laser firing, and the raw channel data may be processed prior to reconstruction to reduce noise while maintaining intensity information useful for quantitative analysis. In some embodiments, photoacoustic processing uses a first processing path in which raw photoacoustic channel data are denoised prior to image reconstruction by applying a first neural network to remove noise contributions (including thermal noise and / or electromagnetic interference), and reconstructed photoacoustic frames are then assembled into one or more projection images, such as a maximum-amplitude projection image, with the denoising configured to preserve original signal intensity for quantitative analysis. In some embodiments, a second processing path reconstructs photoacoustic frames into a three-dimensional volume and applies a second neural network that operates on volumetric data to enhance vasculature representation, improve vessel continuity, and reduce artifacts including skin-related signal components and background noise, thereby producing an output suitable for visualization and morphometric analysis.

[0045] In some embodiments, the first neural network is a two-dimensional network (for example, a U-net-type architecture) applied to channel data on a frame-by-frame basis, and the second neural network is a three-dimensional network applied to volumetric data. In some embodiments, the first neural network and the second neural network include 2-10 encoding / decoding stages, 8-256 feature channels per stage, and convolution kernels having sizes such as 3x3 (2D) and 3x3x3 (3D), with training performed using simulated data augmented with experimentally measured system noise. In some embodiments, the second neural network is trained using volumetric inputs that already incorporate representative system noise, and an output of the first neural network is not used as an input to the second neural network.

[0046] In some embodiments, the two-dimensional neural network receives photoacoustic raw channel data as input and outputs noise-reduced channel data that are then used for reconstructing photoacoustic image frames. In some embodiments, thethree-dimensional neural network receives a reconstructed volumetric representation and outputs an enhanced volumetric representation in which vasculature is more continuous and backgroundAttorney Docket No.: 011520.02022artifacts are reduced, including reductions in skin-related signal components. In some embodiments, the three-dimensional neural network training incorporates representative system noise, and the output of the two-dimensional neural network is not used as input to the three-dimensional neural network.

[0047] In some embodiments, the controller generates a maximum-amplitude projection image from photoacoustic image frames, and the maximum-amplitude projection is used for visualization and for quantitative analysis in which signal intensity is maintained. In some embodiments, projection images are depth-encoded for visualization, and in some embodiments projection images are used as part of morphometric measurements such as vessel branching point density or other measures derived from vascular structure.

[0048] In some embodiments, the scan trajectory includes an overlap region between scan passes, and the controller combines photoacoustic receive data from the first longitudinal scan and the second longitudinal scan to form a composite three-dimensional photoacoustic representation of the breast. In some embodiments, the controller similarly forms athree-dimensional ultrasound representation from ultrasound echo data acquired along the scan trajectory', and co-registered output is generated that spatially aligns the compositethree-dimensional photoacoustic representation with the three-dimensional ultrasound representation.

[0049] In some embodiments, acquisition timing and sequencing are selected based on the signal chain configuration. For example, some embodiments acquire photoacoustic signals and ultrasound echo signals in a coordinated sequence at each scan position, while other embodiments acquire photoacoustic signals for a scan and acquire ultrasound signals for a scan in separate time intervals while the patient remains positioned at the imaging window. In embodiments that use separate acquisitions, the controller may record or identify landmarks in both data sets and align breast outlines and nipple positions to produce co-registered output. Using landmarks for alignment, in many cases, no adjustments were needed to align the photoacoustic and ultrasound images.

[0050] In some embodiments, the controller and associated computing hardware perform the reconstruction and data processing operations described herein using one or more processors and memory storing program instructions. The program instructions may cause the one or more processors to receive photoacoustic and ultrasound data, apply one or more neural networks,Attorney Docket No.: 011520.02022perform reconstruction operations, form volumetric representations, generate projection images, and generate co-registered output, including overlay displays. In some embodiments, the program instructions are stored in a non-transitory computer-readable medium and executed by the one or more processors to carry out one or more methods described herein, including methods that denoise photoacoustic raw channel data prior to reconstruction and enhance volumetric vasculature using a three-dimensional neural network.

[0051] The present disclosure also contemplates variations in system geometry and scan control that remain consistent with the claim language. In some embodiments, the ultrasound transducer array is positioned behind the imaging window within the coupling medium, while the optical-acoustic combiner aligns illumination and acoustic detection within a compact head. In some embodiments, the scan trajectory follows a U-shaped coverage path to provide the field of view described herein, and in other embodiments a larger transducer array reduces the number of scan passes used to cover the breast. A larger array may enable a single scan to obtain a large field of view.

[0052] Throughout this description, terms such as '‘configured to’’ and ‘'may” indicate that embodiments may include additional components or operations beyond those explicitly described, while still producing the outputs described herein. Terms such as “about” and “approximately,” when used, refer to engineering tolerances and variations arising from patient anatomy, positioning, device calibration, and signal processing, and may be interpreted in a manner consistent with measurement uncertainty and ordinary practice in medical imaging system design. In this disclosure, the term “optical component” (and “one or more optical components”) refers to one or more physical, structural optical elements or assemblies that shape, direct, transmit, distribute, filter, or otherwise guide optical energy from a light source toward a region of interest, including configurations that spatially relate the optical path to an acoustic transmit / receive path of an ultrasound transducer. Optical components may be implemented as discrete parts or as an integrated optical subassembly and may include, by way of example and without limitation, one or more reflectors (including mirrors and dichroic mirrors), beam splitters, prisms, wedges, light guides, light pipes, optical waveguides, optical windows, diffusers, apertures, baffles, lenses (including cylindrical lenses and GRIN lenses), collimators, fiber-optic components (including one or more fiber bundles, fiber legs, fiber terminations, and / or fiber output optics), and / or optical filters (including wavelength-selective filters). The optical components may be mounted using brackets, fixtures, kinematic mounts, orAttorney Docket No.: 011520.02022other supports to maintain a desired alignment during scanning, and may be positioned in air, within the coupling medium, or behind an imaging window or film depending on the implementation.

[0053] In some embodiments, the one or more optical components form anoptical-acoustic combiner configured to align illumination and acoustic detection in a compact geometry. For example, the optical-acoustic combiner may include a first reflector that reflects pulsed optical energy toward a shared propagation region and a second reflector that transmits at least a portion of the pulsed optical energy toward tissue while reflecting acoustic signals toward the ultrasound transducer. In other embodiments, alignment of the optical path and the ultrasound path may be achieved by other techniques, such as, for example, delivering light through one or more fiber bundles directed to illuminate from a side, from a perimeter, through a window, through a light guide, or through an optically transmissive portion of an acoustic structure, including arrangements that yield coaxial, coplanar, overlapping, or otherwise geometrically related illumination and acoustic paths. The optical components may be configured to provide line illumination, spot illumination, or area illumination, may include multiple output ports, and may be arranged to deliver illumination at one or more angles relative to the ultrasound transducer. The optical components may be selected and arranged to provide desired illumination distribution at the imaging window and / or within tissue, and may include components that compensate for refraction or reflection at one or more interfaces.

[0054] In an aspect, the present disclosure may be embodied as a system for breast imaging. The system includes a pulsed light source configured to irradiate a region of interest in breast tissue. The pulsed light source may be configured to transmit light having a wavelength within the near infrared (NIR) range. For example, the pulsed light source may have a wavelength of between 680 nm and 1100 nm, inclusive (for example, 750 nm, 800 nm, 850 nm, or 1064 nm). The pulsed light source may be a laser, such as a near infrared laser (for example, a yttrium aluminum garnet (“YAG” laser), a Ti:Sapphire laser, laser diodes, dye lasers, etc.) The pulsed light source may be configured to provide light having a pulse width from 1 ns to 100 ns, inclusive (for example, from 5 ns to 10 ns, inclusive, such as 10 ns).

[0055] The system includes an ultrasound transducer configured to transmit and receive acoustic signals from the region of interest. The ultrasound transducer may receive ultrasound signals induced by the pulsed light source (i.e.. the photoacoustic (PA) effect) and to receiveAttorney Docket No.: 011520.02022ultrasound signals resulting from reflection of signals transmitted by the ultrasound transducer (i.e., echo signals). In some embodiments, the ultrasound transducer is a linear array transducer (for example, a 128-element linear array).

[0056] The system may include one or more optical components configured to align an optical path of the pulsed light source and an ultrasound path of the ultrasound transducer so as to be co-planar. For example, with reference to Figure IB, the one or more optical components may include a first mirror configured to reflect light from the pulsed light source to a signal path, and a second mirror configured to reflect ultrasound signals to the ultrasound transducer from the signal path. The second mirror may be configured to transmit at least a portion of light from the pulsed light source (i.e., at least partially transparent to the light of the pulsed light source).

[0057] A controller is in electronic communication with pulsed light source and the ultrasound transducer. The controller is configured to obtain a first image of the region of interest via photoacoustic techniques and to obtain a second image of the region of interest via ultrasound techniques. For example, the controller may be programmed to trigger a series of light pulse from the pulsed light source, wherein each light pulse of the series of light pulses is at a different position of the region of interest so as to scan the region of interest. The controller receives a series of first acoustic signals from the ultrasound transducer, wherein each first acoustic signal is at a different position of the region of interest and each first acoustic signal corresponds to a light pulse of the series of light pulses. In this way, the controller may reconstruct the first image of the region of interest based on the received series of first acoustic signals.

[0058] With respect to the second image, the controller may transmit a series of ultrasound pulse signals from the ultrasound transducer, wherein each ultrasound pulse signal of the series of ultrasound pulse signals is at a different position of the region of interest so as to scan the region of interest. A series of second acoustic signals is received at the controller from the ultrasound transducer. Each second acoustic signal is at a different position of the region of interest and each second acoustic signal corresponds to a ultrasound pulse signal of the series of ultrasound pulses signals. The controller reconstructs the second image of the region of interest based on the received series of second acoustic signals. In some embodiments, each ultrasound pulse signal of the series of ultrasound pulse signals has a wide-beam transmit sequence comprising ultrasound waves at three different steering angles (e.g., (-15°, 0°, and 15°), withAttorney Docket No.: 011520.02022multiple (e.g., 42) wide-beam transmissions per angle across a lateral length of the ultrasound transducer. The processor may be further configured to spatially compound and average the received second acoustic signals to form a single ultrasound frame.

[0059] The system may include a vessel (contained) configured to contain a coupling medium (e.g., de-ionized water). The vessel is configured to (acoustically) couple the breast tissue with the ultrasound transducer. The vessel may have a transparent window configured to allow transmission of light pulses from within the vessel to the sample such that the light from the pulsed light source is transmitted through the window to the region of interest. The system may include a scanning stage configured to translate the one or more optical components along one or more dimensions to move the optical path for scanning the region of interest.

[0060] In an aspect, the present disclosure may be embodied as a system for automated breast imaging of a patient, for example, a patient in a standing pose. The system includes a pulsed light source configured to irradiate a region of interest in breast tissue. The pulsed light source may be configured to transmit light having a wavelength within the near infrared (NIR) range. For example, the pulsed light source may have a wavelength of between 680 nm and 1300 nm, inclusive (for example, 750 nm, 800 nm, 850 nm, or 1064 nm). The pulsed light source may be a laser, such as a near infrared laser (for example, a yttrium aluminum garnet (“YAG” laser), a Ti: Sapphire laser, laser diodes, light-emitting diode, dye lasers, etc.) The pulsed light source may be configured to provide light having a pulse width from 1 ns to 200 ns, inclusive (for example, from 5 ns to 10 ns, inclusive, such as 10 ns).

[0061] The system includes an ultrasound transducer configured to transmit and receive acoustic signals from the region of interest. The ultrasound transducer may receive ultrasound signals induced by the pulsed light source (i.e., the photoacoustic (PA) effect) and to receive ultrasound signals resulting from reflection of signals transmitted by the ultrasound transducer (i.e., echo signals). In some embodiments, the ultrasound transducer is a linear array transducer (for example, a 128-element linear array) or a curved array transducer.

[0062] The system may include one or more optical components configured to align an optical path of the pulsed light source and an ultrasound path of the ultrasound transducer so as to be co-planar or co-axial. For example, with reference to Figure IB. the one or more optical components may include a first mirror configured to reflect light from the pulsed light source to a signal path, and a second mirror configured to reflect ultrasound signals to the ultrasoundAttorney Docket No.: 011520.02022transducer from the signal path. The second mirror may be configured to transmit at least a portion of light from the pulsed light source (i.e., at least partially transparent to the light of the pulsed light source).

[0063] A controller is in electronic communication with pulsed light source and the ultrasound transducer. The controller may be configured to conduct photoacoustic and ultrasound imaging in an interleaved manner. The controller may be configured to obtain a first image of the region of interest via photoacoustic techniques and to obtain a second image of the region of interest via ultrasound techniques. For example, the controller may be programmed to trigger a light pulse from the pulsed light source, wherein the light pulse irradiates a position of the region of interest. The controller receives a first acoustic signal from the ultrasound transducer, wherein the first acoustic signal corresponds to (i.e.. results from the light pulse) the light pulse. In this way, the controller may reconstruct a first image (photoacoustic image) of the region of interest based on the received first acoustic signal.

[0064] With respect to the second image, the controller may transmit a series of ultrasound pulse signals from the ultrasound transducer, wherein each ultrasound pulse signal of the series of ultrasound pulse signals is at a position of a cross-section of the region of interest so as to scan the entire cross-section. A series of second acoustic signals is received at the controller from the ultrasound transducer. Each second acoustic signal is at a position of the cross-section of the region of interest and each second acoustic signal corresponds to an ultrasound pulse signal of the series of ultrasound pulses signals. The controller reconstructs an image (i.e., the second image) of the region of interest based on the received series of second acoustic signals. In some embodiments, each ultrasound pulse signal of the series of ultrasound pulse signals has a wide-beam transmit sequence comprising ultrasound waves at different steering angles (e.g, (-15°, 0°, and 15°), with multiple (e.g.. 42) wide-beam transmissions per angle across a lateral length of the ultrasound transducer. The processor may be further configured to spatially compound and average the received second acoustic signals to form a single ultrasound frame. The controller may repeat the photoacoustic steps and ultrasound steps to scan each cross-section of the region of interest.

[0065] In some embodiments, the controller may be programmed to trigger a series of light pulse from the pulsed light source, wherein each light pulse of the series of light pulses is at a different position of the region of interest so as to scan the region of interest. The controllerAttorney Docket No.: 011520.02022receives a series of first acoustic signals from the ultrasound transducer, wherein each first acoustic signal is at a different position of the region of interest and each first acoustic signal corresponds to a light pulse of the series of light pulses. In this way, the controller may reconstruct the first image of the region of interest based on the received series of first acoustic signals.

[0066] With respect to the second image, the controller may transmit a series of ultrasound pulse signals from the ultrasound transducer, wherein each ultrasound pulse signal of the series of ultrasound pulse signals is at a different position of the region of interest so as to scan the region of interest. A series of second acoustic signals is received at the controller from the ultrasound transducer. Each second acoustic signal is at a different position of the region of interest and each second acoustic signal corresponds to a ultrasound pulse signal of the series of ultrasound pulses signals. The controller reconstructs the second image of the region of interest based on the received series of second acoustic signals. In some embodiments, each ultrasound pulse signal of the series of ultrasound pulse signals has a wide-beam transmit sequence comprising ultrasound waves at three different steering angles (e.g. (-15°, 0°, and 15°), with multiple (e.g.. 42) wide-beam transmissions per angle across a lateral length of the ultrasound transducer. The processor may be further configured to spatially compound and average the received second acoustic signals to form a single ultrasound frame.

[0067] The system may include a vessel (contained) configured to contain a coupling medium (e.g., de-ionized water, mineral oil, etc.) The vessel may be configured to (acoustically) couple the breast tissue with the ultrasound transducer. The vessel may have a transparent window configured to allow transmission of light pulses from within the vessel to the sample such that the light from the pulsed light source is transmitted through the window to the region of interest. The system may include a scanning stage configured to translate the one or more optical components along one or more dimensions to move the optical path for scanning the region of interest.

[0068] In another aspect, the present disclosure may be embodied as a computer-implemented method of breast imaging of a patient (for example, a patient in a standing pose). The method includes triggering a light pulse from a pulsed light source, wherein the light pulse irradiates a position of a region of interest. In some embodiments, the pulsed light source is configured to transmit light having a wavelength within the near infrared range. In someAttorney Docket No.: 011520.02022embodiments, the pulsed light source is a laser, such as a near infrared laser (for example, a yttrium aluminum garnet (“YAG” laser), a Ti:Sapphire laser, laser diodes, dye lasers, lightemitting diode, etc.) In some embodiments, the pulsed light source has a wavelength of between 680 nm and 1300 nm, inclusive (for example, 750 nm, 800 nm, 850 nm, or 1064 nm). In some embodiments, the pulsed light source is configured to provide light having a pulse width from 1 ns to 200 ns, inclusive (for example, from 5 ns to 10 ns, inclusive, such as 10 ns).

[0069] A first acoustic signal is received from an ultrasound transducer, wherein the first acoustic signal corresponds to (i.e., results from) the light pulse. In some embodiments, the ultrasound transducer is a linear array transducer or a curved array transducer (for example, a 128-element linear array).

[0070] The method includes reconstructing a photoacoustic image of the region of interest based on the received first acoustic signal.

[0071] The method includes transmitting a series of ultrasound pulse signals from the ultrasound transducer, wherein each ultrasound pulse signal of the series of ultrasound pulse signals is at a position of a cross-section of the region of interest so as to scan the entire crosssection.

[0072] A series of second acoustic signals is received from the ultrasound transducer, wherein each second acoustic signal is at a position of the cross-section of the region of interest and each second acoustic signal corresponds to an ultrasound pulse signal of the series of ultrasound pulses signals.

[0073] An image of the region of interest is reconstructed based on the received series of second acoustic signals.

[0074] In some embodiments, the method includes repeating the photoacoustic steps and ultrasound steps to scan each cross-section of the region of interest.

[0075] In some embodiments, the method includes translating, using a stage, the one or more optical components along one or more dimensions to move the optical path for scanning the region of interest.Attorney Docket No.: 011520.02022

[0076] In some embodiments, each ultrasound pulse signal of the series of ultrasound pulse signals has a wide-beam transmit sequence comprising ultrasound waves at different steering angles (e.g., (-15°, 0°, and 15°), with multiple (e.g., 42) wide-beam transmissions per angle across a lateral length of the ultrasound transducer.

[0077] In some embodiments, the method includes spatially compounding and averaging the received second acoustic signals to form a single ultrasound frame.

[0078] In some embodiments, the second image is a 3D volumetric image of the region of interest constructed by stacking cross-sectional B-model ultrasound images along a scanning direction. In some embodiments, multiple second images may be used to create a 3D volumetric image of the region of interest constructed by stacking the second images (e.g.. cross-sectional B-model ultrasound images) along a scanning direction.

[0079] In some embodiments, the method includes denoising the received first acoustic signal before reconstruction of the photoacoustic image. For example, received first acoustic signal may be denoised using a neural network.Further Discussion

[0080] Automated Photoacoustic (PA) tomography (PAT) and ultrasound (US) are promising dual-modality techniques to overcome many of these limitations.

[0081] The photoacoustic effect refers to the process of converting light energy into sound. Its discovery dates back to the late 1800s. The mechanism involves using a short-pulsed laser to provide excitation light, which is absorbed by molecules such as hemoglobin, lipid, or melanin. This absorption leads to thermoelastic expansion and the generation of acoustic waves that propagate through the tissue. Ultrasound transducer arrays detect these acoustic waves, and the received signals are used to reconstruct an image of the distribution of optical absorbers. The photoacoustic technique combines the benefits of high optical absorption contrast and high acoustic resolution, making it a valuable tool for breast cancer imaging. Most PA breast imaging systems utilize near-infrared wavelengths, which provide a good balance between blood absorption and tissue penetration. The acquired hemoglobin map can help to identify tissue malignancy. In general, breast cancer tissue contains more hemoglobin than benign abnormalities or normal tissue, likely due to tumor angiogenesis. The use of PAT to evaluateAttorney Docket No.: 011520.02022these additional tumor features could reduce the need for unnecessary' biopsies, assist in diagnosing breast cancer, and support the monitoring of therapies and drug development.

[0082] Over the past decade, using the photoacoustic principle for breast cancer imaging has become increasingly prevalent. Many of the existing systems require the patient to be in a prone position, which has the drawback of requiring larger clinical space and longer imaging preparation time. In addition, most of these systems do not have native ultrasound imaging capability. For example, the Twente PAM 2 and SBH-PACT systems provided 3D vasculature images of the breast, but both lacked ultrasound data for structural analysis of the lesion. The PAI-04 and LOUISA-3D systems provided both PA and US images. However, the two modalities were captured by two different transducers with different fields of view and spatial resolutions. There are also systems that image patients in a supine pose and offer dual-modal PA and US imaging using the same transducer. For example, the Imagio system uses hand-probe linear arrays to simultaneously acquire PA and US data. However, the handheld imaging scheme is operator-dependent and cannot provide volumetric features for comprehensive 3D tumor information.

[0083] Our group previously developed a photoacoustic dual-scan mammoscope (DSM) system that used two linear transducer arrays to scan the breast in a craniocaudal view, providing naturally co-registered volumetric PA and US images with the patient in standing position. However, this system works better in breasts with large cup sizes. Additionally, although the system delivers cranio-caudal images similar to a mammogram, many patients w ould prefer a method that does not require compression. To overcome this limitation, the present design provides a frontal view of the breast with minimal compression.

[0084] Embodiments of the present system, named OneTouch-PAT, build on the advantages of DSM, such as standing pose imaging position and co-registered PA / US images. In addition, it provides enhanced PA / US image quality, a larger field of view, and the ability to image a wide range of breast cup sizes. With the OneTouch-PAT system, patients just need to stand in front of the imaging w indow and gently contact their breasts to the w indow. In an example, the ultrasound transducer array (e.g., linear, curved, etc.), placed behind the window, will scan the breast, using, for example, a U-shaped path, providing a large field of view of 17 cm by 15 cm in approximately one minute. In some embodiments, ultrasound arrays may be sized sufficiently to obtain a desired field of view in a single pass, whereas smaller arrays mayAttorney Docket No.: 011520.02022require alternative scan paths (e.g, U-shaped path, etc.) The OneTouch system was first tested on four healthy subjects with different breast cup sizes and skin colors to optimize the data processing algorithm. Then, we further validated the system at collaborating clinics by imaging 61 patients with confirmed breast cancer. Our results indicate that the OneTouch system can accommodate different breast variations and tumor features, which can be seen in both PA and US images.

[0085] The system offers a simple and comfortable imaging process, requiring patients to gently attach their breast to the imaging window, where a U-shaped scanning transducer array captures high-resolution 3D PA / US images within approximately one minute over a 17 cm x 15 cm field of view. A clinical study was conducted involving 61 patients to demonstrate the OneTouch-PAT system’s ability to visualize the tumor features in both PA and US images and provide statistically significant differences in vascular intensity and distribution between malignant and healthy breast tissue.

[0086] Methods

[0087] The following discussion provides details of a non-limiting embodiment of the present disclosure and is intended only as an example.

[0088] System Configuration

[0089] The OneTouch-PAT includes a portable laser, a water tank with an imaging window in the front, a height-adjustable lift table, and two linear scanning stages. The imaging window is sealed with a clear film (e.g., 0.002-inch thick) that allows both acoustic and light transmission. A schematic of an embodiment of the system is displayed in Figure la. The imaging platform w as secured on an optical breadboard, which was fixed to a mobile lift table with an adjustable height. The water tank had a depth of 13 cm. a width of 25 cm, and a height of 20 cm. This design provided space for the imaging probe to move freely in the water tank. The transducer and optical fiber bundle were combined using a compact double-reflector design with tw o dichroic mirrors mounted at 45 degrees (e.g., with respect to surface of the ultrasound transducer), allowing for coplanar illumination and acoustic detection. A close-up view of the imaging cross-section is provided in Figure lb. The combined transducer and fiber bundle were placed into the w ater tank and connected to tw o motorized translation stages through T-slottedAttorney Docket No.: 011520.02022aluminum frames. The scanning was controlled by an Arduino board, which was programmed to synchronize with the laser firing.

[0090] The ultrasound transducer used for both PA and US acquisition was a custom-made 128-element linear array (Imasonic, Inc.) with 86-mm lateral length, 2.25 MHz central frequency, and >65% bandwidth. Each element had an arc-shaped aperture with 40 mm elevation focusing, 15 mm element height, and 0.67 mm pitch. The long elevation focus allows optimal acoustic performance across the scanning. The light source was a compact Nd: YAG laser (Quantel model CFR400) with 1064 nm output, 20 Hz pulse repetition rate, and 10 ns pulse width. The laser beam was fed to a fiber bundle with line output. The laser fluence on the skin was approximately 40 rnJ / 'cm2. which was within the safe limits for laser exposure specified by the American National Standards Institutes (ANSI) (100 mJ / cm2at 1064 nm). To synchronize the OneTouch-PAT system, we used the laser’s reference output to trigger both the data acquisition systems and the linear translation stages. The Verasonics Vantage data acquisition (DAQ) system was utilized for both PA and US data acquisition. The acquired raw data was reconstructed and processed in MATLAB. As the One-touch system shared the same transducer and light deliver}' scheme of DSM, it has the same lateral resolution of 1 mm and elevation resolution of 1 to 2 mm

[0021] , which can be further improved by the methods mentioned below.

[0091] Imaging Procedure

[0092] Before imaging, the patient stood upright and was instructed to lean forward to gently attach one breast to the imaging window. Both breast and imaging windows will be preapplied with green ultrasound gel (Parker Laboratories, Inc.) to ensure optimal coupling. To ensure patient comfort, the water tank may be filled with a warm coupling medium — in this case distilled water (but could be mineral oil, or other media as discussed here). The patient will be instructed to raise the arm on the side being imaged to better expose the breast tissue. A raised bar attached to the imaging platform will be available for the patient to hold, enhancing comfort during the procedure. During imaging, the OneTouch-PAT system utilizes a round-trip scanning procedure to achieve a field of view of 17cm * 15 cm. The scanning process involves three steps. First, the transducer starts at the bottom left side of the breast and scans upward at 8 mm / s over 17 cm (Figure 1c). Second, the scanning switches to the second translation stage to perform a 7-cm horizontal shift to the right. Finally, the transducer scans downward to cover the right side ofAttorney Docket No.: 011520.02022the breast (Figure Id). The scanning results are combined to form the final imaging data (Figure le). A single round-trip scan takes approximately 52 seconds to complete.

[0093] High-Performance Ultrasound Imaging

[0094] Multi-angle wide-beam imaging strategies were implemented to obtain high-quality US images. Based on the same transducer, a wide-beam transmit sequence was used to send US waves at three different steering angles (-15°, 0°, and 15°), with 42 wide-beam transmissions per angle across the lateral length of the transducer. This large number of transmit beams reduces side lobe levels in the US image, allowing for high-quality US without sacrificing acquisition time. The data were spatially compounded and averaged to form a single US frame. As the PA and US images were acquired from the same transducer, the two modalities are naturally co-registered. Stacking the cross-sectional B-mode US image along the scanning direction creates a 3D volumetric US image of the breast tissue. Compared to the multi-angle plane imaging approach in our previous DSM system, the wide-beam imaging process provided better US image quality and can reveal more details of the breast tissue

[0095] Photoacoustic Data Acquisition and Processing

[0096] As PA signals are much weaker than pulse-echo US, they have a low signal-to-noise ratio. To further enhance the signals, we added a 128-channel 40-dB preamplifier (Uegion AMP, Photosound Technologies) to the Verasonics system. The preamplifier could not handle ultrasound firing, so in a test embodiment, PA and US were performed separately. The patient remained still in front of the imaging window during PA and US scans (each takes 1 minute). This allows the system to acquire both PA and US data at the same position. To ensure accurate registration of PA and US images, we record key anatomical landmarks, including the compressed breast tissue shape, nipple location, and the tumor regions during the two scans. During the image fusion, we align the breast tissue outlines and verity’ that the nipple positions match in both PA and US images. In most of our patient studies, no adjustments were needed to align the PA and US images.

[0097] Noises from PA imaging include thermal noise and electromagnetic interference (EMI) noise from laser firing. Both will contaminate the true PA signals. Additionally, the poor elevational resolution of linear arrays along the scanning direction will degrade the 3D vascular images. To address these challenges, two deep learning networks were built for OneTouch-PAT.Attorney Docket No.: 011520.02022

[0098] One 2D fully-dense Unet (2DFD) was utilized to denoise the raw channel data before image reconstruction. The 2DFD architecture was the same as in our previous publication, except that this study focuses on noise removal instead of resolution improvement. During training, we fed the network with simulated data added with experimentally acquired system noise. This allows the network to leam the noise characteristics of the imaging setup. After training, the network was applied to the experimental breast data to remove noise. Since only noise was removed from the processing, this method can maintain the original vessel intensity for quantitative analysis.

[0099] For better visualization of the 3D vessel structure, we also incorporated our recently developed 3D fully-dense (3DFD) Unet, which will enhance vessel structure and improve vessel continuity. The 3DFD processed data was used for vessel visualization and branching point analysis.

[0100] The two data processing approaches are shown in Figure 2 for comparison. The first row shows the 2DFD U-net model. The second row indicates the data processing procedure for 3DFD Unet. The 3D-trained networks can better preserve the vascular structure in the 3D space by exploring the data's volumetric information rather than cross-sectional images.Meanwhile, the overall image uality was improved as imaging artifacts and noises were further removed. It should be noted that the 3DFD Unet training already incorporated system noise into training. Therefore, we did not use the 2D FD-Unet processed data as input for the 3DFD Unet.

[0101] Tumor Analysis

[0102] We began by comparing our ultrasound images with the clinical US report to identify morphologically similar lesions in our data. Once we identified the relevant frame numbers in the experimental US images, we examined the corresponding PA vascular features around the tumor region. Next, we conducted a quantitative analysis that included quantification of vessel contrast, average vessel signal intensity, vessel branching points, and standard deviation of vessel and background. The statistical analysis w as then done by comparing these features with the results of the healthy contralateral breast.

[0103] Most of the quantitative analysis was performed on the 2DFD data, except for the branching point which calculated from the 3DFD data. To identify the branching points, we used the MATLAB function “bwmorph” to detect branching points from the maximum intensifyAttorney Docket No.: 011520.02022projection (MIP) image. To eliminate false positives caused by overlapping vessels at different depths, the detected 2D points were further validated in 3D by analyzing their local neighborhoods within the volumetric data. Points without sufficient 3D continuity were excluded from the analysis. The branching point density was calculated by dividing the total number of branching points by the total breast area in the projection image.

[0104] Human Subject Recruitment

[0105] Human experiments in this study were approved by the Institutional Review Board of the Roswell Park Comprehensive Cancer Center and University at Buffalo, which also overlooked patient imaging studies conducted at Windsong Radiology. All human subjects provided informed consent after fully understanding the implications of their participation. Windsong patients were recruited by patient navigators, while Roswell Park patients were recruited by study nurses and breast surgeons. The entire imaging procedure for both breasts was less than 10 minutes. The study enrolled four healthy volunteers from University at Buffalo. The healthy volunteer study aimed to test the performance of the OneTouch-PAT system in different cup sizes and verify the reliability of the deep learning data processing method. After that, 61 patients diagnosed with breast cancer participated in the clinical testing. The comprehensive characterization of the patient information is listed in Table I. Results from five patients were selected to demonstrate the dual-modal PA / US imaging capabilities, while all patients’ PA results were used for the statistical analysis.

[0106] Results

[0107] Results are presented in three sections. First, we evaluated OneTouch-PAT on four healthy subjects with vary ing breast sizes and skin tones, demonstrating its robustness across different breast compositions using optimized data processing algorithms. Next, a clinical validation study was conducted, showing that OneTouch-PAT enables clear visualization of tumor features in both 2D and 3D data. Finally, a quantitative analysis of 61 patients revealed that photoacoustic imaging provides statistically significant differences in signal intensity and vessel distribution betw een malignant and healthy breast tissue.Table 1: Summary of patent data.Attorney Docket No.: 011520.02022<ILC: Invasive Lobular Carcinoma; IDC: Invasive Ductal Carcinoma;DCIS: Ductal Carcinoma In Situ; LCIS: Lobular Carcinoma In Situ;ADH: Atypical Ductal Hyperplasia. LUMA: Luminal A; LUMB:Luminal B; TNBC: Triple-negative breast cancer.

[0108] Data Processing and Healthy Subject Results

[0109] Four healthy volunteers with different breast cup sizes and skin tones were recruited to demonstrate the system's performance. The results are shown in Figure 3, where images were projected along the axial direction and color-encoded with depth (blue represents shallow and red represents deep). The OneTouch-PAT system provides consistent imaging of blood vessels at different cup sizes. For instance, blood vessels can be clearly visualized at the top row of the small cup size A, as well as in the bottom row of cup size D. The subject with a cup size of B has a dark skin color, which would provide a stronger background signal. The deep learning process eliminates most of the skin signals, thereby revealing deeper vascular features.Attorney Docket No.: 011520.02022

[0110] Patient Imaging Results

[0111] In this section, we present the morphological and vascular tumor features in 2D and 3D. Five representative cases are presented to cover three molecular subtypes of breast cancer: LUMA, LUMB, and TNBC.

[0112] Figure 4 displays PA&US results from two patients with LUMA breast tumors. Case 1 involved a patient with fair skin color and cup size C, having a 7x5x6 mm3tumor (LUMA) on the right breast at 12 o’clock, 6 cm from the nipple. The clinical US report (Figure 4a) shows an oval-circumscribed hypoechoic mass with central echogenicity. The same feature is observed in the One-Touch US image (Figure 4b). Figure 4c presents the PA features overlaid on top of the cross-sectional US image where some blood vasculatures can be seen. However, the presence of the vessel is dependent on the orientation of the cross-section. In comparison, the 3D renderings in Figure 4d and Figure 4e show more peripheral vessels, highlighting the importance of volumetric imaging. It should be noted that the orange-red colored vessels at the top left comer of Figure 4e might be due to natural curvature of the breast as our plastic membrane was flexible and did not conform the breast to a flat surface.

[0113] The second LUMA case involved a patient with fair skin color and cup size D, having a 15 x 15 x 17 mm3tumor on the left breast at 10 o’ clock, 3 cm from the nipple. The clinical US image shows a lobulated, partially well-circumscribed, heterogeneously enhancing mass (Figure 4f), which is also visible in the One-Touch US image (Figure 4g). Similar to the first case, we noticed more obvious vascular patterns in the volumetric image than the cross-sectional images (Figure 4h-j). In both LUMA cases, we observed that the peripheral vessels of the tumor were more prominent than the internal vessels.

[0114] Figure 5 presents PA&US images of two patients (case 3&4) with LUMB subtype. In case 3, the patient (breast cup size A and fair skin color) has a 22x18x24 mm3tumor behind the left nipple with overlying skin thickening. The clinical US image (Figure 5a) shows an irregular hypogenic lesion which matches with the One-Touch US image (Figure 5b). The PA cross-sectional and volumetric images in Figure 5c-e show richer vasculature around the tumor core. In case 4, the patient (cup D breast and fair skin color) has a 32x17x19 mm3spiculate mammographic mass in the left breast. Again, a hypogenic lesion with an irregular shape is shown in both clinical (Figure 5f) and One-Touch US images (Figure 5g). The PA images show that the tumors have abundant vessels surrounding the periphery and fewer vessels in the centralAttorney Docket No.: 011520.02022region (Figure 5h-j). In particular, compared to LUMA cases, we noted a more pronounced presence of feeding vessels leading to the tumor core in both LUMB cases. This vascular richness in LUMB tumors suggests a higher angiogenic activity, likely contributing to their higher proliferation rates and more aggressive tumor behavior, as indicated by their elevated Ki-67 indices.

[0115] Figure 6 illustrates the PA&US results of a malignant lesion of the TNBC subtype. Case 5 involved a patient with fair skin color and cup size C. The patient had a lobulated sonographic mass with indistinct margins located at 11 o'clock in the right breast, measuring 22x18x22 mm3. Both clinical (Figure 6a) and One-Touch ultrasound (Figure 6b) images revealed a ~20 mm lesion with a circumscribed margin and round shape; both are characteristic features of TNBC. Within the tumoral region, we observed spotty high-intensity photoacoustic signals (Figure 6c), which are likely attributed to intertumoral hemoglobin enhancement. Interestingly, we did not observe clear feeding vessels in this TNBC case (Figure 6d-e), potentially due to the lack of ER and HER2 expression, which created a less structure vascular network.

[0116] Statistical Analysis

[0117] Statistical analysis was conducted using MATLAB with a type I error rate of 0.05 (a). Upper-tailed t-tests were utilized to compare different parameters for healthy and tumorbearing breasts. Based on our earlier study on the DSM system, we analyzed the regional PA intensity (Ave Intensity), the background PA intensity (Ave Background), the standard deviation of the background PA intensity (STD Background), and the extracted vessel intensity (Ave Vessel) across 61 patients (Table II). The data from the tumor-bearing breast was normalized by dividing it by the corresponding data from the healthy breast. Figure 7 provides a summary of our statistical results.

[0118] As shown in Figure 7a, the malignant breasts exhibit higher relative photoacoustic intensity (> 1) compared to the healthy breasts. Our analysis revealed that tumor-bearing breasts exhibited stronger regional signal intensity (P = 0.0002 for upper tailed t-test) and vessel intensity (P = 0.0062) than healthy breasts. This finding suggests the presence of larger vessels and abundant microvessels in malignant breasts. This observation is consistent with current research that indicates an increase in regional vascularity in malignant breasts. Additionally, the mean signal intensity' of the background (regions with no visible vessel) (P = 0.0007) and theAttorney Docket No.: 011520.02022standard deviation (P = 0.0176) were higher in malignant breasts (Figure 7a), which suggests a strong variation in tissue signals potentially due to the growth of microvasculature. These results are consistent with our earlier reports in the DSM system and indicate that the OneTouch setup can be used to differentiate healthy breasts and tumor-bearing breasts.

[0119] Moreover, we analyzed the tumor-bearing breast tissue on different breast densities. In Figure 7b. the heterogeneous breast exhibited slightly higher average intensity and vessel intensity compared to the scattered fibroglandular breast (extremely dense breast cases were excluded as only 2 cases were available). This difference highlights the greater vascular complexity and density in malignant heterogeneously dense tissue. This observation agrees with studies reporting that angiogenesis-related proteins are upregulated in dense breast tissue.Table II: Summary of Quantitative PA Parameters Name ExplanationAve Intensity’ Average PA intensity within the whole breast.Ave Vessel Average PA intensity within the segmentedvessel regionAve Background Average PA intensity within the non-vesselregionSTD Background Standard deviation of the PA intensity' withinthe non-vessel region.

[0120] As PA imaging is sensitive to skin color, we made a comparison plot based on skin tone (Figure 7c). As we do not have sufficient cases of medium skin tone, we only compared fair and dark skin cases. It can be seen that the average PA intensity ratio is slightly lower in the darker tone cases, indicating that the background skin signals might have reduced the contrast between malignant and healthy breasts. However, the mean vessel ratio is slightly higher in the dark cases, indicating that once all the background signals were removed, the vessel contrast would still stand out. This result highlights the importance of skin effect compensation in PA breast imaging. While we observed a slight difference in the above-mentioned mean ratios, we did not observe a statistically significant difference between the two breast density' or skin color groups. These results indicate that the PA technology is not sensitive to these effects.

[0121] Besides quantification of the PA signal intensity, we also analyzed the vessel branching points. This investigation was performed on the 3D Unet processed data, which improved image resolution and vessel continuity. We noticed that the tumor-bearing breast exhibited higher average branching point density (49.1 / 100 cm2) than the contralateral healthyAttorney Docket No.: 011520.02022patient (38.2 / 100 cm2) (P < 0.005). This observation is consistent with the previous reports from Yamaga et al., who observed 31.7 and 27.0 / 100 cm2density for malignant and healthy tissue, respectively. However, our overall density is higher. We believe the difference in system resolution and branching point quantification methods induced this difference. In particular, the previous report relied on an operator to manually count the branching point, while our method is completely automated, enabling the detection of subtle vessels and corresponding branching points that can be difficult to distinguish with unaided eyes.

[0122] In this study, we developed and evaluated the performance of a new PA / US breast imaging system, OneTouch-PAT. Compared to DSM, the new system offers better patient comfort and higher ultrasound image quality. In addition, deep learning techniques were utilized to enhance the visualization of breast vasculature. The OneTouch can image the vasculature of the breast at a depth of, for example, 3 cm with a field of view of 15 x 17 cm2. The PA and US images from OneTouch-PAT were combined and compared to clinical ultrasound images to localize the tumor site. Representative results from five patients were highlighted to demonstrate the advantage of 3D imaging. In most cases, we noticed blood vessels surrounding the tumor run irregularly, which are characteristics of tumor neovascularization.

[0123] Compared to handheld PA-US breast imaging systems, the automated breast scanning removed the operator dependence and provided a more comprehensive view of the breast vasculature. As can be seen in Figures 4-6, the cross-sectional view can only reveal limited vascular information, while the 3D view provides complete vasculature around the malignant lesion. By sectioning across the suspicious regions, w e can get a complete picture of the morphological information of the lesion using US. The tumor shapes and structure agree with typical findings in the US regarding tumor subtypes, while the addition of PA vasculature provides additional functional information. Our case studies indicated that the most significant image pattern associated with LUMA and LUMB malignancies w as the presence of feeding vessels. On the other hand, TNBC malignancies were obser ed with spotty signals in the tumoral region. This may be partially attributed to the lack of ER and HER2 expression, which is associated with diminished angiogenic signaling and a less organized vascular architecture, resulting in poorly structured or fragmented vascular networks. These results agree with cross-sectional studies from a larger patient population, highlighting the promise of combining volumetric PA and US images for breast tumor subtype identification and personalized treatment planning.Attorney Docket No.: 011520.02022

[0124] Overall, the Onetouch-PAT system offers a portable and effective solution for breast cancer imaging, providing precise and accurate images with a large field of view and deep penetration depth. In addition, the system provides a native combination of PA and US. Relying on the US images, we can identify suspicious lesions and then overlay the anatomical structure of the US image on the PA image. Overall, the OneTouch PAT system was able to identify unique malignancy-associated image features within the tumoral region in the five presented cases across three different breast tumor subtypes.

[0125] In summary, we disclosed an automated PA and US system that images patients in the standing pose. Compared to X-ray mammogram, the One-Touch system does not require painful compression and the sensitivity does not degrade in dense breast tissue. Compared to MRI, the system does not require radiation or contrast injection, and it is not as expensive or time-consuming. Given the automated scanning nature, operator variability is almost negligible in Onetouch-PAT system compared to handheld ultrasound. We also developed a data processing approach for better processing of PA data and enhancement of vascular features. The combination of the hardware and software design allows OneTouch-PAT to provide coregistered PA and US images in 3D. US provides morphological features of the breast, while the PA overlay shows the vascular patterns around the tumor, potentially providing more accurate classification of tumor grade and subtypes. Our results also demonstrated that PA imaging provides statistically significant differences in signal intensity and vessel distribution between malignant and healthy breast tissues. These results highlight the potential of PA as a complementary tool for diagnostic breast cancer imaging.

[0126] Combining features from both modalities, we can easily identify the suspicious lesion and confirm PA features around the lesion. Our statistical analysis from 61 patients and the highlighted patient cases indicate that OneTouch can be anew imaging modality for breast cancer diagnosis and screening.

[0127] Although the present disclosure has been described with respect to one or more particular embodiments, it will be understood that other embodiments of the present disclosure may be made without departing from the spirit and scope of the present disclosure.

Claims

Attorney Docket No.: 011520.02022What is claimed is:

1. An automated system for breast imaging of a patient in a standing pose, comprising:a pulsed light source configured to irradiate a region of interest in breast tissue;an ultrasound transducer configured to transmit ultrasound pulses into, and receive acoustic signals from, the region of interest:one or more optical components configured to align an optical path of the pulsed light source and an ultrasound path of the ultrasound transducer such that the optical path and the ultrasound path are at least one of co-planar or co-axial; anda controller in electronic communication with the pulsed light source and the ultrasound transducer, wherein the controller is configured to acquire photoacoustic data and ultrasound data in an interleaved acquisition sequence that, for each of a plurality of scan positions corresponding to a respective cross-section of the region of interest, comprises: (a) triggering one or more light pulses from the pulsed light source to irradiate a respective illumination region at the scan position, wherein illumination regions corresponding to successive light pulses overlap;(b) receiving, from the ultrasound transducer, light-generated acoustic signals corresponding to the one or more light pulses;(c) reconstructing, based on the received light-generated acoustic signals, a photoacoustic image of the cross-section at the scan position;(d) after receiving the light-generated acoustic signals, transmitting a plurality of ultrasound pulse signals from the ultrasound transducer, the plurality of ultrasound pulse signals being associated with the cross-section at the scan position; (e) receiving, from the ultrasound transducer, a plurality of echo signals corresponding to the plurality of ultrasound pulse signals; and(f) reconstructing, based on the received echo signals, an ultrasound image of the cross-section at the scan position;wherein operations (a)-(f) form one frame including the photoacoustic image and the ultrasound image of the cross-section at the scan position; andwherein the controller is further configured to repeat operations (a)-(f) at different scan positions to scan the region of interest.

2. The system of claim 1, wherein the interleaved acquisition sequence comprises alternating photoacoustic acquisition and ultrasound acquisition at each scan position.Attorney Docket No.: 011520.020223. The system of claim 1, wherein the interleaved acquisition sequence comprises performing photoacoustic acquisition and ultrasound acquisition sequentially at the scan position while the patient remains in the standing pose.

4. The system of claim 1, wherein the controller is configured to reconstruct, from frames acquired at the plurality of scan positions, a three-dimensional photoacoustic representation and a three-dimensional ultrasound representation.

5. The system of claim 4, wherein the controller is configured to generate co-registered output that spatially aligns the three-dimensional photoacoustic representation with thethree-dimensional ultrasound representation.

6. The system of claim 1, further comprising a positioning subsystem configured to translate at least one of the ultrasound transducer or the one or more optical components along a scan trajectory to implement the plurality of scan positions.

7. The system of claim 6, wherein the positioning subsystem comprises one or more translation stages.

8. The system of claim 1, further comprising a vessel configured to contain an acoustic coupling medium and having an imaging window positioned to receive the breast of the standing patient.

9. The system of claim 8, wherein the imaging window is configured to permit transmission of the pulsed optical energy and acoustic energy.

10. The system of claim 8, wherein the imaging window is sealed by a film configured to contact tissue and permit acoustic and optical transmission.

11. The system of claim 8, wherein the acoustic coupling medium comprises water or mineral oil.

12. The system of claim 8, wherein the acoustic coupling medium comprises heavy water (D2O) or other liquid having lower optical attenuation at 1064 nm than water.

13. The system of claim 8, further comprising a height- adjustable support platform configured to position the imaging window relative to the patient.Attorney Docket No.: 011520.0202214. The system of claim 1, wherein the ultrasound transducer comprises an ultrasound transducer array.

15. The system of claim 14, wherein the ultrasound transducer array is a linear array transducer or a curved array.

16. The system of claim 1, wherein the one or more optical components comprise an optical -acoustic combiner comprising a first reflector configured to reflect the pulsed optical energy toward a shared propagation region and a second reflector configured to transmit at least a portion of the pulsed optical energy' while reflecting acoustic signals toward the ultrasound transducer.

17. The system of claim 16, wherein the first reflector and the second reflector comprise dichroic mirrors.

18. The system of claim 1, wherein the pulsed light source is configured to output near-infrared light.

19. The system of claim 1, wherein the controller is configured to perform dual speed of sound back projection reconstruction using a first speed of sound for acoustic propagation through a coupling-medium region and a second speed of sound for acoustic propagation through a tissue region.

20. The system of claim 1, wherein the pulsed light source is configured to emit optical pulses at a plurality of wavelengths.

21. The system of claim 1, wherein the plurality of wavelengths includes at least two wavelengths selected from about 680 nm to about 1300 nm.

22. The system of claim 1, wherein the controller is configured to reconstruct a respective photoacoustic image or photoacoustic image volume for each wavelength of the plurality of wavelengths.

23. A dual-modality' breast imaging system, comprising:a vessel configured to contain an acoustic coupling medium and having an imaging window positioned to receive a breast of a standing subject, the imaging window configured to permit transmission of optical energy' and acoustic energy;Attorney Docket No.: 011520.02022a pulsed light source configured to deliver pulsed optical energy toward breast tissue through the imaging window;an ultrasound transducer array positioned in the vessel and configured to (i) receive photoacoustic signals generated in response to the pulsed optical energy and (ii) receive ultrasound echo signals;a positioning subsystem coupled to the ultrasound transducer array and configured to translate the ultrasound transducer array along a scan trajectory that includes:(a) a first longitudinal scan along the breast.(b) a lateral translation, and(c) a second longitudinal scan along the breast in a direction opposite the first longitudinal scan; anda controller configured to:acquire photoacoustic receive data during at least the first longitudinal scan and the second longitudinal scan,combine photoacoustic receive data from the first longitudinal scan and the second longitudinal scan to form a composite three-dimensional photoacoustic representation of the breast in a frontal view having a field of view of at least 15 cm by 17 cm, acquire ultrasound echo data along the scan trajectory to form a three-dimensional ultrasound representation of the breast, andoutput co-registered three-dimensional photoacoustic and ultrasound representations.

24. The system of claim 23, wherein the imaging window is sealed by a film configured to contact tissue and permit acoustic and optical transmission.

25. The system of claim 23, wherein the coupling medium comprises water or mineral oil.

26. The system of claim 23, wherein the coupling medium comprises heavy' water (D2O) or other liquid having lower optical attenuation at 1064 nm than water.

27. The system of claim 23, further comprising a height- adjustable support platform configured to position the imaging window relative to the standing subject.

28. The system of claim 23, wherein the positioning subsystem compnses first and second translation stages, and wherein the scan trajectory is executed by translating along the firstAttorney Docket No.: 011520.02022translation stage for each of the first longitudinal scan and the second longitudinal scan and translating along the second translation stage for the lateral translation.

29. The system of claim 23, wherein the controller is configured to combine datasets from the first longitudinal scan and the second longitudinal scan using an overlap region between the scans.

30. The system of claim 23, wherein the ultrasound transducer array is a linear array transducer or a curved array.

31. The system of claim 23, wherein the controller is configured to form an ultrasound frame using a multi-angle wide-beam transmit sequence and spatial compounding.

32. The system of claim 31, wherein the multi-angle wide-beam transmit sequence includes at least three steering angles.

33. The system of claim 31, wherein the controller is configured to compound and average echo data acquired from a plurality of transmissions per steering angle to form the ultrasound frame.

34. The system of claim 23, further comprising an optical-acoustic combiner positioned to align an optical path of the pulsed light source with an acoustic receive path of the ultrasound transducer array.

35. The system of claim 34, wherein the optical-acoustic combiner comprises a first dichroic mirror configured to reflect the pulsed optical energy toward a shared propagation region and a second dichroic mirror configured to transmit at least a portion of the pulsed optical energy and reflect photoacoustic signals toward the ultrasound transducer array.

36. The system of claim 23, wherein the pulsed light source is configured to provide light having a wavelength in a near-infrared range.

37. The system of claim 23, wherein the controller is configured to generate an overlay output comprising photoacoustic intensity overlaid on a grayscale ultrasound image.

38. The system of claim 23, wherein the controller is configured to output the co-registered three-dimensional photoacoustic and ultrasound representations within about one minute for a scan of a breast.Attorney Docket No.: 011520.0202239. The system of claim 23, wherein the controller is configured to perform dual speed of sound back projection reconstruction using a first speed of sound for acoustic propagation through a coupling-medium region and a second speed of sound for acoustic propagation through a tissue region.

40. The system of claim 23, wherein the pulsed light source is configured to emit optical pulses at a plurality of wavelengths.

41. The system of claim 23, wherein the plurality of wavelengths includes at least two wavelengths selected from about 680 nm to about 1300 nm.

42. The system of claim 23, wherein the controller is configured to reconstruct a respective photoacoustic image or photoacoustic image volume for each wavelength of the plurality of wavelengths.

43. A computer-implemented method for generating a dual-modality breast image set, comprising:receiving, from an ultrasound transducer array, photoacoustic raw channel data acquired during a scan of a breast while the breast is positioned against an imaging window; applying a first neural network to the photoacoustic raw channel data to reduce noise prior to image reconstruction;reconstructing, based on the noise-reduced photoacoustic raw channel data, a plurality of photoacoustic image frames and generating a three-dimensional photoacoustic image volume based on the plurality of photoacoustic image frames;applying a second neural network to the three-dimensional photoacoustic image volume to enhance vascular structure and reduce background or skin-related signal components; receiving ultrasound echo data acquired with the ultrasound transducer array while the breast remains positioned against the imaging window;reconstructing, from the ultrasound echo data, a three-dimensional ultrasound image volume;andgenerating co-registered output comprising the enhanced three-dimensional photoacoustic image volume spatially aligned with the three-dimensional ultrasound image volume.

44. The method of claim 43, wherein applying the first neural network comprises applying a two-dimensional neural network to the photoacoustic raw channel data.Attorney Docket No.: 011520.0202245. The method of claim 43, wherein applying the first neural network comprises applying a two-dimensional fully-dense U-net to the photoacoustic raw channel data.

46. The method of claim 43, wherein applying the second neural network comprises applying a three-dimensional neural network to volumetric photoacoustic data.

47. The method of claim 43, wherein applying the second neural network comprises applying a three-dimensional fully-dense U-net to the three-dimensional photoacoustic image volume.

48. The method of claim 43, wherein generating the three-dimensional photoacoustic image volume comprises stacking reconstructed photoacoustic image frames acquired at a plurality of scan positions.

49. The method of claim 43, further comprising generating a maximum-amplitude projection image from the plurality of photoacoustic image frames, wherein the maximum-amplitude projection preserves photoacoustic signal intensity for quantitative analysis.

50. The method of claim 43, wherein receiving ultrasound echo data comprises receiving ultrasound echo data acquired using a multi-angle wide-beam transmit sequence and spatial compounding.

51. The method of claim 43, wherein receiving photoacoustic raw channel data comprises receiving photoacoustic raw channel data acquired during (i) a first longitudinal scan, (ii) a lateral translation, and (ni) a second longitudinal scan opposite the first longitudinal scan, and wherein generating the three-dimensional photoacoustic image volume comprises combining photoacoustic data from the first longitudinal scan and the second longitudinal scan.

52. The method of claim 43, further comprising aligning the three-dimensional photoacoustic image volume with the three-dimensional ultrasound image volume using at least one anatomical landmark.

53. The method of claim 52, wherein the at least one anatomical landmark comprises a nipple position, a breast tissue outline, or both.

54. The method of claim 43, further comprising generating a display output that overlays photoacoustic intensity on a grayscale ultrasound image.Attorney Docket No.: 011520.0202255. The method of claim 43, wherein reconstructing comprises back projection using a first speed of sound for a coupling medium region and a second speed of sound for a tissue region.

56. The method of claim 43, further comprising defining the coupling medium region based on a fixed geometry of an imaging window and a transducer position.

57. The method of claim 43, wherein the first speed of sound is about 1.3-1.5 mm / ps and the second speed of sound is about 1.45-1.60 mm / ps.

58. The method of claim 43, wherein receiving photoacoustic raw channel data includes receiving photoacoustic raw channel data corresponding to optical excitation at a plurality of wavelengths.

59. The method of claim 43, further including reconstructing a photoacoustic image volume for each of the plurality of wavelengths.

60. The method of claim 43, further including estimating an oxygenation parameter based on the photoacoustic image volumes for the plurality of wavelengths and outputting an oxygenation map aligned with the three-dimensional ultrasound image volume.

61. A non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform the method of any of claims 37-60.

Images