Magnetoencephalography (MEG) is a pivotal neuroimaging technique for diagnosing and treating brain disorders, known for its precise measurement of the brain’s magnetic fields due to electrical activity. Accurate brain source localization is essential for neurosurgical planning, but the MEG inverse problem—determining brain source locations from MEG data—is complex and inherently ill-posed. This article introduces a novel, data-driven approach to enhance MEG source localization and brain activity characterization. We compare three encoder models, VGGNet, ViT, and ResNet, assessing their performance across varying noise levels. Our findings reveal the effectiveness of neural networks in addressing challenging neuroimaging problems, underscoring their potential in advancing MEG applications.
Ultrasound computed tomography (USCT) is an emerging imaging modality that holds great promise for breast imaging. Full-waveform inversion (FWI)-based image reconstruction methods incorporate accurate wave physics to produce high spatial resolution quantitative images of speed of sound or other acoustic properties of breast tissues from USCT measurement data. However, FWI is computationally burdensome and requires a good initial guess of the speed of sound distribution due to the nonconvex nature of the underlying optimization problem (cycle-skipping). Alternatively, the use of a simplified linear model, such as the Born approximation, allows the image reconstruction problem to be formulated as a convex optimization problem, but sacrifices accuracy. This work proposes utilizing a convolutional neural network (CNN) to correct pressure data and accurately reconstruct images using a simplified forward model, thus combining the benefits of accurate reconstructions from traditional FWI methods with the reduced computational complexity of inversion with simplified models. Furthermore, applying this correction to the measurements before inversion avoids issues inherent to other deep learning reconstruction methods that first invert and then apply correction to the images. Specifically, correction in the measurement domain is well-defined by a mathematical model and avoids hallucinations by an improperly learned image prior. This reconstruction approach was validated with a set of anatomically realistic test images and compared to traditional reconstruction methods (FWI and uncorrected Born inversion), a data-driven learned reconstruction method, and a machine learning method for artifact correction in the image domain after reconstructing using an inaccurate physics model.
Current ultrasound imaging techniques face challenges in producing clear brain images, primarily due to the contrast in sound velocity between the skull and brain tissues, and the difficulty in effectively coupling large probes with skulls. The reverse time migration (RTM) technique, known for its effectiveness in the geophysics community, is utilized to address this coupling issue. In addition, we propose the use of smaller probes capable of generating limited ultrasound brain image fragments from various angles. Subsequently, we have developed a new brain imaging method, termed BrainPuzzle, to restore brain images from these limited fragments. Unlike traditional transformer-based image generation models, BrainPuzzle not only uses a transformer to recognize and rearrange the fragments into their correct positions but also integrates a graph convolutional network (GCN) to automatically capture the spatial relationships among the fragments, thereby enhancing the model’s capabilities. Furthermore, we introduce the concept of using the RTM method to generate these ultrasound brain image fragments. The experimental results based on two distinct sets of generated datasets, demonstrate the exceptional performance of the proposed method in reconstructing the complete brain images from the fragments of ultrasound brain images.
Ultrasound computed tomography (USCT) is an emerging imaging modality that holds great promise for breast imaging. Full-waveform inversion (FWI)-based image reconstruction methods incorporate accurate wave physics to produce high spatial resolution quantitative images of speed of sound or other acoustic properties of the breast tissues. However, FWI reconstruction is computationally expensive, which limits its application in a clinical setting. This contribution investigates using the use of a convolutional neural network (CNN) to learn a mapping from USCT data to speed of sound estimates. The CNN was trained using a supervised approach that employed a large set of anatomically and physiologically realistic numerical breast phantoms (NBPs) and simulated USCT measurements. Once trained, this CNN can then be evaluated for real-time FWI image reconstruction from USCT data. The performance of the proposed method was assessed and compared against FWI using a hold-out sample of 41 NBPs and corresponding USCT images. Accuracy was measured using relative mean square error (RMSE) and structural self-similarity index measure (SSIM). This numerical experiment demonstrates that a supervised learning model can achieve accuracy comparable to FWI, while significantly reducing computational time and memory requirements.
KEYWORDS: Ultrasonography, Ultrasound tomography, Breast, Transducers, Breast cancer, Data acquisition, Tomography, Algorithm development, Signal attenuation, Design for manufacturability, Prototyping, Mammography, In vivo imaging
Breast ultrasound tomography is an emerging imaging modality to reconstruct the sound speed, density, and ultrasound attenuation of the breast in addition to ultrasound reflection/beamforming images for breast cancer detection and characterization. We recently designed and manufactured a new synthetic-aperture breast ultrasound tomography prototype with two parallel transducer arrays consisting of a total of 768 transducer elements. The transducer arrays are translated vertically to scan the breast in a warm water tank from the chest wall/axillary region to the nipple region to acquire ultrasound transmission and reflection data for whole-breast ultrasound tomography imaging. The distance of these two ultrasound transducer arrays is adjustable for scanning breasts with different sizes. We use our breast ultrasound tomography prototype to acquire phantom and in vivo patient ultrasound data to study its feasibility for breast imaging. We apply our recently developed ultrasound imaging and tomography algorithms to ultrasound data acquired using our breast ultrasound tomography system. Our in vivo patient imaging results demonstrate that our breast ultrasound tomography can detect breast lesions shown on clinical ultrasound and mammographic images.
Ultrasound waveform tomography with the total-variation regularization could improve reconstructions of tumor margins,
but the reconstructions usually contain unwanted blocky artifacts. We develop a new ultrasound waveform tomography
method with a second-order total-generalized-variation regularization scheme to improve tomographic reconstructions
of breast tumors and remove blocky artifacts in reconstruction results. We validate our new method using numerical
phantom data and real phantom data acquired using our synthetic-aperture breast ultrasound tomography system with
two parallel transducer arrays. Compared to reconstructions of ultrasound waveform tomography with modified total-variation
regularization, our new ultrasound waveform tomography yields accurate sound-speed reconstruction results
with significantly reduced artifacts.
The sound-speed distribution of the breast can be used for characterizing breast tumors, because they typically have a higher sound speed than normal breast tissue. This is understood to be the result of remodeling of the extracellular matrix surrounding tumors. Breast sound-speed distribution can be reconstructed using ultrasound bent-ray tomography (USRT). We have recently demonstrated that USRT, using arrival times of both transmission and reflection data, significantly improves image quality. To further improve the robustness of tomographic reconstructions, we develop a USRT method using a modified total-variation (MTV) regularization scheme. Regularization is often used in solving inverse problems by introducing restrictions such as for smoothness. Tikhonov regularization is a widely used regularization scheme that tends to smooth tomographic images, but oversmoothing can obscure critical diagnostic detail such as tumor margins. Total-variation (TV) regularization is another common regularization scheme that preserves tumor margins, but at the cost of increased image noise. Our new USRT with MTV regularization is a Tikhonov-TV hybrid, reducing image noise while preserving margins. We apply our new method to ultrasound transmission data from numerical phantoms, and compare the results with those obtained using Tikhonov regularization.
KEYWORDS: Ultrasonography, Transducers, Breast, Ultrasound tomography, Imaging systems, Data acquisition, Tomography, Prototyping, Mammography, Breast cancer
Ultrasound tomography has great potential to provide quantitative estimations of physical properties of breast tumors for accurate characterization of breast cancer. We design and manufacture a new synthetic-aperture breast ultrasound tomography system with two parallel transducer arrays. The distance of these two transducer arrays is adjustable for scanning breasts with different sizes. The ultrasound transducer arrays are translated vertically to scan the entire breast slice by slice and acquires ultrasound transmission and reflection data for whole-breast ultrasound imaging and tomographic reconstructions. We use the system to acquire patient data at the University of New Mexico Hospital for clinical studies. We present some preliminary imaging results of in vivo patient ultrasound data. Our preliminary clinical imaging results show promising of our breast ultrasound tomography system with two parallel transducer arrays for breast cancer imaging and characterization.
Ultrasound transmission tomography usually generates low-resolution breast images. We improve sound-speed reconstructions using ultrasound waveform tomography with both transmission and reflection data. We validate the improvement using computer-generated synthetic-aperture ultrasound transmission and reflection data for numerical breast phantoms. Our tomography results demonstrate that using both transmission and reflection data in ultrasound waveform tomography greatly enhances the resolution and accuracy of tomographic reconstructions compared to ultrasound waveform tomography using either transmission data or reflection data alone. To verify the capability of our novel ultrasound waveform tomography, we design and manufacture a new synthetic-aperture breast ultrasound tomography system with two parallel transducer arrays for clinical studies. The distance of the two transducer arrays is adjustable for accommodating different sizes of the breast. The parallel transducer arrays also allow us to easily scan the axillary region to evaluate the status of axillary lymph nodes and detect breast cancer in the axillary region. However, synthetic-aperture ultrasound reflection data acquired by firing each transducer element sequentially are usually much weaker than transmission data, and have much lower signal-to-noise ratios than the latter. We develop a numerical virtual-point-source method to enhance ultrasound reflection data using synthetic-aperture ultrasound data acquired by firing each transducer element sequentially. Synthetic-aperture ultrasound reflection data for a breast phantom obtained using our numerical virtual-point-source method reveals many coherent ultrasound reflection waveforms that are weak or invisible in the original synthetic-aperture ultrasound data. Ultrasound waveform tomography using both transmission and reflection data together with numerical virtual-point-source method has great potential to produce high-resolution tomographic reconstructions in clinical studies of breast ultrasound tomography.
Regularization is often needed in breast ultrasound waveform tomography to improve tomographic reconstructions. A global regularization parameter may lead to either over-regularization or under-regularization in different regions in the imaging domain. We develop a new ultrasound waveform tomography method with spatially-variant regularization. Our new method employs different regularization parameters in different regions of the breast so that each regularization parameter is optimal for the local region. Our numerical examples demonstrate the improvement of ultrasound waveform tomography using the spatially-variant modified total-variation regularization for sound-speed reconstructions of large and small breast tumors, particularly when their sizes are significantly different from one another.
Early detection of breast cancer is the key to reducing the cancer mortality rate. With increasing computational power, waveform inversion becomes feasible for high-resolution ultrasound tomography. Because of the limited measurement geometry, ultrasound waveform tomography is usually ill-posed, which requires certain computational methodologies to stabilize waveform inversion. We develop a new ultrasound waveform tomography method using a modified total-variation regularization for detecting and characterizing small breast tumors. To solve the minimization problem, we use an alternating-minimization algorithm in which the original optimization is equivalently decomposed
into two simple subproblems. We use numerical breast-phantom data to demonstrate the improved capability of our new tomography method for accurately reconstructs the sound-speed values and shapes of small tumors.
Ultrasound could be an attractive imaging modality for detecting breast microcalcifications, but it requires significant
improvement in image resolution and quality. Recently, we have used tissue-equivalent phantoms to demonstrate that
synthetic-aperture ultrasound has the potential to detect small targets. In this paper, we study the in vivo imaging capability
of a real-time synthetic-aperture ultrasound system for detecting breast microcalcifications. This LANL's (Los Alamos
National Laboratory's) custom built synthetic-aperture ultrasound system has a maximum frame rate of 25 Hz, and is one
of the very first medical devices capable of acquiring synthetic-aperture ultrasound data and forming ultrasound images in
real time, making the synthetic-aperture ultrasound feasible for clinical applications. We recruit patients whose screening
mammograms show breast microcalcifications, and use LANL's synthetic-aperture ultrasound system to scan the regions
with microcalcifications. Our preliminary in vivo patient imaging results demonstrate that synthetic-aperture ultrasound is
a promising imaging modality for detecting breast microcalcifications.
Ultrasound waveform tomography takes wave propagation effects into account during image reconstruction,
and has the potential to produce accurate estimates of the sound speeds of small breast tumors.
However, waveform tomography is computationally time-consuming for large datasets acquired
using a synthetic-aperture ultrasound tomography system that consists of hundreds to thousands of
transducer elements. We introduce a source encoding approach to ultrasound waveform tomography
to significantly improve the computational efficiency. The method simultaneously simulates ultrasound
waveforms emitted from multiple transducer elements. To distinguish the effect of different
sources, we apply a random phase to each source. The random phase helps eliminate the unwanted
cross interferences produced by different sources. This approach greatly reduces the computational
time of ultrasound waveform tomography to one tenth of that for the original waveform tomography,
and makes it feasible for ultrasound waveform tomography in clinical applications.
Waveform tomography has the potential to quantitatively reconstruct the sound speed values of breast
tumors. It is difficult to obtain quantitative values of the sound speed of breast tumors when their
sizes are in the order of, or smaller than, the ultrasound wavelength. Because of the ill-posedness
of the full-waveform inversion, regularization techniques are usually used to improve reconstruction.
We develop an ultrasound waveform tomography method with the total-variation regularization to
improve sound-speed reconstructions of small breast tumors. Our numerical examples demonstrate
that our ultrasound waveform tomography with the total-variation regularization is a promising tool
for quantitative estimation of the sound speed of small breast tumors.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.