Conventional wisdom dictates that imaging hardware should be optimized by use of an ideal observer (IO) that exploits full statistical knowledge of the class of objects to be imaged, without consideration of the reconstruction method to be employed. However, accurate and tractable models of the complete object statistics are often difficult to determine in practice. Moreover, in imaging systems that employ compressive sensing concepts, imaging hardware and (sparse) image reconstruction are innately coupled technologies. We have previously proposed a sparsity-driven ideal observer (SDIO) that can be employed to optimize hardware by use of a stochastic object model that describes object sparsity. The SDIO and sparse reconstruction method can therefore be "matched" in the sense that they both utilize the same statistical information regarding the class of objects to be imaged. To efficiently compute SDIO performance, the posterior distribution is estimated by use of computational tools developed recently for variational Bayesian inference. Subsequently, the SDIO test statistic can be computed semi-analytically. The advantages of employing the SDIO instead of a Hotelling observer are systematically demonstrated in case studies in which magnetic resonance imaging (MRI) data acquisition schemes are optimized for signal detection tasks.
Proc. SPIE. 10132, Medical Imaging 2017: Physics of Medical Imaging
KEYWORDS: Optical imaging, Data modeling, Imaging systems, Magnetic resonance imaging, Image segmentation, Reliability, Image restoration, Medical image reconstruction, Image analysis, Medical imaging, Monte Carlo methods, Reconstruction algorithms, Probability theory, Brain mapping, Brain, Bayesian inference
Point estimates, such as the maximum a posteriori (MAP) estimate, are commonly computed in image re-
construction tasks. However, such point estimates provide no information about the range of highly probable
solutions, namely the uncertainty in the computed estimate. Bayesian inference methods that seek to compute
the posterior probability distribution function (PDF) of the object can provide exactly this information, but
are generally computationally intractable. Markov Chain Monte Carlo (MCMC) methods, which avoid explicit
posterior computation by directly sampling from the PDF, require considerable expertise to run in a proper
way. This work investigates a computationally efficient variational Bayesian inference approach for computing
the posterior image variance with application to MRI. The methodology employs a sparse object prior model
that is consistent with the model assumed in most sparse reconstruction methods. The posterior variance map
generated by the proposed method provides valuable information that reveals how data-acquisition parameters
and the specification of the object prior affect the reliability of a reconstructed MAP image. The proposed
method is demonstrated by use of computer-simulated MRI data.
Photoacoustic computed tomography (PACT) and ultrasound computed tomography (USCT) are emerging modalities for breast imaging. As in all emerging imaging technologies, computer-simulation studies play a critically important role in developing and optimizing the designs of hardware and image reconstruction methods for PACT and USCT. Using computer-simulations, the parameters of an imaging system can be systematically and comprehensively explored in a way that is generally not possible through experimentation. When conducting such studies, numerical phantoms are employed to represent the physical properties of the patient or object to-be-imaged that influence the measured image data. It is highly desirable to utilize numerical phantoms that are realistic, especially when task-based measures of image quality are to be utilized to guide system design. However, most reported computer-simulation studies of PACT and USCT breast imaging employ simple numerical phantoms that oversimplify the complex anatomical structures in the human female breast. We develop and implement a methodology for generating anatomically realistic numerical breast phantoms from clinical contrast-enhanced magnetic resonance imaging data. The phantoms will depict vascular structures and the volumetric distribution of different tissue types in the breast. By assigning optical and acoustic parameters to different tissue structures, both optical and acoustic breast phantoms will be established for use in PACT and USCT studies.
We describe the ongoing development and performance of a high-pulse-energy wavelength-cycling laser system for three-dimensional optoacoustic tomography of the breast. Joule-level energies are desired for achieving the required penetration depths while maintaining safe fluence levels. Wavelength cycling provides a pulse sequence which repeatedly alternates between two wavelengths (approximately 756 and 797 nm) that provide differential imaging. This improves co-registration of captured differential images and quantification of blood oxygen saturation. New design features have been developed for and incorporated into a clinical prototype laser system, to improve efficacy and ease of use in the clinic. We describe the benefits of these features for operation with a clinical pilot optoacoustic / ultrasound dual-modality three-dimensional imaging system.
Proc. SPIE. 9708, Photons Plus Ultrasound: Imaging and Sensing 2016
KEYWORDS: Signal to noise ratio, Breast, Breast cancer, Tumors, Tissues, Imaging systems, Computing systems, Computer simulations, Tomography, Monte Carlo methods, Transducers, Photoacoustic tomography, Mammography, Tissue optics, Detection theory, Signal detection
The hybrid nature of optoacoustic tomography (OAT) brings together the advantages of both optical imaging and ultrasound imaging, making it a promising tool for breast cancer imaging. It is advocated in the modern imaging science literature to utilize objective, or task-based, measures of system performance to guide the optimization of hardware design and image reconstruction algorithms. In this work, we investigate this approach to assess the performance of OAT breast imaging systems. In particular, we apply principles from signal detection theory to compute the detectability of a simulated tumor at different depths within a breast, for two different system designs. The signal-to-noise ratio of the test statistic computed by a numerical observer is employed as the task-specific summary measure of system performance. A numerical breast model is employed that contains both slowly varying background and vessel structures as the background model, and superimpose a deterministic signal to emulate a tumor. This study demonstrates how signal detection performance of a numerical observer will vary as a function of signal depth and imaging system characteristics. The described methodology can be employed readily to systematically optimize other OAT imaging systems for tumor detection tasks.
Because optoacoustic tomography (OAT) can provide functional information based on hemoglobin contrast, it is a promising imaging modality for breast cancer diagnosis. Developing an effective OAT breast imaging system requires balancing multiple design constraints, which can be expensive and time-consuming. Therefore, computer- simulation studies are often conducted to facilitate this task. However, most existing computer-simulation studies of OAT breast imaging employ simple phantoms such as spheres or cylinders that over-simplify the complex anatomical structures in breasts, thus limiting the value of these studies in guiding real-world system design. In this work, we propose a method to generate realistic numerical breast phantoms for OAT research based on clinical magnetic resonance imaging (MRI) data. The phantoms include a skin layer that defines breast-air boundary, major vessel branches that affect light absorption in the breast, and fatty tissue and fibroglandular tissue whose acoustical heterogeneity perturbs acoustic wave propagation. By assigning realistic optical and acoustic parameters to different tissue types, we establish both optic and acoustic breast phantoms, which will be exported into standard data formats for cross-platform usage.
The Bayesian ideal observer (IO) has been widely advocated to guide hardware optimization. However, except for special cases, computation of the IO test statistic is computationally burdensome and requires an appropriate stochastic object model that may be difficult to determine in practice. Modern reconstruction methods, referred to as sparse reconstruction methods, exploit the fact that objects of interest typically possess sparse representations and have proven to be highly effective at reconstructing images from under-sampled measurement data. Moreover, in computed imaging approaches that employ compressive sensing concepts, imaging hardware and image reconstruction are innately coupled technologies. In this work, we propose a sparsity-driven IO (SD-IO) to guide the optimization of data acquisition parameters for modern computed imaging systems. The SD-IO employs a variational Bayesian inference method to estimate the posterior distribution and calculates an approximate likelihood ratio analytically as its test statistic. Since it assumes knowledge of low-level statistical properties of the object that are related to sparsity, the SD-IO exploits the same statistical information regarding the object that is utilized by highly effective sparse image reconstruction methods. Preliminary simulation results are presented to demonstrate the feasibility of the SD-IO calculation.
Optoacoustic tomography (OAT) is a promising imaging modality for human breast cancer imaging, with higher resolution and deeper penetration compared to other optical imaging modalities such as diffuse optical tomography or optical coherence tomography. It yields a resolution of 1 mm at depth up to 2 cm. But there is an inherent conflict between the limitations imposed on laser power and the need to adequately penetrate a substantial portion of the breast. To achieve sufficient penetration at every view angle, instead of illuminating the whole breast all at once, sometimes illumination is focused onto a small region of the breast and rotated along with the transducer array to cover the entire object. This paper evaluates the effect of this rotating partial illumination design on OAT image reconstruction. The optical process is simulated by conducting Monte Carlo simulations on a numerical phantom mimicking a real breast, with various specially designed illumination schemes. The acoustic process is simulated by incorporating the transducer's spatial impulse response. Iterative reconstruction is applied to estimate the OAT image. We conclude that rotating partial illumination introduces inconsistency into the system equation, and the degree of inconsistency determines the reconstruction quality.
In this work, we investigate a novel reconstruction method for laser-induced ultrasound computed tomography (USCT) breast imaging that circumvents limitations of existing methods that rely on ray-tracing. There is currently great interest in developing hybrid imaging systems that combine optoacoustic tomography (OAT) and USCT. There are two primary motivations for this: (1) the speed-of-sound (SOS) distribution reconstructed by USCT can provide complementary diagnostic information; and (2) the reconstructed SOS distribution can be incorporated in the OAT reconstruction algorithm to improve OAT image quality. However, image reconstruction in USCT remains challenging. The majority of existing approaches for USCT breast imaging involve ray-tracing to establish the imaging operator. This process is cumbersome and can lead to inaccuracies in the reconstructed SOS images in the presence of multiple ray-paths and/or shadow zones. To circumvent these problems, we implemented a partial differential equation-based Eulerian approach to USCT that was proposed in the mathematics literature but never investigated for medical imaging applications. This method operates by directly inverting the Eikonal equation without ray-tracing. A numerical implementation of this method was developed and compared to existing reconstruction methods for USCT breast imaging. We demonstrated the ability of the new method to reconstruct SOS maps from TOF data obtained by a hybrid OAT/USCT imager built by our team.
Photoacoustic computed tomography (PACT) provides structural and functional information when used in small animal brain imaging. Acoustic distortion caused by bone structures largely limits the deep brain image quality. In our work, we present ex vivo PACT images of freshly excised mouse brain, intending that can serve as a gold standard for future PACT in vivo studies on small animal brain imaging. Our results show that structures such as the striatum, hippocampus, ventricles, and cerebellum can be clearly di erentiated. An artery feature called the Circle of Willis, located at the bottom of the brain, can also be seen. These results indicate that if acoustic distortion can be accurately accounted for, PACT should be able to image the entire mouse brain with rich structural information.