SPSubok Park

Subok Park

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Subok Park has a total of 18 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Model observer design for multi-signal detection in the presence of anatomical noise

    by Gezheng Wen, Mia K Markey, Subok Park

    As psychophysical studies are resource-intensive to conduct, model observers are commonly used to assess and optimize medical imaging quality. Model observers are typically designed to detect at most one signal. However, in clinical practice, there may be multiple abnormalities in a single image set (e.g. multifocal multicentric (MFMC) breast cancer), which can impact treatment planning. Prevalence of signals can be different across anatomical regions, and human observers do not know the number or location of signals a priori . As new imaging techniques have the potential to improve multiple-signal detection (e.g. digital breast tomosynthesis may be more effective for diagnosis of MFMC than mammography), image quality assessment approaches addressing such tasks are needed. In this study, we present a model observer to detect multiple signals in an image dataset. A novel implementation of partial least squares (PLS) was developed to estimate different sets of efficient channels directly from the images. The PLS channels are adaptive to the characteristics of signals and the background, and they capture the interactions among signal locations. Corresponding linear decision templates are employed to generate both image-level and location-specific scores on the presence of signals. Our results show that: (1) the model observer can achieve high performance with a reasonably small number of channels; (2) the model observer with PLS channels outperforms that with benchmark modified Laguerre–Gauss channels, especially when realistic signal shapes and complex background statistics are involved; (3) the tasks of clinical interest, and other constraints such as sample size would alter the optimal design of the model observer.


  • A virtual trial framework for quantifying the detectability of masses in breast tomosynthesis projection data

    by Stefano Young, Predrag R. Bakic, Kyle J. Myers, Robert J. Jennings, Subok Park

    Purpose:: Digital breast tomosynthesis (DBT) is a promising breast cancer screening tool that has already begun making inroads into clinical practice. However, there is ongoing debate over how to quantitatively evaluate and optimize these systems, because different definitions of image quality can lead to different optimal design strategies. Powerful and accurate tools are desired to extend our understanding of DBT system optimization and validate published design principles.Methods:: The authors developed a virtual trial framework for task‐specific DBT assessment that uses digital phantoms, open‐source x‐ray transport codes, and a projection‐space, spatial‐domain observer model for quantitative system evaluation. The authors considered evaluation of reconstruction algorithms as a separate problem and focused on the information content in the raw, unfiltered projection images. Specifically, the authors investigated the effects of scan angle and number of angular projections on detectability of a small (3 mm diameter) signal embedded in randomly‐varying anatomical backgrounds. Detectability was measured by the area under the receiver‐operating characteristic curve (AUC). Experiments were repeated for three test cases where the detectability‐limiting factor was anatomical variability, quantum noise, or electronic noise. The authors also juxtaposed the virtual trial framework with other published studies to illustrate its advantages and disadvantages.Results:: The large number of variables in a virtual DBT study make it difficult to directly compare different authors’ results, so each result must be interpreted within the context of the specific virtual trial framework. The following results apply to 25% density phantoms with 5.15 cm compressed thickness and 500 μm3 voxels (larger 500 μm2 detector pixels were used to avoid voxel‐edge artifacts): 1. For raw, unfiltered projection images in the anatomical‐variability‐limited regime, AUC appeared to remain constant or increase slightly with scan angle. 2. In the same regime, when the authors fixed the scan angle, AUC increased asymptotically with the number of projections. The threshold number of projections for asymptotic AUC performance depended on the scan angle. In the quantum‐ and electronic‐noise dominant regimes, AUC behaviors as a function of scan angle and number of projections sometimes differed from the anatomy‐limited regime. For example, with a fixed scan angle, AUC generally decreased with the number of projections in the electronic‐noise dominant regime. These results are intended to demonstrate the capabilities of the virtual trial framework, not to be used as optimization rules for DBT.Conclusions:: The authors have demonstrated a novel simulation framework and tools for evaluating DBT systems in an objective, task‐specific manner. This framework facilitates further investigation of image quality tradeoffs in DBT.


  • Investigating the feasibility of using partial least squares as a method of extracting salient information for the evaluation of digital breast tomosynthesis

    by George Z. Zhang, Kyle J. Myers, Subok Park

    Digital breast tomosynthesis (DBT) has shown promise for improving the detection of breast cancer, but it has not yet been fully optimized due to a large space of system parameters to explore. A task-based statistical approach1 is a rigorous method for evaluating and optimizing this promising imaging technique with the use of optimal observers such as the Hotelling observer (HO). However, the high data dimensionality found in DBT has been the bottleneck for the use of a task-based approach in DBT evaluation. To reduce data dimensionality while extracting salient information for performing a given task, efficient channels have to be used for the HO. In the past few years, 2D Laguerre-Gauss (LG) channels, which are a complete basis for stationary backgrounds and rotationally symmetric signals, have been utilized for DBT evaluation2, 3 . But since background and signal statistics from DBT data are neither stationary nor rotationally symmetric, LG channels may not be efficient in providing reliable performance trends as a function of system parameters. Recently, partial least squares (PLS) has been shown to generate efficient channels for the Hotelling observer in detection tasks involving random backgrounds and signals.4 In this study, we investigate the use of PLS as a method for extracting salient information from DBT in order to better evaluate such systems.


  • A statistical, task‐based evaluation method for three‐dimensional x‐ray breast imaging systems using variable‐background phantoms

    by Subok Park, Robert Jennings, Haimo Liu, Aldo Badano, Kyle Myers

    Purpose: For the last few years, development and optimization of three‐dimensional (3D) x‐ray breast imaging systems, such as digital breast tomosynthesis (DBT) and computed tomography, have drawn much attention from the medical imaging community, either academia or industry. However, there is still much room for understanding how to best optimize and evaluate the devices over a large space of many different system parameters and geometries. Current evaluation methods, which work well for 2D systems, do not incorporate the depth information from the 3D imaging systems. Therefore, it is critical to develop a statistically sound evaluation method to investigate the usefulness of inclusion of depth and background‐variability information into the assessment and optimization of the 3D systems.Methods: In this paper, we present a mathematical framework for a statistical assessment of planar and 3D x‐ray breast imaging systems. Our method is based on statistical decision theory, in particular, making use of the ideal linear observer called the Hotelling observer. We also present a physical phantom that consists of spheres of different sizes and materials for producing an ensemble of randomly varying backgrounds to be imaged for a given patient class. Lastly, we demonstrate our evaluation method in comparing laboratory mammography and three‐angle DBT systems for signal detection tasks using the phantom's projection data. We compare the variable phantom case to that of a phantom of the same dimensions filled with water, which we call the uniform phantom, based on the performance of the Hotelling observer as a function of signal size and intensity.Results: Detectability trends calculated using the variable and uniform phantom methods are different from each other for both mammography and DBT systems.Conclusions: Our results indicate that measuring the system's detection performance with consideration of background variability may lead to differences in system performance estimates and comparisons. For the assessment of 3D systems, to accurately determine trade offs between image quality and radiation dose, it is critical to incorporate randomness arising from the imaging chain including background variability into system performance calculations.


  • Open Access

    Efficient estimation of ideal-observer performance in classification tasks involving high-dimensional complex backgrounds

    by Subok Park, Eric Clarkson

    The Bayesian ideal observer is optimal among all observers and sets an absolute upper bound for the performance of any observer in classification tasks [ Van Trees, Detection, Estimation, and Modulation Theory, Part I (Academic, 1968). ]. Therefore, the ideal observer should be used for objective image quality assessment whenever possible. However, computation of ideal-observer performance is difficult in practice because this observer requires the full description of unknown, statistical properties of high-dimensional, complex data arising in real life problems. Previously, Markov-chain Monte Carlo (MCMC) methods were developed by Kupinski et al. [J. Opt. Soc. Am. A 20, 430(2003) ] and by Park et al. [J. Opt. Soc. Am. A 24, B136 (2007) and IEEE Trans. Med. Imaging 28, 657 (2009) ] to estimate the performance of the ideal observer and the channelized ideal observer (CIO), respectively, in classification tasks involving non-Gaussian random backgrounds. However, both algorithms had the disadvantage of long computation times. We propose a fast MCMC for real-time estimation of the likelihood ratio for the CIO. Our simulation results show that our method has the potential to speed up ideal-observer performance in tasks involving complex data when efficient channels are used for the CIO.


  • Singular Vectors of a Linear Imaging System as Efficient Channels for the Bayesian Ideal Observer

    by Subok Park, Joel M. Witten, Kyle J. Myers

    The Bayesian ideal observer provides an absolute upper bound for diagnostic performance of an imaging system and hence should be used for the assessment of image quality whenever possible. However, computation of ideal-observer performance in clinical tasks is difficult since the probability density functions of the data required for this observer are often unknown in tasks involving realistic, complex backgrounds. Moreover, the high dimensionality of the integrals that need to be calculated for the observer makes the computation more difficult. The ideal observer constrained to a set of channels, which we call a channelized-ideal observer (CIO), can reduce the dimensionality of the problem. These channels are called efficient if the CIO can approximate ideal-observer performance. In this paper, we propose a method to choose efficient channels for the ideal observer based on a singular value decomposition of a linear imaging system. As a demonstration, we test our method on detection tasks using non-Gaussian lumpy backgrounds and signals of Gaussian and elliptical profiles. Our simulation results show that singular vectors associated with either the background or the signal are highly efficient for the ideal observer for detecting both types of signals. In addition, this CIO outperforms a channelized-Hotelling observer with the same channels.


  • Incorporating Human Contrast Sensitivity in Model Observers for Detection Tasks

    by Subok Park, Aldo Badano, Brandon D. Gallas, Kyle J. Myers

    Contrast sensitivity of the human visual system is a characteristic that can adversely affect human performance in detection tasks. In this paper, we propose a method for incorporating human contrast sensitivity in anthropomorphic model observers. In our method, we model human contrast sensitivity using the Barten model with the mean luminance of a region of interest centered at the signal location. In addition, one free parameter is varied to control the effect of the contrast sensitivity on the model observer's performance. We investigate our model of human contrast sensitivity in a channelized-Hotelling observer (CHO) with difference-of-Gaussian channels. We call the CHO incorporating the contrast sensitivity a contrast-sensitive CHO (CS-CHO). The human data from a psychophysical study by Park are used for comparing the performance of the CS-CHO to human performance. That study used Gaussian signals with six different signal intensities in non-Gaussian lumpy backgrounds. A value of the free parameter is chosen to match the performance of the CS-CHO to the mean human performance only at the strongest signal. Results show that the CS-CHO with the chosen value of the free parameter predicts the mean human performance at the five lower signal intensities. Our results show that the CS-CHO predicts human performance well as a function of signal intensity.


  • Singular vectors of a linear imaging system as efficient channels for the ideal observer in detection tasks involving non-Gaussian distributed lumpy images

    by Joel M. Witten, Subok Park, Kyle J. Myers

    The Bayesian ideal observer sets an upper bound for diagnostic performance of an imaging system in binary detection tasks. Thus, this observer should be used for image quality assessment whenever possible. However, it is difficult to compute ideal-observer performance because the probability density functions of the data, required for the observer, are often unknown in tasks involving complex backgrounds. Furthermore, the dimension of the integrals that need to be calculated for the observer is huge. To attempt to reduce the dimensionality of the problem, and yet still approximate ideal-observer performance, a channelized-ideal observer (CIO) with Laguerre-Gauss channels was previously investigated for detecting a Gaussian signal at a known location in non-Gaussian lumpy images. While the CIO with Laguerre-Gauss channels had, in some cases, approximated ideal-observer performance, there was still a gap between the mean performance of the ideal observer and the CIO. Moreover, it is not clear how to choose efficient channels for the ideal observer. In the current work, we investigate the use of singular vectors of a linear imaging system as efficient channels for the ideal observer in the same tasks. Singular value decomposition of the imaging system is performed to obtain its singular vectors. Singular vectors most relevant to the signal and background images are chosen as candidate channels. Results indicate that the singular vectors are not only more efficient than Laguerre-Gauss channels, but are also highly efficient for the ideal observer. The results further demonstrate that singular vectors strongly associated with the signal-only image are the most efficient channels.

  • Channelized-ideal observer using Laguerre-Gauss channels in detection tasks involving non-Gaussian distributed lumpy backgrounds and a Gaussian signal

    by Subok Park, Harrison H. Barrett, Eric Clarkson, Matthew A. Kupinski, Kyle J. Myers

    We investigate a channelized-ideal observer (CIO) with Laguerre–Gauss (LG) channels to approximate ideal-observer performance in detection tasks involving non-Gaussian distributed lumpy backgrounds and a Gaussian signal. A Markov-chain Monte Carlo approach is employed to determine the performance of both the ideal observer and the CIO using a large number of LG channels. Our results indicate that the CIO with LG channels can approximate ideal-observer performance within error bars, depending on the imaging system, object, and channel parameters. The CIO also outperforms a channelized-Hotelling observer using the same channels. In addition, an alternative approach for estimating the CIO is investigated. This approach makes use of the characteristic functions of channelized data and employs an approximation method to the area under the receiver operating characteristic curve. The alternative approach provides good estimates of the performance of the CIO with five LG channels. However, for large channel cases, more efficient computational methods need to be developed for the CIO to become useful in practice.


  • Efficiency of the human observer for detecting a Gaussian signal at a known location in non-Gaussian distributed lumpy backgrounds

    by Subok Park, Brandon D. Gallas, Aldo Badano, Nicholas A. Petrick, Kyle J. Myers

    A previous study [J. Opt. Soc. Am. A22, 3 (2005)] has shown that human efficiency for detecting a Gaussian signal at a known location in non-Gaussian distributed lumpy backgrounds is approximately 4%. This human efficiency is much less than the reported 40% efficiency that has been documented for Gaussian-distributed lumpy backgrounds [J. Opt. Soc. Am. A16, 694 (1999) and J. Opt. Soc. Am. A18, 473 (2001)]. We conducted a psychophysical study with a number of changes, specifically in display-device calibration and data scaling, from the design of the aforementioned study. Human efficiency relative to the ideal observer was found again to be approximately 5%. Our variance analysis indicates that neither scaling nor display made a statistically significant difference in human performance for the task. We conclude that the non-Gaussian distributed lumpy background is a major factor in our low human-efficiency results.