Task-based Measures Of Image Quality Research Articles

Investigating the use of signal detection information in supervised learning-based image denoising with consideration of task-shift.

Recently, learning-based denoising methods that incorporate task-relevant information into the training procedure have been developed to enhance the utility of the denoised images. However, this line of research is relatively new and underdeveloped, and some fundamental issues remain unexplored. Our purpose is to yield insights into general issues related to these task-informed methods. This includes understanding the impact of denoising on objective measures of image quality (IQ) when the specified task at inference time is different from that employed for model training, a phenomenon we refer to as "task-shift." A virtual imaging test bed comprising a stylized computational model of a chest X-ray computed tomography imaging system was employed to enable a controlled and tractable study design. A canonical, fully supervised, convolutional neural network-based denoising method was purposely adopted to understand the underlying issues that may be relevant to a variety of applications and more advanced denoising or image reconstruction methods. Signal detection and signal detection-localization tasks under signal-known-statistically with background-known-statistically conditions were considered, and several distinct types of numerical observers were employed to compute estimates of the task performance. Studies were designed to reveal how a task-informed transfer-learning approach can influence the tradeoff between conventional and task-based measures of image quality within the context of the considered tasks. In addition, the impact of task-shift on these image quality measures was assessed. The results indicated that certain tradeoffs can be achieved such that the resulting AUC value was significantly improved and the degradation of physical IQ measures was statistically insignificant. It was also observed that introducing task-shift degrades the task performance as expected. The degradation was significant when a relatively simple task was considered for network training and observer performance on a more complex one was assessed at inference time. The presented results indicate that the task-informed training method can improve the observer performance while providing control over the tradeoff between traditional and task-based measures of image quality. The behavior of a task-informed model fine-tuning procedure was demonstrated, and the impact of task-shift on task-based image quality measures was investigated.

Applying Self-Supervised Learning to Image Quality Assessment in Chest CT Imaging.

Many new reconstruction techniques have been deployed to allow low-dose CT examinations. Such reconstruction techniques exhibit nonlinear properties, which strengthen the need for a task-based measure of image quality. The Hotelling observer (HO) is the optimal linear observer and provides a lower bound of the Bayesian ideal observer detection performance. However, its computational complexity impedes its widespread practical usage. To address this issue, we proposed a self-supervised learning (SSL)-based model observer to provide accurate estimates of HO performance in very low-dose chest CT images. Our approach involved a two-stage model combining a convolutional denoising auto-encoder (CDAE) for feature extraction and dimensionality reduction and a support vector machine for classification. To evaluate this approach, we conducted signal detection tasks employing chest CT images with different noise structures generated by computer-based simulations. We compared this approach with two supervised learning-based methods: a single-layer neural network (SLNN) and a convolutional neural network (CNN). The results showed that the CDAE-based model was able to achieve similar detection performance to the HO. In addition, it outperformed both SLNN and CNN when a reduced number of training images was considered. The proposed approach holds promise for optimizing low-dose CT protocols across scanner platforms.

Task-dependent estimability index to assess the quality of cardiac computed tomography angiography for quantifying coronary stenosis.

.Purpose: Quantifying stenosis in cardiac computed tomography angiography (CTA) images remains a difficult task, as image noise and cardiac motion can degrade image quality and distort underlying anatomic information. The purpose of this study was to develop a computational framework to objectively assess the precision of quantifying coronary stenosis in cardiac CTA.Approach: The framework used models of coronary vessels and plaques, asymmetric motion point spread functions, CT image blur (task-based modulation transfer functions) and noise (noise-power spectrums), and an automated maximum-likelihood estimator implemented as a matched template squared-difference operator. These factors were integrated into an estimability index () as a task-based measure of image quality in cardiac CTA. The index was applied to assess how well it can to predict the quality of 132 clinical cases selected from the Prospective Multicenter Imaging Study for Evaluation of Chest Pain trial. The cases were divided into two cohorts, high quality and low quality, based on clinical scores and the concordance of clinical evaluations of cases by experienced cardiac imagers. The framework was also used to ascertain protocol factors for CTA Biomarker initiative of the Quantitative Imaging Biomarker Alliance (QIBA).Results: The index categorized the patient datasets with an area under the curve of 0.985, an accuracy of 0.977, and an optimal threshold of 25.58 corresponding to a stenosis estimation precision (standard deviation) of 3.91%. Data resampling and training–test validation methods demonstrated stable classifier thresholds and receiver operating curve performance. The framework was successfully applicable to the QIBA objective.Conclusions: A computational framework to objectively quantify stenosis estimation task performance was successfully implemented and was reflective of clinical results in the context of a prominent clinical trial with diverse sites, readers, scanners, acquisition protocols, and patients. It also demonstrated the potential for prospective optimization of imaging protocols toward targeted precision and measurement consistency in cardiac CT images.

Approximating the Ideal Observer and Hotelling Observer for Binary Signal Detection Tasks by Use of Supervised Learning Methods.

It is widely accepted that the optimization of medical imaging system performance should be guided by task-based measures of image quality (IQ). Task-based measures of IQ quantify the ability of an observer to perform a specific task, such as detection or estimation of a signal (e.g., a tumor). For binary signal detection tasks, the Bayesian Ideal Observer (IO) sets an upper limit of observer performance and has been advocated for use in optimizing medical imaging systems and data-acquisition designs. Except in special cases, the determination of the IO test statistic is analytically intractable. Markov-chain Monte Carlo (MCMC) techniques can be employed to approximate the IO detection performance, but their reported applications have been limited to relatively simple object models. In cases where the IO test statistic is difficult to compute, the Hotelling Observer (HO) can be employed. To compute the HO test statistic, potentially large covariance matrices must be accurately estimated and subsequently inverted, which can present computational challenges. This paper investigates the supervised learning-based methodologies for approximating the IO and HO test statistics. Convolutional neural networks (CNNs) and single-layer neural networks (SLNNs) are employed to approximate the IO and HO test statistics, respectively. The numerical simulations were conducted for both signal-known-exactly (SKE) and signal-known-statistically (SKS) signal detection tasks. The considered background models include the lumpy object model and the clustered lumpy object model. The measurement noise models considered are Gaussian, Laplacian, and mixed Poisson-Gaussian. The performances of the supervised learning methods are assessed via receiver operating characteristic (ROC) analysis, and the results are compared to those produced by the use of traditional numerical methods or analytical calculations when feasible. The potential advantages of the proposed supervised learning approaches for approximating the IO and HO test statistics are discussed.

Reconstruction-Aware Imaging System Ranking by Use of a Sparsity-Driven Numerical Observer Enabled by Variational Bayesian Inference.

It is widely accepted that optimization of imaging system performance should be guided by task-based measures of image quality. It has been advocated that imaging hardware or data-acquisition designs should be optimized by use of an ideal observer that exploits full statistical knowledge of the measurement noise and class of objects to be imaged, without consideration of the reconstruction method. In practice, accurate and tractable models of the complete object statistics are often difficult to determine. Moreover, in imaging systems that employ compressive sensing concepts, imaging hardware and sparse image reconstruction are innately coupled technologies. In this paper, a sparsity-driven observer (SDO) that can be employed to optimize hardware by use of a stochastic object model describing object sparsity is described and investigated. The SDO 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 the SDO test statistic, computational tools developed recently for variational Bayesian inference with sparse linear models are adopted. The use of the SDO to rank data-acquisition designs in a stylized example as motivated by magnetic resonance imaging is demonstrated. This paper reveals that the SDO can produce rankings that are consistent with visual assessments of the reconstructed images but different from those produced by use of the traditionally employed Hotelling observer.

Learning-based stochastic object models for characterizing anatomical variations

It is widely known that the optimization of imaging systems based on objective, task-based measures of image quality via computer-simulation requires the use of a stochastic object model (SOM). However, the development of computationally tractable SOMs that can accurately model the statistical variations in human anatomy within a specified ensemble of patients remains a challenging task. Previously reported numerical anatomic models lack the ability to accurately model inter-patient and inter-organ variations in human anatomy among a broad patient population, mainly because they are established on image data corresponding to a few of patients and individual anatomic organs. This may introduce phantom-specific bias into computer-simulation studies, where the study result is heavily dependent on which phantom is used. In certain applications, however, databases of high-quality volumetric images and organ contours are available that can facilitate this SOM development. In this work, a novel and tractable methodology for learning a SOM and generating numerical phantoms from a set of volumetric training images is developed. The proposed methodology learns geometric attribute distributions (GAD) of human anatomic organs from a broad patient population, which characterize both centroid relationships between neighboring organs and anatomic shape similarity of individual organs among patients. By randomly sampling the learned centroid and shape GADs with the constraints of the respective principal attribute variations learned from the training data, an ensemble of stochastic objects can be created. The randomness in organ shape and position reflects the learned variability of human anatomy. To demonstrate the methodology, a SOM of an adult male pelvis is computed and examples of corresponding numerical phantoms are created.

Generation of anatomically realistic numerical phantoms for photoacoustic and ultrasonic breast imaging.

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.

SU-F-207-16: CT Protocols Optimization Using Model Observer

Purpose: To quantitatively evaluate the performance of different CT protocols using task-based measures of image quality. This work studies the task of size and the contrast estimation of different iodine concentration rods inserted in head- and body-sized phantoms using different imaging protocols. These protocols are designed to have the same dose level (CTDIvol) but using different X-ray tube voltage settings (kVp). Methods: Different concentrations of iodine objects inserted in a head size phantom and a body size phantom are imaged on a 64-slice commercial CT scanner. Scanning protocols with various tube voltages (80, 100, and 120 kVp) and current settings are selected, which output the same absorbed dose level (CTDIvol). Because the phantom design (size of the iodine objects, the air gap between the inserted objects and the phantom) is not ideal for a model observer study, the acquired CT images are used to generate simulation images with four different sizes and five different contracts iodine objects. For each type of the objects, 500 images (100 × 100 pixels) are generated for the observer study. The observer selected in this study is the channelized scanning linear observer which could be applied to estimate the size and the contrast. The figure of merit used is the correct estimation ratio. The mean and the variance are estimated by the shuffle method. Results: The results indicate that the protocols with 100 kVp tube voltage setting provides the best performance for iodine insert size and contrast estimation for both head and body phantom cases. Conclusion: This work presents a practical and robust quantitative approach using channelized scanning linear observer to study contrast and size estimation performance from different CT protocols. Different protocols at same CTDIvol setting could Result in different image quality performance. The relationship between the absorbed dose and the diagnostic image quality is not linear.

RADIANCE AND PHOTON NOISE: Imaging in geometrical optics, physical optics, quantum optics and radiology.

A fundamental way of describing a photon-limited imaging system is in terms of a Poisson random process in spatial, angular and wavelength variables. The mean of this random process is the spectral radiance. The principle of conservation of radiance then allows a full characterization of the noise in the image (conditional on viewing a specified object). To elucidate these connections, we first review the definitions and basic properties of radiance as defined in terms of geometrical optics, radiology, physical optics and quantum optics. The propagation and conservation laws for radiance in each of these domains are reviewed. Then we distinguish four categories of imaging detectors that all respond in some way to the incident radiance, including the new category of photon-processing detectors. The relation between the radiance and the statistical properties of the detector output is discussed and related to task-based measures of image quality and the information content of a single detected photon.

Assessing image quality and dose reduction of a new x-ray computed tomography iterative reconstruction algorithm using model observers.

A number of different techniques have been developed to reduce radiation dose in x-ray computed tomography (CT) imaging. In this paper, the authors will compare task-based measures of image quality of CT images reconstructed by two algorithms: conventional filtered back projection (FBP), and a new iterative reconstruction algorithm (IR). To assess image quality, the authors used the performance of a channelized Hotelling observer acting on reconstructed image slices. The selected channels are dense difference Gaussian channels (DDOG).A body phantom and a head phantom were imaged 50 times at different dose levels to obtain the data needed to assess image quality. The phantoms consisted of uniform backgrounds with low contrast signals embedded at various locations. The tasks the observer model performed included (1) detection of a signal of known location and shape, and (2) detection and localization of a signal of known shape. The employed DDOG channels are based on the response of the human visual system. Performance was assessed using the areas under ROC curves and areas under localization ROC curves. For signal known exactly (SKE) and location unknown/signal shape known tasks with circular signals of different sizes and contrasts, the authors' task-based measures showed that a FBP equivalent image quality can be achieved at lower dose levels using the IR algorithm. For the SKE case, the range of dose reduction is 50%-67% (head phantom) and 68%-82% (body phantom). For the study of location unknown/signal shape known, the dose reduction range can be reached at 67%-75% for head phantom and 67%-77% for body phantom case. These results suggest that the IR images at lower dose settings can reach the same image quality when compared to full dose conventional FBP images. The work presented provides an objective way to quantitatively assess the image quality of a newly introduced CT IR algorithm. The performance of the model observers using the IR images was always higher than that seen using the FBP images in the authors' SKE and SKE location unknown detection tasks. To achieve a FBP-equivalent image quality in CT systems, the authors can lower the radiation dose by using this IR image reconstruction algorithm. Further studies are warranted using clinical data and human observer to validate these results for more complicated and realistic tasks.

Fisher information and surrogate figures of merit for the task-based assessment of image quality

Fisher information can be used as a surrogate for task-based measures of image quality based on ideal observer performance. A new and improved derivation of the Fisher information approximation for ideal-observer detectability is provided. This approximation depends only on the presence of a weak signal and does not depend on Gaussian statistical assumptions. This is also not an asymptotic result and therefore applies to imaging, where there is typically only one dataset, albeit a large one. Applications to statistical mixture models for image data are presented. For Gaussian and Poisson mixture models the results are used to connect reconstruction error with ideal-observer detection performance. When the task is the estimation of signal parameters of a weak signal, the ensemble mean squared error of the posterior mean estimator can also be expanded in powers of the signal amplitude. There is no linear term in this expansion, and it is shown that the quadratic term involves a Fisher information kernel that generalizes the standard Fisher information. Applications to imaging mixture models reveal a close connection between ideal performance on these estimation tasks and detection tasks for the same signals. Finally, for tasks that combine detection and estimation, we may also define a detectability that measures performance on this combined task and an ideal observer that maximizes this detectability. This detectability may also be expanded in powers of the signal amplitude, and the quadratic term again involves the Fisher information kernel. Applications of this approximation to imaging mixture models show a relation with the pure detection and pure estimation tasks for the same signals.

ADAPTIVE SMALL-ANIMAL SPECT/CT.

We are exploring the concept of adaptive multimodality imaging, a form of non-linear optimization where the imaging configuration is automatically adjusted in response to the object. Preliminary studies suggest that substantial improvement in objective, task-based measures of image quality can result. We describe here our work to add motorized adjustment capabilities and a matching CT to our existing FastSPECT II system to form an adaptive small-animal SPECT/CT.

A Task-Based Approach to Adaptive and Multimodality Imaging: Computation techniques are proposed for figures-of-merit to establish feasibility and optimize use of multiple imaging systems for disease diagnosis and treatment-monitoring.

Multimodality imaging is becoming increasingly important in medical imaging. Since the motivation for combining multiple imaging modalities is generally to improve diagnostic or prognostic accuracy, the benefits of multimodality imaging cannot be assessed through the display of example images. Instead, we must use objective, task-based measures of image quality to draw valid conclusions about system performance. In this paper, we will present a general framework for utilizing objective, task-based measures of image quality in assessing multimodality and adaptive imaging systems. We introduce a classification scheme for multimodality and adaptive imaging systems and provide a mathematical description of the imaging chain along with block diagrams to provide a visual illustration. We show that the task-based methodology developed for evaluating single-modality imaging can be applied, with minor modifications, to multimodality and adaptive imaging. We discuss strategies for practical implementing of task-based methods to assess and optimize multimodality imaging systems.

Experimental determination of object statistics from noisy images.

Modern imaging systems rely on complicated hardware and sophisticated image-processing methods to produce images. Owing to this complexity in the imaging chain, there are numerous variables in both the hardware and the software that need to be determined. We advocate a task-based approach to measuring and optimizing image quality in which one analyzes the ability of an observer to perform a task. Ideally, a task-based measure of image quality would account for all sources of variation in the imaging system, including object variability. Often, researchers ignore object variability even though it is known to have a large effect on task performance. The more accurate the statistical description of the objects, the more believable the task-based results will be. We have developed methods to fit statistical models of objects, using only noisy image data and a well-characterized imaging system. The results of these techniques could eventually be used to optimize both the hardware and the software components of imaging systems.

Application of task-based measures of image quality to optimization and evaluation of three-dimensional reconstruction-based compensation methods in myocardial perfusion SPECT.

In this paper, we apply the channelized Hotelling observer (CHO) using a defect detection task to the optimization and evaluation of three-dimensional iterative reconstruction-based compensation methods for myocardial perfusion single-photon emission computed tomography (SPECT). We used a population of 24 mathematical cardiac-torso phantoms that realistically model the activity and attenuation distribution in three classes of patients: females, and males with flat diaphragms and raised diaphragms. Projection data were generated and subsequently reconstructed using methods based on the ordered subsets-expectation maximization (OSEM) algorithm. The methods evaluated included compensation for attenuation, detector response blurring, and scatter in various combinations. We applied the CHO to optimize the number of iterations for OSEM and the cutoff frequency and order of a three-dimensional postreconstruction Butterworth filter. Using the optimal parameters, we then compared the compensation methods. The index of comparison in these studies was the area under the receiver operating characteristics curve (AUC) for the CHO. We found that attenuation compensation with either detector response or scatter compensation gave statistically significant increases in the AUC compared to attenuation compensation alone. The greatest increase in the AUC occurred when all three compensations were applied. These results indicate that compensation for detector response and scatter, in addition to attenuation compensation, will improve defect detectability in myocardial SPECT images.