Theoretical study of lesion detectability of MAP reconstruction using computer observers

PET image reconstruction using weighted nuclear norm maximization and deep learning prior

The ill-posed Positron emission tomography (PET) reconstruction problem usually results in limited resolution and significant noise. Recently, deep neural networks have been incorporated into PET iterative reconstruction framework to improve the image quality. In this paper, we propose a new neural network-based iterative reconstruction method by using weighted nuclear norm (WNN) maximization, which aims to recover the image details in the reconstruction process. The novelty of our method is the application of WNN maximization rather than WNN minimization in PET image reconstruction. Meanwhile, a neural network is used to control the noise originated from WNN maximization. Our method is evaluated on simulated and clinical datasets. The simulation results show that the proposed approach outperforms state-of-the-art neural network-based iterative methods by achieving the best contrast/noise tradeoff with a remarkable contrast improvement on the lesion contrast recovery. The study on clinical datasets also demonstrates that our method can recover lesions of different sizes while suppressing noise in various low-dose PET image reconstruction tasks. Our code is available athttps://github.com/Kuangxd/PETReconstruction.

Lesion detection and quantification performance of the Tachyon-I time-of-flight PET scanner: phantom and human studies

The first generation Tachyon PET (Tachyon-I) is a demonstration single-ring PET scanner that reaches a coincidence timing resolution of 314 ps using LSO scintillator crystals coupled to conventional photomultiplier tubes. The objective of this study was to quantify the improvement in both lesion detection and quantification performance resulting from the improved time-of-flight (TOF) capability of the Tachyon-I scanner. We developed a quantitative TOF image reconstruction method for the Tachyon-I and evaluated its TOF gain for lesion detection and quantification. Scans of either a standard NEMA torso phantom or healthy volunteers were used as the normal background data. Separately scanned point source and sphere data were superimposed onto the phantom or human data after accounting for the object attenuation. We used the bootstrap method to generate multiple independent noisy datasets with and without a lesion present. The signal-to-noise ratio (SNR) of a channelized hotelling observer (CHO) was calculated for each lesion size and location combination to evaluate the lesion detection performance. The bias versus standard deviation trade-off of each lesion uptake was also calculated to evaluate the quantification performance. The resulting CHO-SNR measurements showed improved performance in lesion detection with better timing resolution. The detection performance was also dependent on the lesion size and location, in addition to the background object size and shape. The results of bias versus noise trade-off showed that the noise (standard deviation) reduction ratio was about 1.1-1.3 over the TOF 500 ps and 1.5-1.9 over the non-TOF modes, similar to the SNR gains for lesion detection. In conclusion, this Tachyon-I PET study demonstrated the benefit of improved time-of-flight capability on lesion detection and ROI quantification for both phantom and human subjects.

Theoretical Analysis of Penalized Maximum-Likelihood Patlak Parametric Image Reconstruction in Dynamic PET for Lesion Detection

Detecting cancerous lesions is a major clinical application of emission tomography. In a previous work, we studied penalized maximum-likelihood (PML) image reconstruction for lesion detection in static PET. Here we extend our theoretical analysis of static PET reconstruction to dynamic PET. We study both the conventional indirect reconstruction and direct reconstruction for Patlak parametric image estimation. In indirect reconstruction, Patlak parametric images are generated by first reconstructing a sequence of dynamic PET images, and then performing Patlak analysis on the time activity curves (TACs) pixel-by-pixel. In direct reconstruction, Patlak parametric images are estimated directly from raw sinogram data by incorporating the Patlak model into the image reconstruction procedure. PML reconstruction is used in both the indirect and direct reconstruction methods. We use a channelized Hotelling observer (CHO) to assess lesion detectability in Patlak parametric images. Simplified expressions for evaluating the lesion detectability have been derived and applied to the selection of the regularization parameter value to maximize detection performance. The proposed method is validated using computer-based Monte Carlo simulations. Good agreements between the theoretical predictions and the Monte Carlo results are observed. Both theoretical predictions and Monte Carlo simulation results show the benefit of the indirect and direct methods under optimized regularization parameters in dynamic PET reconstruction for lesion detection, when compared with the conventional static PET reconstruction.

Prospective regularization design in prior-image-based reconstruction

Prior-image-based reconstruction (PIBR) methods leveraging patient-specific anatomical information from previous imaging studies and/or sequences have demonstrated dramatic improvements in dose utilization and image quality for low-fidelity data. However, a proper balance of information from the prior images and information from the measurements is required (e.g. through careful tuning of regularization parameters). Inappropriate selection of reconstruction parameters can lead to detrimental effects including false structures and failure to improve image quality. Traditional methods based on heuristics are subject to error and sub-optimal solutions, while exhaustive searches require a large number of computationally intensive image reconstructions. In this work, we propose a novel method that prospectively estimates the optimal amount of prior image information for accurate admission of specific anatomical changes in PIBR without performing full image reconstructions. This method leverages an analytical approximation to the implicitly defined PIBR estimator, and introduces a predictive performance metric leveraging this analytical form and knowledge of a particular presumed anatomical change whose accurate reconstruction is sought. Additionally, since model-based PIBR approaches tend to be space-variant, a spatially varying prior image strength map is proposed to optimally admit changes everywhere in the image (eliminating the need to know change locations a priori). Studies were conducted in both an ellipse phantom and a realistic thorax phantom emulating a lung nodule surveillance scenario. The proposed method demonstrated accurate estimation of the optimal prior image strength while achieving a substantial computational speedup (about a factor of 20) compared to traditional exhaustive search. Moreover, the use of the proposed prior strength map in PIBR demonstrated accurate reconstruction of anatomical changes without foreknowledge of change locations in phantoms where the optimal parameters vary spatially by an order of magnitude or more. In a series of studies designed to explore potential unknowns associated with accurate PIBR, optimal prior image strength was found to vary with attenuation differences associated with anatomical change but exhibited only small variations as a function of the shape and size of the change. The results suggest that, given a target change attenuation, prospective patient-, change-, and data-specific customization of the prior image strength can be performed to ensure reliable reconstruction of specific anatomical changes.

Regularization designs for uniform spatial resolution and noise properties in statistical image reconstruction for 3-D X-ray CT

Statistical image reconstruction methods for X-ray computed tomography (CT) provide improved spatial resolution and noise properties over conventional filtered back-projection (FBP) reconstruction, along with other potential advantages such as reduced patient dose and artifacts. Conventional regularized image reconstruction leads to spatially variant spatial resolution and noise characteristics because of interactions between the system models and the regularization. Previous regularization design methods aiming to solve such issues mostly rely on circulant approximations of the Fisher information matrix that are very inaccurate for undersampled geometries like short-scan cone-beam CT. This paper extends the regularization method proposed in to 3-D cone-beam CT by introducing a hypothetical scanning geometry that helps address the sampling properties. The proposed regularization designs were compared with the original method in with both phantom simulation and clinical reconstruction in 3-D axial X-ray CT. The proposed regularization methods yield improved spatial resolution or noise uniformity in statistical image reconstruction for short-scan axial cone-beam CT.

Task-based detectability in CT image reconstruction by filtered backprojection and penalized likelihood estimation

PURPOSE

Nonstationarity is an important aspect of imaging performance in CT and cone-beam CT (CBCT), especially for systems employing iterative reconstruction. This work presents a theoretical framework for both filtered-backprojection (FBP) and penalized-likelihood (PL) reconstruction that includes explicit descriptions of nonstationary noise, spatial resolution, and task-based detectability index. Potential utility of the model was demonstrated in the optimal selection of regularization parameters in PL reconstruction.

METHODS

Analytical models for local modulation transfer function (MTF) and noise-power spectrum (NPS) were investigated for both FBP and PL reconstruction, including explicit dependence on the object and spatial location. For FBP, a cascaded systems analysis framework was adapted to account for nonstationarity by separately calculating fluence and system gains for each ray passing through any given voxel. For PL, the point-spread function and covariance were derived using the implicit function theorem and first-order Taylor expansion according to Fessler ["Mean and variance of implicitly defined biased estimators (such as penalized maximum likelihood): Applications to tomography," IEEE Trans. Image Process. 5(3), 493-506 (1996)]. Detectability index was calculated for a variety of simple tasks. The model for PL was used in selecting the regularization strength parameter to optimize task-based performance, with both a constant and a spatially varying regularization map.

RESULTS

Theoretical models of FBP and PL were validated in 2D simulated fan-beam data and found to yield accurate predictions of local MTF and NPS as a function of the object and the spatial location. The NPS for both FBP and PL exhibit similar anisotropic nature depending on the pathlength (and therefore, the object and spatial location within the object) traversed by each ray, with the PL NPS experiencing greater smoothing along directions with higher noise. The MTF of FBP is isotropic and independent of location to a first order approximation, whereas the MTF of PL is anisotropic in a manner complementary to the NPS. Task-based detectability demonstrates dependence on the task, object, spatial location, and smoothing parameters. A spatially varying regularization "map" designed from locally optimal regularization can improve overall detectability beyond that achievable with the commonly used constant regularization parameter.

CONCLUSIONS

Analytical models for task-based FBP and PL reconstruction are predictive of nonstationary noise and resolution characteristics, providing a valuable framework for understanding and optimizing system performance in CT and CBCT.

Quantifying the Importance of the Statistical Assumption in Statistical X-ray CT Image Reconstruction

Statistical image reconstruction (SIR) is a promising approach to reducing radiation dose in clinical computerized tomography (CT) scans. Clinical CT scanners use energy-integrating detectors. The CT signal follows a compound Poisson distribution, its probability density function (PDF) does not have an analytical form hence cannot be used in an SIR method. The goal of this work is to quantify the effects of using an approximate statistical assumption in SIR methods for clinical CT applications. We apply a pseudo-Ideal Observer (pIO) to simulated CT projection data of the fanbeam geometry at different dose levels. The simulation models the polychromatic X-ray tube spectrum, the effects of the bowtie filter, and the energy-integrating detectors. The pIO uses a pseudo likelihood function (pLF) to calculate the pseudo likelihood ratio, which is the decision variable used by the pIO in a binary detection task. The pLF is an approximation to the true LF of the underlying data. The pIO has inferior performance than the IO unless the pLF coincides with the LF; this performance difference quantifies the closeness between the pseudo likelihood and the exact one. Using lesion detectability in a signal known exactly, background known exactly binary detection task as a figure-of-merit, our results show that at down to 0.1% of a reference tube current level I0, the pIO that uses a Poisson approximation, or a matched variance Gaussian approximation in either the transmission or the line integral domain, achieves 99% the performance of the IO. The constant variance Gaussian approximation has only 70%-80% of the IO performance. At tube currents lower than 0.1% I0, the performance difference is more substantial. We conclude that at current clinical dose levels, it is important to account for the mean-dependent variance in CT projection data in SIR problem formulation, the exact PDF of the CT signal is not as important.

Quantitative statistical methods for image quality assessment

Quantitative measures of image quality and reliability are critical for both qualitative interpretation and quantitative analysis of medical images. While, in theory, it is possible to analyze reconstructed images by means of Monte Carlo simulations using a large number of noise realizations, the associated computational burden makes this approach impractical. Additionally, this approach is less meaningful in clinical scenarios, where multiple noise realizations are generally unavailable. The practical alternative is to compute closed-form analytical expressions for image quality measures. The objective of this paper is to review statistical analysis techniques that enable us to compute two key metrics: resolution (determined from the local impulse response) and covariance. The underlying methods include fixed-point approaches, which compute these metrics at a fixed point (the unique and stable solution) independent of the iterative algorithm employed, and iteration-based approaches, which yield results that are dependent on the algorithm, initialization, and number of iterations. We also explore extensions of some of these methods to a range of special contexts, including dynamic and motion-compensated image reconstruction. While most of the discussed techniques were developed for emission tomography, the general methods are extensible to other imaging modalities as well. In addition to enabling image characterization, these analysis techniques allow us to control and enhance imaging system performance. We review practical applications where performance improvement is achieved by applying these ideas to the contexts of both hardware (optimizing scanner design) and image reconstruction (designing regularization functions that produce uniform resolution or maximize task-specific figures of merit).

SPECT system optimization against a discrete parameter space

In this paper, we present an analytical approach for optimizing the design of a static SPECT system or optimizing the sampling strategy with a variable/adaptive SPECT imaging hardware against an arbitrarily given set of system parameters. This approach has three key aspects. First, it is designed to operate over a discretized system parameter space. Second, we have introduced an artificial concept of virtual detector as the basic building block of an imaging system. With a SPECT system described as a collection of the virtual detectors, one can convert the task of system optimization into a process of finding the optimum imaging time distribution (ITD) across all virtual detectors. Thirdly, the optimization problem (finding the optimum ITD) could be solved with a block-iterative approach or other nonlinear optimization algorithms. In essence, the resultant optimum ITD could provide a quantitative measure of the relative importance (or effectiveness) of the virtual detectors and help to identify the system configuration or sampling strategy that leads to an optimum imaging performance. Although we are using SPECT imaging as a platform to demonstrate the system optimization strategy, this development also provides a useful framework for system optimization problems in other modalities, such as positron emission tomography and x-ray computed tomography (Moore et al (2009 IEEE Nucl. Sci. Symp. Conf. Rec. pp 4154-7), Freed et al (2008 Med. Phys. 35 1912-25)).

Effect of Scan Time on Oncologic Lesion Detection in Whole-Body PET

Lesion-detection performance in oncologic PET depends in part upon count statistics, with shorter scans having higher noise and reduced lesion detectability. However, advanced techniques such as time-of-flight (TOF) and point spread function (PSF) modeling can improve lesion detection. This work investigates the relationship between reducing count levels (as a surrogate for scan time) and reconstructing with PSF model and TOF. A series of twenty-four whole-body phantom scans was acquired on a Biograph mCT TOF PET/CT scanner using the experimental methodology prescribed for the Utah PET Lesion Detection Database. Six scans were acquired each day over four days, with up to 23 (68)Ge shell-less lesions (diam. 6, 8, 10, 12, 16 mm) distributed throughout the phantom thorax and pelvis. Each scan acquired 6 bed positions at 240 s/bed in listmode format. The listmode files were then statistically pruned, preserving Poisson statistics, to equivalent count levels for scan times of 180 s, 120 s, 90 s, 60 s, 45 s, 30 s, and 15 s per bed field-of-view, corresponding to whole-body scan times of 1.5-24 min. Each dataset was reconstructed using ordinary Poisson line-of-response (LOR) OSEM, with PSF model, with TOF, and with PSF+TOF. Localization receiver operating characteristics (LROC) analysis was then performed using the channelized non-prewhitened (CNPW) observer. The results were analyzed to delineate the relationship between scan time, reconstruction method, and strength of post-reconstruction filter. Lesion-detection performance degraded as scan time was reduced, and progressively stronger filters were required to maximize performance for the shorter scans. PSF modeling and TOF were found to improve detection performance, but the degree of improvement for TOF was much larger than for PSF for the large phantom used in this study. Notably, the images using TOF provided equivalent lesion-detection performance to the images without TOF for scan durations 40% shorter, suggesting that TOF may offset, at least in part, the need for longer scan times in larger patients.

Quantitative accuracy of MAP reconstruction for dynamic PET imaging in small animals

PURPOSE

Iterative reconstruction algorithms are becoming more commonly employed in positron emission tomography (PET) imaging; however, the quantitative accuracy of the reconstructed images still requires validation for various levels of contrast and counting statistics.

METHODS

The authors present an evaluation of the quantitative accuracy of the 3D maximum a posteriori (3D-MAP) image reconstruction algorithm for dynamic PET imaging with comparisons to two of the most widely used reconstruction algorithms: the 2D filtered-backprojection (2D-FBP) and 2D-ordered subsets expectation maximization (2D-OSEM) on the Siemens microPET scanners. The study was performed for various levels of count density encountered in typical dynamic scanning as well as the imaging of cardiac activity concentration in small animal studies on the Focus 120. Specially designed phantoms were used for evaluation of the spatial resolution, image quality, and quantitative accuracy. A normal mouse was employed to evaluate the accuracy of the blood time activity concentration extracted from left ventricle regions of interest (ROIs) within the images as compared to the actual blood activity concentration measured from arterial blood sampling.

RESULTS

For MAP reconstructions, the spatial resolution and contrast have been found to reach a stable value after 20 iterations independent of the β values (i.e., hyper parameter which controls the weight of the penalty term) and count density within the frame. The spatial resolution obtained with 3D-MAP reaches values of ∼1.0 mm with a β of 0.01 while the 2D-FBP has value of 1.8 mm and 2D-OSEM has a value of 1.6 mm. It has been observed that the lower the hyper parameter β used in MAP, more iterations are needed to reach the stable noise level (i.e., image roughness). The spatial resolution is improved by using a lower β value at the expense of higher image noise. However, with similar noise level the spatial resolution achieved by 3D-MAP was observed to be better than that by 2D-FBP or 2D-OSEM. Using an image quality phantom containing hot spheres, the estimated activity concentration in the largest sphere has the expected concentration relative to the background area for all the MAP images. The obtained recovery coefficients have been also shown to be almost independent of the count density. 2D-FBP and 2D-OSEM do not perform as well, yielding recovery coefficients lower than those observed with 3D-MAP (approximately 33% lower for the smallest sphere). However, a small positive bias was observed in MAP reconstructed images for frames of very low count density. This bias is present in the uniform area for count density of less than 0.05 × 10(6) counts/ml. For the dynamic mouse study, it was observed that 3D-MAP (even gated at diastole) cannot predict accurately the blood activity concentration due to residual spill-over activity from the myocardium into the left ventricle (approximately 15%). However, 3D-MAP predicts blood activity concentration closer to blood sampling than 2D-FBP.

CONCLUSIONS

The authors observed that 3D-MAP produces more accurate activity concentration estimates than 2D-FBP or 2D-OSEM at all practical levels of statistics and contrasts due to improved spatial resolution leading to lesser partial volume effect.

Adaptive imaging for lesion detection using a zoom-in PET system

Positron emission tomography (PET) has become a leading modality in molecular imaging. Demands for further improvements in spatial resolution and sensitivity remain high with growing number of applications. In this paper we present a novel PET system design that integrates a high-resolution depth-of-interaction (DOI) detector into an existing PET system to obtain higher-resolution and higher-sensitivity images in a target region around the face of the high-resolution detector. A unique feature of the proposed PET system is that the high-resolution detector can be adaptively positioned based on the detectability or quantitative accuracy of a feature of interest. This paper focuses on the signal-known-exactly, background-known-exactly (SKE-BKE) detection task. We perform theoretical analysis of lesion detectability using computer observers, and then develop methods that can efficiently calculate the optimal position of the high-resolution detector that maximizes the lesion detectability. We simulated incorporation of a high-resolution DOI detector into the microPET II scanner. Quantitative results verified that the new system has better performance than the microPET II scanner in terms of spatial resolution and lesion detectability, and that the optimal position for lesion detection can be reliably predicted by the proposed method.

High-resolution adaptive PET imaging

While the performance of small animal PET systems has been improved impressively in terms of spatial resolution and sensitivity, demands for further improvements remain high with growing number of applications. Here we propose a novel PET system design that integrates a high-resolution detector into an existing PET system to obtain higher-resolution images in a target region. The high-resolution detector will be adaptively positioned based on the detectability or quantitative accuracy of a feature of interest. The proposed system will be particularly effective for studying human cancers using animal models where tumors are often grown near the skin surface and therefore permit close contact with the high resolution detector. It will also be useful for the high-resolution brain imaging in rodents. In this paper, we present the theoretical analysis and Monte Carlo simulation studies of the performance of the proposed system.

Theoretical analysis and simulation study of a high-resolution zoom-in PET system

We study a novel PET system that integrates a high-resolution zoom-in detector into an existing PET scanner to provide higher resolution and sensitivity in a target region. In contrast to a full-ring PET insert, the proposed system is designed to focus on the target region close to the face of the high-resolution detector. The proposed design is easier to implement than a full-ring insert and provides flexibility for adaptive PET imaging. We developed a maximum a posteriori (MAP) image reconstruction method for the proposed system. Theoretical analysis of the resolution and noise properties of the MAP reconstruction is performed. We show that the proposed PET system offers better performance in terms of resolution-noise tradeoff and lesion detectability. The results are validated using computer simulations.

Lesion detection in dynamic FDG-PET using matched subspace detection

We describe a matched subspace detection algorithm to assist in the detection of small tumors in dynamic positron emission tomography (PET) images. The algorithm is designed to differentiate tumors from background using the time activity curves (TACs) that characterize the uptake of PET tracers. TACs are modeled using linear subspaces with additive Gaussian noise. Using TACs from a primary tumor region of interest (ROI) and one or more background ROIs, each identified by a human observer, two linear subspaces are identified. Applying a matched subspace detector to these identified subspaces on a voxel-by-voxel basis throughout the dynamic image produces a test statistic at each voxel which on thresholding indicates potential locations of secondary or metastatic tumors. The detector is derived for three cases: using a single TAC with white noise of unknown variance, using a single TAC with known noise covariance, and detection using multiple TACs within a small ROI with known noise covariance. The noise covariance is estimated for the reconstructed image from the observed sinogram data. To evaluate the proposed method, a simulation-based receiver operating characteristic (ROC) study for dynamic PET tumor detection is designed. The detector uses a dynamic sequence of frame-by-frame 2-D reconstructions as input. We compare the performance of the subspace detectors with that of a Hotelling observer applied to a single frame image and of the Patlak method applied to the dynamic data. We also show examples of the application of each detection approach to clinical PET data from a breast cancer patient with metastatic disease.

Analysis of Resolution and Noise Properties of Nonquadratically Regularized Image Reconstruction Methods for PET

We present accurate and efficient methods for estimating the spatial resolution and noise properties of nonquadratically regularized image reconstruction for positron emission tomography (PET). It is well known that quadratic regularization tends to over-smooth sharp edges. Many types of edge-preserving nonquadratic penalties have been proposed to overcome this problem. However, there has been little research on the quantitative analysis of nonquadratic regularization due to its nonlinearity. In contrast, quadratically regularized estimators are approximately linear and are well understood in terms of resolution and variance properties. We derive new approximate expressions for the linearized local perturbation response (LLPR) and variance using the Taylor expansion with the remainder term. Although the expressions are implicit, we can use them to accurately predict resolution and variance for nonquadratic regularization where the conventional expressions based on the first-order Taylor truncation fail. They also motivate us to extend the use of a certainty-based modified penalty to nonquadratic regularization cases in order to achieve spatially uniform perturbation responses, analogous to uniform spatial resolution in quadratic regularization. Finally, we develop computationally efficient methods for predicting resolution and variance of nonquadratically regularized reconstruction and present simulations that illustrate the validity of these methods.

A channelized Hotelling observer study of lesion detection in SPECT MAP reconstruction using anatomical priors

In emission tomography, anatomical side information, in the form of organ and lesion boundaries, derived from intra-patient coregistered CT or MR scans can be incorporated into the reconstruction. Our interest is in exploring the efficacy of such side information for lesion detectability. To assess detectability we used the SNR of a channelized Hotelling observer and a signal-known exactly/background-known exactly detection task. In simulation studies, we incorporated anatomical side information into a SPECT MAP (maximum a posteriori) reconstruction by smoothing within but not across organ or lesion boundaries. A non-anatomical prior was applied by uniform smoothing across the entire image. We investigated whether the use of anatomical priors with organ boundaries alone or with perfect lesion boundaries alone would change lesion detectability relative to the case of a prior with no anatomical information. Furthermore, we investigated whether any such detectability changes for the organ-boundary case would be a function of the distance of the lesion to the organ boundary. We also investigated whether any detectability changes for the lesion-boundary case would be a function of the degree of proximity, i.e. a difference in the radius of the true functional lesion and the radius of the anatomical lesion boundary. Our results showed almost no detectability difference with versus without organ boundaries at any lesion-to-organ boundary distance. Our results also showed no difference in lesion detectability with and without lesion boundaries, and no variation of lesion detectability with degree of proximity.

Fast predictions of variance images for fan-beam transmission tomography with quadratic regularization

Accurate predictions of image variances can be useful for reconstruction algorithm analysis and for the design of regularization methods. Computing the predicted variance at every pixel using matrix-based approximations [1] is impractical. Even most recently adopted methods that are based on local discrete Fourier approximations are impractical since they would require a forward and backprojection and two fast Fourier transform (FFT) calculations for every pixel, particularly for shift-variant systems like fan-beam tomography. This paper describes new "analytical" approaches to predicting the approximate variance maps of 2-D images that are reconstructed by penalized-likelihood estimation with quadratic regularization in fan-beam geometries. The simplest of the proposed analytical approaches requires computation equivalent to one backprojection and some summations, so it is computationally practical even for the data sizes in X-ray computed tomography (CT). Simulation results show that it gives accurate predictions of the variance maps. The parallel-beam geometry is a simple special case of the fan-beam analysis. The analysis is also applicable to 2-D positron emission tomography (PET).

Penalized maximum-likelihood image reconstruction for lesion detection

Detecting cancerous lesions is one major application in emission tomography. In this paper, we study penalized maximum-likelihood image reconstruction for this important clinical task. Compared to analytical reconstruction methods, statistical approaches can improve the image quality by accurately modelling the photon detection process and measurement noise in imaging systems. To explore the full potential of penalized maximum-likelihood image reconstruction for lesion detection, we derived simplified theoretical expressions that allow fast evaluation of the detectability of a random lesion. The theoretical results are used to design the regularization parameters to improve lesion detectability. We conducted computer-based Monte Carlo simulations to compare the proposed penalty function, conventional penalty function, and a penalty function for isotropic point spread function. The lesion detectability is measured by a channelized Hotelling observer. The results show that the proposed penalty function outperforms the other penalty functions for lesion detection. The relative improvement is dependent on the size of the lesion. However, we found that the penalty function optimized for a 5 mm lesion still outperforms the other two penalty functions for detecting a 14 mm lesion. Therefore, it is feasible to use the penalty function designed for small lesions in image reconstruction, because detection of large lesions is relatively easy.

Theoretical study of penalized-likelihood image reconstruction for region of interest quantification

Region of interest (ROI) quantification is an important task in emission tomography (e.g., positron emission tomography and single photon emission computed tomography). It is essential for exploring clinical factors such as tumor activity, growth rate, and the efficacy of therapeutic interventions. Statistical image reconstruction methods based on the penalized maximum-likelihood (PML) or maximum a posteriori principle have been developed for emission tomography to deal with the low signal-to-noise ratio of the emission data. Similar to the filter cut-off frequency in the filtered backprojection method, the regularization parameter in PML reconstruction controls the resolution and noise tradeoff and, hence, affects ROI quantification. In this paper, we theoretically analyze the performance of ROI quantification in PML reconstructions. Building on previous work, we derive simplified theoretical expressions for the bias, variance, and ensemble mean-squared-error (EMSE) of the estimated total activity in an ROI that is surrounded by a uniform background. When the mean and covariance matrix of the activity inside the ROI are known, the theoretical expressions are readily computable and allow for fast evaluation of image quality for ROI quantification with different regularization parameters. The optimum regularization parameter can then be selected to minimize the EMSE. Computer simulations are conducted for small ROIs with variable uniform uptake. The results show that the theoretical predictions match the Monte Carlo results reasonably well.

Analysis of observer performance in known-location tasks for tomographic image reconstruction

We consider the task of detecting a statistically varying signal of known location on a statistically varying background in a reconstructed tomographic image. We analyze the performance of linear observer models in this task. We show that, if one chooses a suitable reconstruction method, a broad family of linear observers can exactly achieve the optimal detection performance attainable with any combination of a linear observer and linear reconstructor. This conclusion encompasses several well-known observer models from the literature, including models with a frequency-selective channel mechanism and certain types of internal noise. Interestingly, the "optimal" reconstruction methods are unregularized and in some cases quite unconventional. These results suggest that, for the purposes of designing regularized reconstruction methods that optimize lesion detectability, known-location tasks are of limited use.

Comparison between MAP and postprocessed ML for image reconstruction in emission tomography when anatomical knowledge is available

Previously, the noise characteristics obtained with penalized-likelihood reconstruction [or maximum a posteriori (MAP)] have been compared to those obtained with postsmoothed maximum-likelihood (ML) reconstruction, for emission tomography applications requiring uniform resolution. It was found that penalized-likelihood reconstruction was not superior to postsmoothed ML. In this paper, a similar comparison is made, but now for applications where the noise suppression is tuned with anatomical information. It is assumed that limited but exact anatomical information is available. Two methods were compared. In the first method, the anatomical information is incorporated in the prior of a MAP-algorithm and is, therefore, imposed during MAP-reconstruction. The second method starts from an unconstrained ML-reconstruction, and imposes the anatomical information in a postprocessing step. The theoretical analysis was verified with simulations: small lesions were inserted in two different objects, and noisy PET data were produced and reconstructed with both methods. The resulting images were analyzed with bias-noise curves, and by computing the detection performance of the nonprewhitening observer and a channelized Hotelling observer. Our analysis and simulations indicate that the postprocessing method is inferior, unless the noise correlations between neighboring pixels are taken into account. This can be done by applying a so-called prewhitening filter. However, because the prewhitening filter is shift variant and object dependent, it seems that MAP reconstruction is the more efficient method.

Fast LROC analysis of Bayesian reconstructed emission tomographic images using model observers

Lesion detection and localization is an important task in emission computed tomography. Detection and localization performance with signal location uncertainty may be summarized by a scalar figure of merit, the area under the localization receiver operating characteristic (LROC) curve, A(LROC). We consider model observers to compute A(LROC) for two-dimensional maximum a posteriori (MAP) reconstructions. Model observers may be used to rapidly prototype studies that use human observers. We address the case background-known-exactly (BKE) and signal known except for location. Our A(LROC) calculation makes use of theoretical expressions for the mean and covariance of the reconstruction and, unlike conventional methods that also use model observers, does not require computation of a large number of sample reconstructions. We validate the results of the procedure by comparison to A(LROC) obtained using a gold-standard Monte Carlo method employing a large set of reconstructed noise samples. Under reasonable simulation conditions, our theoretical calculation is about one to two orders of magnitude faster than the conventional Monte Carlo method.

Efficient calculation of resolution and covariance for penalized-likelihood reconstruction in fully 3-D SPECT

Resolution and covariance predictors have been derived previously for penalized-likelihood estimators. These predictors can provide accurate approximations to the local resolution properties and covariance functions for tomographic systems given a good estimate of the mean measurements. Although these predictors may be evaluated iteratively, circulant approximations are often made for practical computation times. However, when numerous evaluations are made repeatedly (as in penalty design or calculation of variance images), these predictors still require large amounts of computing time. In Stayman and Fessler (2000), we discussed methods for precomputing a large portion of the predictor for shift-invariant system geometries. In this paper, we generalize the efficient procedure discussed in Stayman and Fessler (2000) to shift-variant single photon emission computed tomography (SPECT) systems. This generalization relies on a new attenuation approximation and several observations on the symmetries in SPECT systems. These new general procedures apply to both two-dimensional and fully three-dimensional (3-D) SPECT models, that may be either precomputed and stored, or written in procedural form. We demonstrate the high accuracy of the predictions based on these methods using a simulated anthropomorphic phantom and fully 3-D SPECT system. The evaluation of these predictors requires significantly less computation time than traditional prediction techniques, once the system geometry specific precomputations have been made.

Accurate estimation of the Fisher information matrix for the PET image reconstruction problem

The Fisher information matrix (FIM) plays a key role in the analysis and applications of statistical image reconstruction methods based on Poisson data models. The elements of the FIM are a function of the reciprocal of the mean values of sinogram elements. Conventional plug-in FIM estimation methods do not work well at low counts, where the FIM estimate is highly sensitive to the reciprocal mean estimates at individual detector pairs. A generalized error look-up table (GELT) method is developed to estimate the reciprocal of the mean of the sinogram data. This approach is also extended to randoms precorrected data. Based on these techniques, an accurate FIM estimate is obtained for both Poisson and randoms precorrected data. As an application, the new GELT method is used to improve resolution uniformity and achieve near-uniform image resolution in low count situations.

Rapid calculation of detectability in Bayesian single photon emission computed tomography

We consider the calculation of lesion detectability using a mathematical model observer, the channelized Hotelling observer (CHO), in a signal-known-exactly/background-known-exactly detection task for single photon emission computed tomography (SPECT). We focus on SPECT images reconstructed with Bayesian maximum a posteriori methods. While model observers are designed to replace time-consuming studies using human observers, the calculation of CHO detectability is usually accomplished using a large number of sample images, which is still time consuming. We develop theoretical expressions for a measure of detectability, the signal-to-noise-ratio (SNR) of a CHO observer, that can be very rapidly evaluated. Key to our expressions are approximations to the reconstructed image covariance. In these approximations, we use methods developed in the PET literature, but modify them to reflect the different nature of attenuation and distance-dependent blur in SPECT. We validate our expressions with Monte Carlo methods. We show that reasonably accurate estimates of the SNR can be obtained at a computational expense equivalent to approximately two projection operations, and that evaluating SNR for subsequent lesion locations requires negligible additional computation.

Analysis of lesion detectability in Bayesian emission reconstruction with nonstationary object variability

Bayesian methods based on the maximum a posteriori principle (also called penalized maximum-likelihood methods) have been developed to improve image quality in emission tomography. To explore the full potential of Bayesian reconstruction for lesion detection, we derive simplified theoretical expressions that allow fast evaluation of the detectability of a lesion in Bayesian reconstruction. This work is builded on the recent progress on the theoretical analysis of image properties of statistical reconstructions and the development of numerical observers. We explicitly model the nonstationary variation of the lesion and background without assuming that they are locally stationary. The results can be used to choose the optimum prior parameters for the maximum lesion detectability. The theoretical results are validated using Monte Carlo simulations. The comparisons show good agreement between the theoretical predictions and the Monte Carlo results. We also demonstrate that the lesion detectability can be reliably estimated using one noisy data set.

A unified noise analysis for iterative image estimation

Iterative image estimation methods have been widely used in emission tomography. Accurate estimation of the uncertainty of the reconstructed images is essential for quantitative applications. While both iteration-based noise analysis and fixed-point noise analysis have been developed, current iteration-based results are limited to only a few algorithms that have an explicit multiplicative update equation and some may not converge to the fixed-point result. This paper presents a theoretical noise analysis that is applicable to a wide range of preconditioned gradient-type algorithms. Under a certain condition, the proposed method does not require an explicit expression of the preconditioner. By deriving the fixed-point expression from the iteration-based result, we show that the proposed iteration-based noise analysis is consistent with fixed-point analysis. Examples in emission tomography and transmission tomography are shown. The results are validated using Monte Carlo simulations.

Ordered subsets Bayesian tomographic reconstruction using 2-D smoothing splines as priors

The ordered subsets expectation maximization (OS-EM) algorithm has enjoyed considerable interest for accelerating the well-known EM algorithm for emission tomography. The OS principle has also been applied to several regularized EM algorithms, such as nonquadratic convex minimization-based maximum a posteriori (MAP) algorithms. However, most of these methods have not been as practical as OS-EM due to their complex optimization methods and difficulties in hyperparameter estimation. We note here that, by relaxing the requirement of imposing sharp edges and using instead useful quadratic spline priors, solutions are much easier to compute, and hyperparameter calculation becomes less of a problem. In this work, we use two-dimensional smoothing splines as priors and apply a method of iterated conditional modes for the optimization. In this case, step sizes or line-search algorithms necessary for gradient-based descent methods are avoided. We also accelerate the resulting algorithm using the OS approach and propose a principled way of scaling smoothing parameters to retain the strength of smoothing for different subset numbers. Our experimental results show that the OS approach applied to our quadratic MAP algorithms provides a considerable acceleration while retaining the advantages of quadratic spline priors.

Theoretical evaluation of the detectability of random lesions in Bayesian emission reconstruction

Detecting cancerous lesion is an important task in positron emission tomography (PET). Bayesian methods based on the maximum a posteriori principle (also called penalized maximum likelihood methods) have been developed to deal with the low signal to noise ratio in the emission data. Similar to the filter cut-off frequency in the filtered backprojection method, the prior parameters in Bayesian reconstruction control the resolution and noise trade-off and hence affect detectability of lesions in reconstructed images. Bayesian reconstructions are difficult to analyze because the resolution and noise properties are nonlinear and object-dependent. Most research has been based on Monte Carlo simulations, which are very time consuming. Building on the recent progress on the theoretical analysis of image properties of statistical reconstructions and the development of numerical observers, here we develop a theoretical approach for fast computation of lesion detectability in Bayesian reconstruction. The results can be used to choose the optimum hyperparameter for the maximum lesion detectability. New in this work is the use of theoretical expressions that explicitly model the statistical variation of the lesion and background without assuming that the object variation is (locally) stationary. The theoretical results are validated using Monte Carlo simulations. The comparisons show good agreement between the theoretical predications and the Monte Carlo results.

Globally convergent image reconstruction for emission tomography using relaxed ordered subsets algorithms

We present two types of globally convergent relaxed ordered subsets (OS) algorithms for penalized-likelihood image reconstruction in emission tomography: modified block sequential regularized expectation-maximization (BSREM) and relaxed OS separable paraboloidal surrogates (OS-SPS). The global convergence proof of the existing BSREM (De Pierro and Yamagishi, 2001) required a few a posteriori assumptions. By modifying the scaling functions of BSREM, we are able to prove the convergence of the modified BSREM under realistic assumptions. Our modification also makes stepsize selection more convenient. In addition, we introduce relaxation into the OS-SPS algorithm (Erdoğan and Fessler, 1999) that otherwise would converge to a limit cycle. We prove the global convergence of diagonally scaled incremental gradient methods of which the relaxed OS-SPS is a special case; main results of the proofs are from (Nedić and Bertsekas, 2001) and (Correa and Lemaréchal, 1993). Simulation results showed that both new algorithms achieve global convergence yet retain the fast initial convergence speed of conventional unrelaxed ordered subsets algorithms.

Rapid Computation of LROC Figures of Merit Using Numerical Observers (for SPECT/PET Reconstruction)

The assessment of PET and SPECT image reconstructions by image quality metrics is typically time consuming, even if methods employing model observers and samples of reconstructions are used to replace human testing. We consider a detection task where the background is known exactly and the signal is known except for location. We develop theoretical formulae to rapidly evaluate two relevant figures of merit, the area under the LROC curve and the probability of correct localization. The formulae can accommodate different forms of model observer. The theory hinges on the fact that we are able to rapidly compute the mean and covariance of the reconstruction. For four forms of model observer, the theoretical expressions are validated by Monte Carlo studies for the case of MAP (maximum a posteriori) reconstruction. The theory method affords a 10(2) - 10(3) speedup relative to methods in which model observers are applied to sample reconstructions.

Iterative versus filtered backprojection reconstruction for statistical parametric mapping of PET activation measurements: a comparative case study

The significance of task-induced cerebral blood flow responses, assessed using statistical parametric mapping, depends, among other things, on the signal-to-noise ratio (SNR) of these responses. Generally, positron emission tomography sinograms of H(2)(15)O activation studies are reconstructed using filtered backprojection (FBP). Alternatively, the acquired data can be reconstructed using an iterative reconstruction procedure. It has been demonstrated that the application of iterative reconstruction methods improves image SNR as compared with FBP. The aim of this study was to compare FBP with iterative reconstruction, to assess the statistical power of H(2)(15)O-PET activation studies using statistical parametric mapping. For this case study, PET data originating from a bimanual motor task were reconstructed using both FBP and maximum likelihood expectation maximization (ML-EM), an iterative algorithm. Both resulting data sets were statistically analyzed using statistical parametric mapping. It was found, with this dataset, that the statistical analysis of the iteratively reconstructed data confirm the a priori expected physiological response. In addition, increased Z scores were obtained in the iteratively reconstructed data. In particular, for the expected task-related response, activation of the posterior border of the left angular gyrus, the Z score increased from 3.00 to 3.96. Furthermore, the number of statistically significant clusters doubled while their volume increased by more than 50%. In conclusion, iterative reconstruction has the potential to increase the statistical power in H(2)(15)O-PET activation studies as compared with FBP reconstruction.