KEYWORDS: Denoising, Education and training, 3D modeling, Single photon emission computed tomography, 3D image processing, Perfusion imaging, Network architectures, 3D acquisition, Mathematical optimization, Image processing
The purpose of this research is to address the critical challenge of improving the detectability of small perfusion defects in deep learning (DL) denoising for low dose Myocardial Perfusion Imaging (MPI) with Single-Photon Emission Computed Tomography (SPECT). By developing a 3D convolutional auto-encoder (CAE) incorporated with an edge-preservation mechanism, the study aims to mitigate potential blurring effects associated with DL-based denoising methods. The CAE is optimized to enhance noise reduction on low-dose SPECT-MPI scans while seeking to maintain the integrity of image-edge features which are vital for preserving subtle myocardial perfusion defects after denoising.
Convolutional neural networks (CNNs) have been previously used as model observers (MO) for the purpose of defect detection in medical images. Due to their limited generalizability, such CNN observers do not possess the ability to recognize whether an input image comes from the same distribution as the data it was trained on, i.e., the ability of having domain awareness. In this paper we propose an adaptive learning approach for training a domain-aware CNN ideal observer (IO). In our approach we use a variant of U-Net CNN which is trained simultaneously for defect localization prediction and for reconstruction of the input image. We demonstrate that the reconstruction mean-squared-error (MSE) by the network can serve as an indicator of how well the observer performs in the defect localization task, which is an important step towards developing a domain-aware MO. Furthermore we propose an adaptive learning approach by automatically selecting datasets on which the model in training has poor reconstruction MSE. Our results show that this adaptive training approach can improve the model performance both in generalization and defect localization compared to a non-adaptive approach, particularly for out-of-distribution images - images that were not seen during the training of the algorithm.
In computerized detection of clustered microcalcifications (MCs) from mammograms, the traditional approach is to apply a pattern detector to locate the presence of individual MCs, which are subsequently grouped into clusters. Such an approach is often susceptible to the occurrence of false positives (FPs) caused by local image patterns that resemble MCs. We investigate the feasibility of a direct detection approach to determining whether an image region contains clustered MCs or not. Toward this goal, we develop a deep convolutional neural network (CNN) as the classifier model to which the input consists of a large image window (1 cm2 in size). The multiple layers in the CNN classifier are trained to automatically extract image features relevant to MCs at different spatial scales. In the experiments, we demonstrated this approach on a dataset consisting of both screen-film mammograms and full-field digital mammograms. We evaluated the detection performance both on classifying image regions of clustered MCs using a receiver operating characteristic (ROC) analysis and on detecting clustered MCs from full mammograms by a free-response receiver operating characteristic analysis. For comparison, we also considered a recently developed MC detector with FP suppression. In classifying image regions of clustered MCs, the CNN classifier achieved 0.971 in the area under the ROC curve, compared to 0.944 for the MC detector. In detecting clustered MCs from full mammograms, at 90% sensitivity, the CNN classifier obtained an FP rate of 0.69 clusters/image, compared to 1.17 clusters/image by the MC detector. These results indicate that using global image features can be more effective in discriminating clustered MCs from FPs caused by various sources, such as linear structures, thereby providing a more accurate detection of clustered MCs on mammograms.
Retrieving a set of known lesions similar to the one being evaluated might be of value for assisting radiologists to distinguish between benign and malignant clustered microcalcifications (MCs) in mammograms. In this work, we investigate how perceptually similar cases with clustered MCs may relate to one another in terms of their underlying characteristics (from disease condition to image features). We first conduct an observer study to collect similarity scores from a group of readers (five radiologists and five non-radiologists) on a set of 2,000 image pairs, which were selected from 222 cases based on their images features. We then explore the potential relationship among the different cases as revealed by their similarity ratings. We apply the multi-dimensional scaling (MDS) technique to embed all the cases in a 2-D plot, in which perceptually similar cases are placed in close vicinity of one another based on their level of similarity. Our results show that cases having different characteristics in their clustered MCs are accordingly placed in different regions in the plot. Moreover, cases of same pathology tend to be clustered together locally, and neighboring cases (which are more similar) tend to be also similar in their clustered MCs (e.g., cluster size and shape). These results indicate that subjective similarity ratings from the readers are well correlated with the image features of the underlying MCs of the cases, and that perceptually similar cases could be of diagnostic value for discriminating between malignant and benign cases.
In this work, we conducted an imaging study to make a direct, quantitative comparison of image features measured by
film and full-field digital mammography (FFDM). We acquired images of cadaver breast specimens containing
simulated microcalcifications using both a GE digital mammography system and a screen-film system. To quantify the
image features, we calculated and compared a set of 12 texture features derived from spatial gray-level dependence
matrices. Our results demonstrate that there is a great degree of agreement between film and FFDM, with the correlation
coefficient of the feature vector (formed by the 12 textural features) being 0.9569 between the two; in addition, a paired
sign test reveals no significant difference between film and FFDM features. These results indicate that textural features
may be interchangeable between film and FFDM for CAD algorithms.
In this paper, we present a numerical observer for image quality assessment, aiming to predict human observer accuracy
in a cardiac perfusion defect detection task for single-photon emission computed tomography (SPECT). In medical
imaging, image quality should be assessed by evaluating the human observer accuracy for a specific diagnostic task.
This approach is known as task-based assessment. Such evaluations are important for optimizing and testing imaging
devices and algorithms. Unfortunately, human observer studies with expert readers are costly and time-demanding. To
address this problem, numerical observers have been developed as a surrogate for human readers to predict human
diagnostic performance. The channelized Hotelling observer (CHO) with internal noise model has been found to predict
human performance well in some situations, but does not always generalize well to unseen data. We have argued in the
past that finding a model to predict human observers could be viewed as a machine learning problem. Following this
approach, in this paper we propose a channelized relevance vector machine (CRVM) to predict human diagnostic scores
in a detection task. We have previously used channelized support vector machines (CSVM) to predict human scores and
have shown that this approach offers better and more robust predictions than the classical CHO method. The comparison
of the proposed CRVM with our previously introduced CSVM method suggests that CRVM can achieve similar
generalization accuracy, while dramatically reducing model complexity and computation time.
In medical imaging, image quality is commonly assessed by measuring the performance of a human observer performing
a specific diagnostic task. However, in practice studies involving human observers are time consuming and difficult to
implement. Therefore, numerical observers have been developed, aiming to predict human diagnostic performance to
facilitate image quality assessment. In this paper, we present a numerical observer for assessment of cardiac motion in
cardiac-gated SPECT images. Cardiac-gated SPECT is a nuclear medicine modality used routinely in the evaluation of
coronary artery disease. Numerical observers have been developed for image quality assessment via analysis of
detectability of myocardial perfusion defects (e.g., the channelized Hotelling observer), but no numerical observer for
cardiac motion assessment has been reported. In this work, we present a method to design a numerical observer aiming
to predict human performance in detection of cardiac motion defects. Cardiac motion is estimated from reconstructed
gated images using a deformable mesh model. Motion features are then extracted from the estimated motion field and
used to train a support vector machine regression model predicting human scores (human observers' confidence in the
presence of the defect). Results show that the proposed method could accurately predict human detection performance
and achieve good generalization properties when tested on data with different levels of post-reconstruction filtering.
In this work, we present a four-dimensional reconstruction technique for cardiac gated SPECT images using a content-adaptive deformable mesh model. Cardiac gated SPECT images are affected by a high level of noise.
Noise reduction methods usually do not account for cardiac motion and therefore introduce motion blur-an artifact
that can decrease diagnostic accuracy. Additionally, image reconstruction methods typically rely on uniform
sampling and Cartesian griding for image representation. The proposed method utilizes a mesh representation
of the images in order to utilize the benefits of content-adaptive nonuniform sampling. The mesh model allows
for accurate representation of important regions while significantly compressing the data. The content-adaptive
deformable mesh model is generated by combining nodes generated on the full torso using pre-reconstructed emission
and attenuation images with nodes accurately sampled on the left ventricle. Ventricular nodes are further
displaced according to cardiac motion using our previously introduced motion estimation technique. The resulting
mesh structure is then used to perform iterative image reconstruction using a mesh-based maximum-likelihood
expectation-maximization algorithm. Finally, motion-compensated post-reconstruction temporal filtering is applied
in the mesh domain using the deformable mesh model. Reconstructed images as well as quantitative
evaluation show that the proposed method offers improved image quality while reducing the data size.
In this paper, we present a numerical observer for assessment of cardiac motion in nuclear medicine. Numerical
observers are used in medical imaging as a surrogate for human observers to automatically measure the diagnostic
quality of medical images. The most commonly used quality measurement is the detection performance in a detection
task. In this work, we present a new numerical observer aiming to measure image quality for the task of cardiac motiondefect
detection in cardiac SPECT imaging. The proposed observer utilizes a linear discriminant on features extracted
from cardiac motion, characterized by a deformable mesh model of the left ventricle and myocardial brightening.
Simulations using synthetic data indicate that the proposed method can effectively capture the cardiac motion and
provide an accurate prediction of the human observer performance.
We present a post-reconstruction motion-compensated spatio-temporal filtering method for noise reduction in cardiac
gated SPECT images. SPECT imaging suffers from low photon count due to radioactive dose limitations resulting in a
high noise level in the reconstructed images. This is especially true in gated cardiac SPECT where the total number of
counts is divided into a number of gates (time frames). Classical spatio-temporal filtering approaches, used in gated
cardiac SPECT for noise reduction, do not accurately account for myocardium motion and brightening and therefore
perform sub-optimally. The proposed post-reconstruction method consists of two steps: motion and brightening
estimation and spatio-temporal motion-compensated filtering. In the first step we utilize a left ventricle model and a
deformable mesh structure. The second step, which consists of motion-compensated spatio-temporal filtering, makes use
of estimated myocardial motion to enable accurate smoothing. Additionally, the algorithm preserves myocardial
brightening, a result of partial volume effect which is widely used as a diagnostic feature. The proposed method is
evaluated quantitatively to assess noise reduction and the influence on estimated ejection fraction.
Photoacoustic tomography (PAT) is a hybrid imaging modality that combines the advantages of both optical
imaging and ultrasound imaging techniques. Most existing reconstruction algorithms assume the speed-of-sound
distribution within the object is homogeneous. In certain practical applications, this assumption may not be
valid and will result in conspicuous image artifacts. In this work, we investigate the possibility of simultaneously
estimating the speed-of-sound and optical absorption properties from data acquired in a PAT experiment. We
propose and numerically implement a time-domain iterative algorithm that can reconstruct both the speed-of-sound and optical absorption distribution, by use of a priori information regarding the geometry of the speed-of-sound map. Computer-simulation results are presented to demonstrate the efficacy of the proposed
reconstruction method.
KEYWORDS: Signal attenuation, Tomography, Sensors, Single photon emission computed tomography, Monte Carlo methods, Data modeling, Collimators, Expectation maximization algorithms, Image restoration, Signal to noise ratio
In this paper, we present a new methodology for calculation of a 2D projection operator for emission tomography
using a content-adaptive mesh model (CAMM). A CAMM is an efficient image representation based on adaptive
sampling and linear interpolation, wherein non-uniform image samples are placed most densely in regions having fine
detail. We have studied CAMM in recent years and shown that a CAMM is an efficient tool for data representation and
tomographic reconstruction. In addition, it can also provide a unified framework for tomographic reconstruction of organs
(e.g., the heart) that undergo non-rigid deformation. In this work we develop a projection operator model suitable for a
CAMM representation such that it accounts for several major degradation factors in data acquisition, namely object
attenuation and depth-dependent blur in detector-collimator response. The projection operator is calculated using a ray-tracing
algorithm. We tested the developed projection operator by using Monte Carlo simulation for single photon
emission tomography (SPECT). The methodology presented here can also be extended to transmission tomography.
KEYWORDS: Single photon emission computed tomography, Data modeling, Heart, Reconstruction algorithms, Signal to noise ratio, Motion estimation, Smoothing, Data acquisition, Cameras, Technetium
In this work we propose a spatio-temporal approach for reconstruction of dynamic gated cardiac SPECT images.
As in traditional gated cardiac SPECT, the cardiac cycle is divided into a number of gate intervals, but the
tracer distribution is treated as a time-varying signal for each gate. Our goal is to produce a dynamic image
sequence that shows both cardiac motion and time-varying tracer distribution. To combat the ill-conditioned
nature of the problem, we use B-spline basis functions to regulate the time activities curves, and apply a joint
MAP estimation approach based on motion compensation for reconstruction of all the different gates. We also
explore the benefit of using a time-varying regularization prior for the gated dynamic images. The proposed
approach is evaluated using a dynamic version of the 4D gated mathematical cardiac torso phantom (gMCAT)
simulating a gated SPECT perfusion acquisition with Technitium-99m labeled Teboroxime.
The human user is an often ignored component of the imaging chain. In medical diagnostic tasks, the human observer plays the role of the decision-maker, forming opinions based on visual assessment of images. In content-based image retrieval, the human user is the ultimate judge of the relevance of images recalled from a database. We argue that data collected from human observers should be used in conjunction with machine-learning algorithms to model and optimize
performance in tasks that involve humans. In essence, we treat the human observer as a nonlinear system to be identified. In this paper, we review our work in two applications of this general idea. In the first, a learning machine is trained to predict the accuracy of human observers in a lesion detection task for purposes of assessing image quality. In the second, a learning machine is trained to predict human users' perception of the similarity of two images for purposes
of content-based image retrieval from a database. In both examples, it is shown that a nonlinear learning machine can accurately identify the nonlinear human system that maps images into numerical values, such as detection performance or image similarity.
Conventional mammography is one of the most widely used diagnostic imaging techniques, but it has serious and well-known shortcomings, which are driving the development of innovative alternatives. Our group has been developing an x-ray imaging approach called multiple-image radiography (MIR), which shows promise as a potential alternative to conventional x-ray imaging (radiography). Like computed tomography (CT), MIR is a computed imaging technique, in which the images are not directly observed, but rather computed algorithmically. Whereas conventional radiography produces just one image depicting absorption effects, MIR simultaneously produces three images, showing separately the effects of absorption, refraction, and ultra-small-angle x-ray scattering. The latter two effects are caused by refractive-index variations in the object, which yield fine image details not seen in standard radiographs. MIR has the added benefits of dramatically lessening radiation dose, virtually eliminating scatter degradation, and lessening the importance of compressing the breast during imaging. In this paper we review progress to date on the MIR technique, focus on the basic physics and signal-processing issues involved in this new imaging method.
In many applications, the illuminant condition of color reproductions is different from that of targets. To achieve the same color perceptions, reproduced colors should be adjusted according to the changes of reference whites determined by the illuminants. In this paper, we consider this color reproduction problem which is also subject to material constraints. By material constraints we mean any constraints that are applied to the amount of inks, lights, voltages, and currents that are used in the generation of color. Color reproduction that is subject to material constraints is called the relaxed color reproduction, because the reproduced colors may not match the targets exactly. An algorithm that is suitable for this task is the method of vector space projections (VSP). In our work, VSP method is directly applied to the control signals of devices. The effects of illuminant variances are also studied. In order to use VSP for constrained color reproduction, we use a novel approach to convert the non-linear constraints in the CIE-Lab space into simpler linear forms. Experimental results demonstrate the feasibility of this method.
Microcalcification (MC) clusters in mammograms can be important early signs of breast cancer in women. Accurate detection of MC clusters is an important but challenging problem. In this paper, we propose the use of a recently developed machine learning technique -- relevance vector machine (RVM) -- for automatic detection of MCs in digitized mammograms. RVM is based on Bayesian estimation theory, and as a feature it can yield a decision function that depends on only a very small number of so-called relevance vectors. We formulate MC detection as a supervised-learning problem, and use RVM to classify if an MC object is present or not at each location in a mammogram image. MC clusters are then identified by grouping the detected MC objects. The proposed method is tested using a database of 141 clinical mammograms, and compared with a support vector machine (SVM) classifier which we developed previously. The detection performance is evaluated using the free-response receiver operating characteristic (FROC) curves. It is demonstrated that the RVM classifier matches closely with the SVM classifier in detection performance, and does so with a much sparser kernel representation than the SVM classifier. Consequently, the RVM classifier greatly reduces the computational complexity, making it more suitable for real-time processing of MC clusters in mammograms.
In this paper we investigate several state-of-the-art machine-learning methods for automated classification of clustered microcalcifications (MCs), aimed to assisting radiologists for more accurate diagnosis of breast cancer in a computer-aided diagnosis (CADx) scheme. The methods we consider include: support vector machine (SVM), kernel Fisher discriminant (KFD), and committee machines (ensemble averaging and AdaBoost), most of which have been developed recently in statistical learning theory. We formulate differentiation of malignant from benign MCs as a supervised learning problem, and apply these learning methods to develop the classification algorithms. As input, these methods use image features automatically extracted from clustered MCs. We test these methods using a database of 697 clinical mammograms from 386 cases, which include a wide spectrum of difficult-to-classify cases. We use receiver operating characteristic (ROC) analysis to evaluate and compare the classification performance by the different methods. In addition, we also investigate how to combine information from multiple-view mammograms of the same case so that the best decision can be made by a classifier. In our experiments, the kernel-based methods (i.e., SVM, KFD) yield the best performance, significantly outperforming a well-established CADx approach based on neural network learning.
In this paper we investigate the benefits of a spatio-temporal approach for reconstruction of cardiac image sequences. We introduce a temporal prior based on motion-compensation to enforce temporal correlations along the curved trajectories that follow the cardiac motion. The image frames in a sequence are reconstructed simultaneously through maximum a posteriori (MAP) estimation. We evaluated the performance of our algorithm using the 4D gated mathematical cardiac-torso (gMCAT) D1.01 phantom to simulate gated SPECT perfusion imaging with Tc-99m-sestamibi. Our experimental results show that the proposed approach could significantly improve the accuracy of reconstructed images without causing cross-frame blurring that may arise form the cardiac motion.
In conventional computed tomography (CT) a single volumetric image representing the linear attenuation coefficient of an object is produced. For weakly absorbing tissues, the attenuation of the X-ray beam may not be the best description of disease-related information. In this work we present a new volumetric imaging method, called multiple-image computed tomography (MICT), that can concurrently produce several images from a set of measurements made with a single X-ray beam. MICT produces three volumetric images that represent the attenuation, refraction, ultra-small-angle scattering properties of an object. The MICT method is implemented to reconstruct images of a physical phantom and a biological object from measurement data produced by a synchroton light source. An iterative reconstruction method is employed for reconstruction of MICT images from experimental data sets that contains enhanced Poisson noise levels that are representative of future benchtop implementations of MICT. We also demonstrated that images produced by the DEI-CT method (the predecessor of MICT) can contain significant artifacts due to ultra-small-angle scattering effects while the corresponding MICT images do not.
In this paper we study a motion-compensated approach for simultaneous reconstruction of image frames in a time sequence. We treat the frames in a sequence collectively as a single function of both space and time, and define a temporal prior to account for the temporal correlations in a sequence. This temporal prior is defined in a form of motion-compensation, aimed to follow the curved trajectories of the object motion through space-time. The image frames are then obtained through estimation using the expectation-maximization (EM) algorithm. The proposed algorithm was evaluated extensively using the 4D gated mathematical cardiac-torso (gMCAT) D1.01 phantom to simulate gated SPECT perfusion imaging with Tc99m. Our experimental results demonstrate that the use of motion compensation for reconstruction can lead to significant improvement in image quality and reconstruction accuracy.
We are comparing two different methods for obtaining the radiologists’ subjective impression of similarity, for application in distinguishing benign from malignant lesions. Thirty pairs of mammographic clustered calcifications were used in this study. These 30 pairs were rated on a 5-point scale as to their similarity, where 1 was nearly identical and 5 was not at all similar. After this, all possible combinations of pairs of pairs were shown to the reader (n=435) and the reader selected which pair was most similar. This experiment was repeated by the observers with at least a week between reading sessions. Using analysis of variance, intra-class correlation coefficients (ICC) were calculated for both absolute scoring method and paired comparison method. In addition, for the paired comparison method, the coefficient of consistency within each reader was calculated. The average coefficient of consistence for the 4 readers was 0.88 (range 0.49-0.97). These results were statistically significant different from guessing at p << 0.0001. The ICC for intra-reader agreement was 0.51 (0.37-0.66 95% CI) for the absolute method and 0.82 (0.73-0.91 95% CI) for the paired comparison method. This difference was statistically significant (p=0.001). For the inter-reader agreement, the ICC for the absolute method was 0.39 (0.21-0.57 95% CI) and 0.37 (0.18-0.56 95% CI) for the paired comparison method. We conclude that humans are able to judge similarity of clustered calcifications in a meaningful way. Further, radiologists had greater intra-reader agreement when using the paired comparison method than when using an absolute rating scale. Differences in the criteria used by different observers to judge similarity and differences in interpreting which calcifications comprise the cluster can lead to low ICC values for inter-reader agreement for both methods.
In this work we explore the use of a content-adaptive mesh model (CAMM) in the classical problem of image restoration. In the proposed framework, we first model the image to be restored by an efficient mesh representation. A CAMM can be viewed as a form of image representation using non-uniform samples, of which the mesh nodes (i.e., image samples) are adaptively placed according to the local content of the image. The image is then restored through estimating the model parameters (i.e., mesh nodal values) from the data. There are several potential advantages of the proposed approach. First, a CAMM provides a spatially-adaptive regularization framework. This is achieved by the fact that the interpolation basis functions in a CAMM have support strictly limited to only those elements that they are associated with. Second, a CAMM provides an efficient, but accurate, representation of the image, thereby greatly reducing the number of parameters to be estimated. In this work we present some exploratory results to demonstrate the proposed approach.
We propose a new algorithm for restoring quantum limited images. The algorithm is based on projection methods but uses constraint sets in which set membership is based on probabilistic measures. Such constraints can be regarded as soft, as opposed to hard constraints in which less latitude is given in defining set membership. We show that the restoration of quantum- limited images has certain similarities to estimating a probability density function. We apply the algorithm to a widely used image phantom and demonstrate that the processed image features less noise without blurred edges.
In conventional block-transform coding, the compressed images are decoded using only the transmitted transform data. In this paper, we formulate image decoding as an image recovery problem. According to this approach, the decoded image is reconstructed using not only the transmitted data, but in addition, the prior knowledge that images before compression do not display blocking artifacts. A spatially-adaptive image recovery algorithm is proposed based on the theory of projections onto convex sets. Numerical experiments demonstrate that the proposed algorithm yields images superior to those from both the JPEG deblocking recommendation and a projection-based image decoding approach.
In low bit-rate applications, MPEG based compressed video exhibits annoying coding artifacts. In this paper, image recovery algorithms based on the theory of projections onto convex sets are proposed to decode video images from MPEG data without coding artifacts. According to this approach, each video frame is reconstructed using not only the transmitted data but also other prior knowledge which is available and not explicitly used by the conventional MPEG decoding algorithm. Numerical experiments demonstrate that the proposed algorithm yield images superior to those from conventional MPEG decoders.
In this paper, we present a set theoretic based approach for solving the classical blind deconvolution problem. In this new approach, every piece of available information about both the source function and the system impulse response function is expressed via a constraint set. The link between generalized projection based algorithms and algorithms studied previously by other researchers is established. Motivated by this relation, a new algorithm is proposed which enjoys good convergent properties. Finally, numerical examples are presented to test the proposed algorithm.
In most block-transform based codecs (coder-decoder) the compressed image is reconstructed using only the transmitted data. In this paper, the reconstruction is formulated as a regularized image recovery problem where both the transmitted data and prior knowledge about the properties of the original image are used. This is accomplished by minimizing an objective function, using iterative algorithms, which captures the smoothness properties of the original image. Experimental results are presented which demonstrate that the proposed regularized algorithms yield reconstructed images with superior quality, both visually and using objective distance metrics, to that of traditional decoders that use only the transmitted transform coefficients.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.