Chung Chan

Regularized image reconstruction with an anatomically adaptive prior for positron emission tomography

The incorporation of accurately aligned anatomical information as a prior to guide reconstruction and noise regularization in positron emission tomography (PET) has been suggested in many previous studies. However, the advantages of this approach can only be realized if the exact lesion outline is also available. In practice, the anatomical imaging modality may be unable to differentiate between normal and pathological tissues, and thus the edges of lesions seen in the anatomical image may not correspond to functional boundaries in the emission image. In this study, we explored an alternative approach to incorporating an anatomical prior into PET image reconstruction. Of particular interest was the realistic situation where lesions are apparent in the emission images but not in the corresponding anatomical images. In the proposed method, regional information obtained from the anatomical prior was used to estimate an anatomically adaptive anisotropic median-diffusion filtering (AAMDF) prior. This smoothing prior was determined and applied adaptively to each anatomical region on the emission image and then assembled to form a prior image for the next iteration in the reconstruction process. We formulated a two-step joint estimation reconstruction scheme to update the estimated image and prior image iteratively. The proposed AAMDF prior was evaluated and compared with maximum a posteriori (MAP) reconstruction methods with and without anatomical side information. In experiments using synthetic and physical phantom data, the AAMDF prior yielded overall higher lesion-to-background contrast and less error in lesion estimation than other algorithms for a comparable level of background noise. We conclude that lesion contrast and quantification can be improved using an anatomically derived smoothing prior without requiring knowledge of the lesion boundary. This may have important implications in clinical PET/CT, where lesion boundaries are often not obtainable from CT images.

Postreconstruction Nonlocal Means Filtering of Whole-Body PET With an Anatomical Prior

Positron emission tomography (PET) images usually suffer from poor signal-to-noise ratio (SNR) due to the high level of noise and low spatial resolution, which adversely affect its performance for lesion detection and quantification. The complementary information present in high-resolution anatomical images from multi-modality imaging systems could potentially be used to improve the ability to detect and/or quantify lesions. However, previous methods that use anatomical priors usually require matched organ/lesion boundaries. In this study, we investigated the use of anatomical information to suppress noise in PET images while preserving both quantitative accuracy and the amplitude of prominent signals that do not have corresponding boundaries on computerized tomography (CT). The proposed approach was realized through a postreconstruction filter based on the nonlocal means (NLM) filter, which reduces noise by computing the weighted average of voxels based on the similarity measurement between patches of voxels within the image. Anatomical knowledge obtained from CT was incorporated to constrain the similarity measurement within a subset of voxels. In contrast to other methods that use anatomical priors, the actual number of neighboring voxels and weights used for smoothing were determined from a robust measurement on PET images within the subset. Thus, the proposed approach can be robust to signal mismatches between PET and CT. A 3-D search scheme was also investigated for the volumetric PET/CT data. The proposed anatomically guided median nonlocal means filter (AMNLM) was first evaluated using a computer phantom and a physical phantom to simulate realistic but challenging situations where small lesions are located in homogeneous regions, which can be detected on PET but not on CT. The proposed method was further assessed with a clinical study of a patient with lung lesions. The performance of the proposed method was compared to Gaussian, edge-preserving bilateral and NLM filters, as well as median nonlocal means (MNLM) filtering without an anatomical prior. The proposed AMNLM method yielded improved lesion contrast and SNR compared with other methods even with imperfect anatomical knowledge, such as missing lesion boundaries and mismatched organ boundaries.

Median non-local means filtering for low SNR image denoising: Application to PET with anatomical knowledge

Denoising low signal-to-noise-ratio (SNR) images is a significant challenge since the intensity gradient due to noise elements may compete with or even exceed the intensity gradient due to features in the images. This situation can often be encountered in photon-limited medical imaging applications such as MLEM reconstructed Positron Emission Tomography (PET) images. In this study, we propose a median non-local means filter for denoising low-SNR images. The proposed method incorporates a median filtering operation indirectly in the nonxadlocal means (NLM) method, which gives more robust estimation of the weights used to average the pixels in the image. For the application of multi-modality imaging such as PET/CT, we further extended the method to incorporate anatomical side information which can be obtained from co-registered CT images without segmentation to preserve abrupt changes between organs on PET images and reduce the computational cost of weight calculations. We applied the proposed method (AMNLM) to a PET/CT simulation, a real physical phantom study and a clinical patient study with lung lesions. The results suggest that the proposed method outperforms the standard Gaussian filtering approach, anisotropic-median diffusion filtering (AMDF) and NLM in terms of visual assessment and trade-off between lesion contrast and noise.

A non-local post-filtering algorithm for PET incorporating anatomical knowledge

The maximum likelihood expectation maximization (MLEM) reconstruction method is known to yield noisy images at high iteration numbers because emission tomographic reconstruction is an ill-posed problem. The noise can be suppressed by post-filtering the ML estimate or imposing a priori knowledge as a constraint within a Bayesian reconstruction framework. Most of these filters and priors are based on weighting the intensity differences between neighbouring pixels within a small local neighbourhood. Therefore, they have limited information to distinguish edges from noise. We investigated the use of a non-local means (NLM) filter for post-filtering MLEM reconstructed positron emission tomography (PET) images. We further investigated the effect of incorporating anatomical side information obtained from co-registered computed tomography (CT) images into the NLM, resulting in an adaptive non-local means (A-NLM) filter which takes into account the variance within each anatomical region on the PET image. In simulated and physical phantom experiments, the A-NLM filter demonstrated superior performance tradeoff between lesion contrast and noise than conventional Gaussian post-filtering and NLM without anatomical prior. We conclude that the A-NLM filter has potential for improved lesion detection over Gaussian post-filtered MLEM images.

Interactive fusion and contrast enhancement for whole body PET/CT data using multi-image pixel composting

The most important application of the dual-modality PET/CT is the ability to efficiently display the fused data. However, in PET/CT fusion, the amount of information displayed is often impaired as the CT data occupies greater range of contrast than that is possible to display without enhancements. A common approach to improving the CT information in the PET/CT fusion is by enhancing the contrast range of the CT data which can improve on the accuracy of structure localization and PET/CT interpretation. In this study, we present an interactive multi-image fusion which optimizes the display of the information from dual-modality PET/CT data. By interactively selecting a specific CT contrast range and assigning the resultant image as a layer, the multi-layers can be constructed and then fused using the multi-image pixel compositing. The enhanced CT data is further fused with the PET data for PET/CT diagnosis. The proposed algorithm is able to simultaneously display greater amount of information from the fused PET/CT data and reveal substantial details of the CT data that would not have been possible with standard PET/CT fusion. The preliminary results are encouraging and show potential in the PET/CT diagnosis and interpretation

Automatic Mobile Device Synchronization and Remote Control System for High-Performance Medical Applications

The field of telemedicine is in need of generic solutions that harness the power of small, easily carried computing devices to increase efficiency and decrease the likelihood of medical errors. Our study resolved to build a framework to bridge the gap between handheld and desktop solutions by developing an automated network protocol that wirelessly propagates application data and images prepared by a powerful workstation to handheld clients for storage, display and collaborative manipulation. To this end, we present the Mobile Active Medical Protocol (MAMP), a framework capable of nigh-effortlessly linking medical workstation solutions to corresponding control interfaces on handheld devices for remote storage, control and display. The ease-of-use, encapsulation and applicability of this automated solution is designed to provide significant benefits to the rapid development of telemedical solutions. Our results demonstrate that the design of this system allows an acceptable data transfer rate, a usable framerate for diagnostic solutions and enough flexibility to enable its use in a wide variety of cases. To this end, we also present a large-scale multi-modality image viewer as an example application based on the MAMP.

Minimum Cross-entropy Reconstruction of PET Images with Anatomically Based Anisotropic Median-Diffusion Filtering

We propose a spatially-variant anisotropic median-diffusion filter prior aided by anatomical knowledge for PET reconstruction. The anisotropic median-diffusion filter is applied locally to an anatomical region which is defined from a co-registered CT image. The individually smoothed regions are then combined to form a prior term in the minimum cross-entropy reconstruction algorithm. A simulated PET thorax phantom with lesions was investigated in terms of bias and contrast versus noise tradeoffs. Compared with MLEM and three other maximum a posteriori (MAP)-like reconstruction algorithms, the proposed algorithm demonstrated better bias-noise tradeoff except when the lesion was close to an anatomical boundary and better contrast-noise tradeoff in all cases.

An anatomically based regionally adaptive prior for MAP reconstruction in emission tomography

The boundary information derived from anatomical images can be incorporated into maximum a posteriori (MAP) reconstruction algorithms to improve the quality of reconstructed images in positron emission tomography (PET). However, challenges arise from mismatches between anatomical (CT) and functional (PET) images which are unavoidable in practice. The aim of this study is to devise a new approach to incorporating anatomical knowledge into emission tomographic reconstruction which is robust to the mismatches while still improving the quality of reconstructed PET images. An anatomically based regionally adaptive regularization MAP (RMAP) is presented. The anatomical knowledge is introduced by labeling the current estimate of the PET image with different anatomical regions derived from the corresponding CT image. An intensity selective non-convex prior is used to model the local smoothness properties adaptively in each anatomical region. The regionally adaptive priors are then combined to form a prior in the Bayesian formulation for the next iteration in the reconstruction. Simulated results show that the proposed algorithm yielded superior lesion contrast recovery, bias- variance tradeoff and robustness to the mismatches between anatomical and functional images compared with MAP with a conventional non-convex prior and MAP with anatomical prior.

Quarantine MAP reconstruction of PET/CT data using dual priors

Maximum a posteriori (MAP) reconstruction makes use of an anatomical prior from CT or MRI imaging to enforce smoothness of reconstructed PET images while preserving anatomical edges. The tendency of this technique to smooth parts of the image between anatomical boundaries may reduce the detectability of functional lesions if, as is commonly the case, the edges of these lesions do not conform to anatomical boundaries. We have investigated the use of a functional prior in addition to an anatomical prior to improve the detection and quantification of lesions in PET imaging. We introduce a new parameter, Q, which controls the weight, β, of the functional prior on a spatially-variant basis, to enable a reduction of the smoothing effect in regions containing lesions. Such regions constitute the functional prior. They can be defined, for example, by applying a threshold to a preliminary reconstructed PET image. They are quarantined from the smoothing of the standard MAP algorithm, and subjected to a lesser degree of smoothing as determined by the combined effects of Q and β. We call this dual-prior technique quarantine MAP reconstruction (QMAP). Thus the method alters the degree of smoothing in specific parts of the image with the aim of enhancing lesion detectability. We have compared the QMAP algorithm in computer simulations with standard One-Step-Late (OSL) MAP reconstruction and OSL-MAP with CT prior information. QMAP provided better lesion contrast than the other algorithms, without altering the properties of other parts of the image.

The impact of reconstruction algorithms on semi-automatic small lesion segmentation for PET: A phantom study

A robust lesion segmentation method is critical for quantification of lesion activity in positron emission tomography (PET), especially for the cases where lesion boundary is not discernible in the corresponding computed tomography (CT). However, lesion delineation in PET is a challenging task, especially for small lesions, due to the low intrinsic resolution, image noise and partial volume effect. The combinations of different reconstruction methods and post-reconstruction smoothing on PET images also affect the segmentation result significantly which has always been overlooked. Therefore, the aim of this study was to investigate the impact of different reconstruction methods on semi-automated small lesion segmentation for PET images. Four conventional segmentation methods were evaluated including region growing technique based on maximum intensity (RGmax) and mean intensity (RGmean) thresholds, Fuzzy c-mean (FCM) and watershed (WS) technique. All these methods were evaluated on a physical phantom scan which was reconstructed with Ordered Subset Expectation Maximization (OSEM) with Gaussian post-smoothing and Maximum a Posteriori (MAP) with quadratic prior respectively. The results demonstrate that: 1) the performance of all the segmentation methods subject to the smoothness constraint applied on the reconstructed images; 2) FCM method applied on MAP reconstructed images yielded overall superior performance than other evaluated combinations.