Search

Deriving accurate attenuation maps for PET/MRI remains a challenging problem because MRI voxel intensities are not related to properties of photon attenuation and bone/air interfaces have similarly low signal. This work presents a learning-based method to derive patient-specific computed tomography (CT) maps from routine T1-weighted MRI in their native space for attenuation correction of brain PET. We developed a machine-learning-based method using a sequence of alternating random forests under the framework of an iterative refinement model. Anatomical feature selection is included in both training and predication stages to achieve optimal performance. To evaluate its accuracy, we retrospectively investigated 17 patients, each of which has been scanned by PET/CT and MR for brain. The PET images were corrected for attenuation on CT images as ground truth, as well as on pseudo CT (PCT) images generated from MR images. The PCT images showed mean average error of 66.1 ± 8.5 HU, average correlation coefficient of 0.974 ± 0.018 and average Dice similarity coefficient (DSC) larger than 0.85 for air, bone and soft tissue. The side-by-side image comparisons and joint histograms demonstrated very good agreement of PET images after correction by PCT and CT. The mean differences of voxel values in selected VOIs were less than 4%, the mean absolute difference of all active area is around 2.5%, and the mean linear correlation coefficient is 0.989 ± 0.017 between PET images corrected by CT and PCT. This work demonstrates a novel learning-based approach to automatically generate CT images from routine T1-weighted MR images based on a random forest regression with patch-based anatomical signatures to effectively capture the relationship between the CT and MR images. Reconstructed PET images using the PCT exhibit errors well below accepted test/retest reliability of PET/CT indicating high quantitative equivalence. [ABSTRACT FROM AUTHOR]

Currently, clinical brain single photon emission computed tomography (SPECT) is mostly performed using rotating dual-head gamma cameras equipped with low-energy-high-resolution parallel-beam collimators (LEHR PAR). The resolution of these systems is rather poor (8-10 mm) and the rotation of the heavy gamma cameras can introduce misalignment errors. Therefore, we designed a static full-ring multi-lofthole brain SPECT insert for an existing ring of LaBr3 (5% Ce) detectors. The novelty of the design is found in the shutter mechanism that makes the system very flexible and eliminates the need for rotating parts. A stationary SPECT insert is not only more robust, it is also easier to integrate in a magnetic resonance imaging system (MRI) for simultaneous SPECT-MRI. The target spatial resolution of our design is 6 mm. In this study we used analytical calculations to optimize the collimator for an existing ring of LaBr3 (5% Ce) detectors. We fixed the target spatial resolution at 6 mm in the center of the field-of-view and maximized the volume sensitivity by changing the collimator radius, the aperture and the number of loftholes. Based on these optimal parameters we simulated phantom data and evaluated the image quality of our multi-lofthole system. We simulated a noiseless uniform and Defrise phantom to assess artifacts and sampling completeness and a noiseless hot-rod phantom to assess the reconstructed spatial resolution. We visually evaluated a simulated noisy Hoffman phantom with two lesions. Then, we evaluated the non-prewhitening matched filter signal-to-noise ratio (NPW-SNR) in two lesion detectability phantoms: one with hot lesions and one with cold lesions. Finally, a contrast-to-noise (CNR) study was performed on a phantom with both hot and cold lesions of different sizes (6-16 mm). All results were compared to a LEHR PAR system. The optimization resulted in a final collimator design with a volume sensitivity of 1.55 × 10-4 cps Bq-1, which is 2.5 times lower than the sensitivity of a dual-head system with LEHR PAR collimators. Spatial resolution, on the other hand, has clearly improved compared to LEHR PAR: with the multi-lofthole system we successfully reconstructed 4 mm hot rods. Although this improved resolution did not result in an unambiguous improvement in CNR or NPWSNR, we believe that the flexibility of the shutter mechanism opens interesting perspectives toward time-multiplexing and integration with MRI. [ABSTRACT FROM AUTHOR]

An advantage of semiconductor-based dedicated cardiac single photon emission computed tomography (SPECT) cameras when compared to conventional Anger cameras is superior energy resolution. This provides the potential for improved separation of the photopeaks in dual radionuclide imaging, such as combined use of 99mTc and 123I . There is, however, the added complexity of tailing effects in the detectors that must be accounted for. In this paper we present a model-based correction algorithm which extracts the useful primary counts of 99mTc and 123I from projection data. Equations describing the in-patient scatter and tailing effects in the detectors are iteratively solved for both radionuclides simultaneously using a maximum a posteriori probability algorithm with one-step-late evaluation. Energy window-dependent parameters for the equations describing in-patient scatter are estimated using Monte Carlo simulations. Parameters for the equations describing tailing effects are estimated using virtually scatter-free experimental measurements on a dedicated cardiac SPECT camera with CdZnTe-detectors. When applied to a phantom study with both 99mTc and 123I, results show that the estimated spatial distribution of events from 99mTc in the 99mTc photopeak energy window is very similar to that measured in a single 99mTc phantom study. The extracted images of primary events display increased cold lesion contrasts for both 99mTc and 123I. [ABSTRACT FROM AUTHOR]