SPECT Imaging of 2-D and 3-D Distributed Sources with Near-Field Coded Aperture Collimation: Computer Simulation and Real Data Validation

Abstract

The imaging of distributed sources with near-field coded aperture (CA) remains extremely challenging and is broadly considered unsuitable for single-photon emission computerized tomography (SPECT). This study proposes a novel CA SPECT reconstruction approach and evaluates the feasibilities of imaging and reconstructing distributed hot sources and cold lesions using near-field CA collimation and iterative image reconstruction. Computer simulations were designed to compare CA and pinhole collimations in two-dimensional radionuclide imaging. Digital phantoms were created and CA images of the phantoms were reconstructed using maximum likelihood expectation maximization (MLEM). Errors and the contrast-to-noise ratio (CNR) were calculated and image resolution was evaluated. An ex vivo rat heart with myocardial infarction was imaged using a micro-SPECT system equipped with a custom-made CA module and a commercial 5-pinhole collimator. Rat CA images were reconstructed via the three-dimensional (3-D) MLEM algorithm developed for CA SPECT with and without correction for a large projection angle, and 5-pinhole images were reconstructed using the commercial software provided by the SPECT system. Phantom images of CA were markedly improved in terms of image quality, quantitative root-mean-squared error, and CNR, as compared to pinhole images. CA and pinhole images yielded similar image resolution, while CA collimation resulted in fewer noise artifacts. CA and pinhole images of the rat heart were well reconstructed and the myocardial perfusion defects could be clearly discerned from 3-D CA and 5-pinhole SPECT images, whereas 5-pinhole SPECT images suffered from severe noise artifacts. Image contrast of CA SPECT was further improved after correction for the large projection angle used in the rat heart imaging. The computer simulations and small-animal imaging study presented herein indicate that the proposed 3-D CA SPECT imaging and reconstruction approaches worked reasonably well, demonstrating the feasibilities of achieving high sensitivity and high resolution SPECT using near-field CA collimation.

Quantitative Comparison of Planar Coded Aperture Imaging Reconstruction Methods

“…Mainly four methods for real-time capable image reconstruction have been proposed within the last few decades. MURA Decoding (also called inverse filtering, or cross-correlation analysis): [3,4,6,18,20,21,30,31,47,48,54], Wiener Filtering [24,34], convolutional Maximum Likelihood Expectation Maximization (MLEM) reconstruction [38,39] and data-driven Deep Learning approaches [34,56,57]. Although the standard MLEM algorithm has also been investigated [28,36], it is not considered in this paper, because its reconstruction time of several hours makes it impractical for the application in mobile systems.…”

“…CAI has been investigated as alternative collimator in single photon emission computed tomography (SPECT): Experiments for imaging the bio-distribution of radioactively labeled compounds in small animals, have been carried out [4,6]. [18,30,47] analyzed the use of a coded aperture 𝛾-camera for the localization of sentinel lymph nodes and [38,39] examined the capability of imaging coldspot lesions in cardiac imaging. Other research fields like x-ray fluorescence spectroscopy [34] or nuclear decommissioning [15,21] analyzed the potential use of CAI as well.…”

Comparison of Low-Pass Filters for SPECT Imaging

“…The lower value of SNR of smaller hot regions as compared to the larger hot regions were recorded. This is because SNR is count statistics, region/lesion size, and depth sensitive [31,32]. Contrary, with the increase in cutoff frequency of the Hamming filter, an increase in SNR of 22.6 mm diameter hot region at 0.40 and 0.50 cycles/cm cutoff frequency was achieved.…”

An Innovative Concept of a 3D-Coded Aperture Imaging System Proposed for Early Breast Cancer Detection

“…The tumor was placed at half of the depth of the breast, and the diameter of the tumor varied from 8 mm to 3 mm. The M-MLEM, an iterative reconstruction algorithm, was used for image reconstruction [ 19 ]. For the 3D reconstruction of the breast phantom, the image slices are reconstructed with a mask and 90 degrees rotated mask, which is also the antimask.…”