Resolution recovery and noise regularization in nuclear cardiology

Radionuclide myocardial perfusion imaging (MPI) is one of the most established and validated diagnostic modalities in the evaluation of patients with hemodynamically significant coronary artery disease (CAD). Growing public awareness and media attention toward the radiation exposure raised appropriate concerns related to the potential harmful effects of the radionuclide use. In 2010, American Society of Nuclear Cardiology (ASNC) issued a statement titled ‘‘Recommendations for reducing radiation exposure in myocardial perfusion imaging’’ addressing these issues. The use of appropriateness criteria for MPI, decreased radiotracer activity administration, utilization of newer solid state cameras, and novel image reconstruction are some of the ways to diminish ionizing radiation exposure in patients and healthcare professionals. This article will address image reconstruction possibilities for myocardial perfusion imaging (Figures 1, 2). The traditional method of image reconstruction in nuclear cardiology has been filtered backprojection (FPB). Its drawbacks are relatively lesser quality, lowcount SPECT (single-photon emission computerized tomography) images related to amplified noise, and inability to correct for photon attenuation and scatter. This reconstruction algorithm does not compensate for the limitations related to the detector, such as collimator geometry and distance from the emitting source. FBP images are prone to blurring and star artifacts, requiring filtering prior to backprojection, potentially decreasing imaging resolution. Low-count images in FBP preclude a further decrease in radiopharmaceutical administration. This limits the ability to diminish the effective radiation dose to the patient as a result, requiring longer acquisition times, which may lead to motion artifacts due to poor patient cooperation. Ideally, the reconstruction method should allow the lowest amount of radiopharmaceutical without loss of image quality and diagnostic accuracy. Following significant improvements in the computational processing power, iterative reconstruction, namely maximal likelihood expectation maximization (MLEM) technique, replaced FBP over a decade ago. Another addition to the iterative method was an introduction of ordered subset expectation maximization (OSEM). It uses a subset of the data at each iteration, producing a faster rate of conversion. OSEM reconstructions allow the detector variables to be incorporated, also attenuation correction maps, scatter, and the variation between source and detector positions. Resolution recovery reconstruction method addresses noise reduction and resolution recovery simultaneously for low-count density data. A statistically based noise suppression algorithm is applied to the images, allowing quality improvement. At the present time, there are several iterative reconstruction packages available on the market: Evolution (GE healthcare, Milwaukee, WI, USA), widebeam reconstruction (UltraSPECT, Haifa, Israel), Astonish (Phillips), Flash 3D (Siemens), and nSPEED (Digirad), all offering resolution recovery and noise reduction. Wide-beam reconstruction incorporates resolution recovery and controls noise during the reconstruction process. The modification is done without applying the post-processing filter and is a great adjunct tool for the low-count statistics studies with inherently higher noise level. It is an iterative reconstruction method designed to focus on resolution recovery and noise reduction at the same time, therefore helping improve image quality in studies with significantly less photon counts. This technology incorporates the physics and geometry modeling of the emission and detection process. There are several factors such modeling focuses on. First, pixel-voxel weighting values are aligned to the Reprint requests: Olga James, MD, Duke University Medical Center, Durham, NC; olga.g.james@dm.duke.edu J Nucl Cardiol 2017;24:138–41. 1071-3581/$34.00 Copyright 2016 American Society of Nuclear Cardiology.