Myocardial perfusion images are the result of a complex reconstruction process. Although sophisticated reconstruction methods are available, including correction for motion, attenuation and scatter correction, these software tools cannot produce “miracles”. It is therefore important to achieve optimal quality of the raw data by selecting the proper matrix size, angular sampling, zoom factor, patient-to-camera distance, energy window settings, and assuring that the camera is properly tuned and maintained through regular quality control procedures. In addition, the acquired projection data should be checked for motion and the presence of high extra-cardiac uptake. This should be done before the patient leaves the department and before reconstruction is commenced.
Camera vendors and various third party companies provide reconstruction software that implement iterative reconstruction based on Ordered Subset Expectation Maximisation or maximum likelihood expectation maximization (OS-EM/ML-EM). The advantage of these algorithms over traditional filtered-back projection is that information about the camera, patient and radiopharmaceutical can be exploited to reconstruct better images. CT images can be incorporated for estimation of attenuation and scatter; the collimator-detector-response can be modelled and used for resolution recovery; noise can be compensated by modelling the underlying characteristics of the decay process.
These reconstruction methods can achieve enhanced image quality that may be traded against shorter acquisition times or reduced administered activity. Fundamental to all these algorithms is the correct choice of user-selectable parameters (typically number of iterations and subsets, regularisation, and filter parameters). Inadequate parameters most likely lead to insufficient image quality and artefacts. As implementations vary considerably across vendors, it is not possible to transfer settings between camera systems without prior validation.
Currently, all major vendors offer the possibility to include resolution recovery (also called count recovery) in the OS-EM/ML-EM algorithm. The increased reconstructed resolution and lower noise allow for slightly lower count statistics (hence lower injected activity) or shorter scan times [35 (link)]. However, such techniques require careful testing against phantom studies performed with validated hardware and software.
Camera vendors and various third party companies provide reconstruction software that implement iterative reconstruction based on Ordered Subset Expectation Maximisation or maximum likelihood expectation maximization (OS-EM/ML-EM). The advantage of these algorithms over traditional filtered-back projection is that information about the camera, patient and radiopharmaceutical can be exploited to reconstruct better images. CT images can be incorporated for estimation of attenuation and scatter; the collimator-detector-response can be modelled and used for resolution recovery; noise can be compensated by modelling the underlying characteristics of the decay process.
These reconstruction methods can achieve enhanced image quality that may be traded against shorter acquisition times or reduced administered activity. Fundamental to all these algorithms is the correct choice of user-selectable parameters (typically number of iterations and subsets, regularisation, and filter parameters). Inadequate parameters most likely lead to insufficient image quality and artefacts. As implementations vary considerably across vendors, it is not possible to transfer settings between camera systems without prior validation.
Currently, all major vendors offer the possibility to include resolution recovery (also called count recovery) in the OS-EM/ML-EM algorithm. The increased reconstructed resolution and lower noise allow for slightly lower count statistics (hence lower injected activity) or shorter scan times [35 (link)]. However, such techniques require careful testing against phantom studies performed with validated hardware and software.
Free full text: Click here
