Table 6.
Classification of attenuation correction methods.
| Category | Method | Definition | Benefit | Limitations | Examples |
|---|---|---|---|---|---|
| Transmission-less based | Uniform fit-ellipse method (UFEM) | Object outline is approximated as an ellipse around its edges. In order to generate the attenuation map, uniform attenuation is assigned Then, uniform attenuation is assigned inside the figure. | Quick and easy Functional for brain studies | Low precision Limited to homogeneous areas | [389] |
| Automated contour detection method (ACDM) | Edge-detection algorithms are used to generate the shape of the object. This allows convex shapes | It is independent of the specialist Functional for brain studies | Low precision but higher than UFEM Only for homogeneous areas | [390,391] | |
| Other methods | Several techniques fit here, such as algebraic reconstruction–based techniques (MLAA or MLACF) or machine-learning techniques | No anatomical information is needed | Possibility of cross-talk between emission data and attenuation map Virtually not used in clinical trials | [392], [393], [394], [395], [396] | |
| Transmission-based | Radionuclide transmission | Apply an external source (PET, SPECT or SPECT/PET) interleaving transmission and emission scanning | Available in most systems. Existence of complementary methods to reduce noise | Need to modify the obtained attenuation coefficients due to be energy-dependent Possibility of errors due to cross-talk of data | [397], [398], [399], [400] |
| CT transmission | It can be addressed by segmenting the CT regions and assigning linear attenuation coefficients to each tissue or transforming the CT image to the attenuation map associated with the radiotracer photon's energy (or a combination of both methods using different scale factors for bone and tissues) | Quick Low noise Good spatial resolution Functional for brain studies | Misregistration due to respiratory movements Erroneus uptake due to patient's possession of unnatural materials (e.g. prostheses) CT photonic energy usually differs from that of the radionuclide used for emission scanning | [401], [402], [403], [404], [405] | |
| MRI transmission | Segmentation-based techniques, where first the PET and MRI images are co-registered and then a segmentation technique is applied. Usually fuzzy is used to divide the image from two to five tissues, assigning to each tissue some attenuation coefficients Atlas-based techniques, where an MR template is used instead of multistep segmentation procedures | High precision Popular for brain studies | Total dependence on co-registration success in the patient's image | [370,[405], [406], [407], [408], [409] |