Model calculations for PET images

Robust analysis techniques can be applied to dynamic PET images, producing parametric images.

Common pixel-by-pixel analysis methods

Models for reversible tracer uptake

Outcome is the volume of distribution, binding potential, or a related parameter.

Without blood sampling

• Simplified reference tissue model (SRTM) can be used to compute BP and R₁ images, using PMOD, or TPC software applying basis function approach (with imgbfbp) or multilinear model approach, solved with NNLS (using imgsrtm); the latter method is sensitive to noise, and thus smoothing would be recommended
• Logan analysis to produce DVR image (sensitive to noise, smoothing recommended)
• Late-scan tissue/reference tissue ratio (equals DVR in bolus+infusion studies)
• AUC tissue/reference tissue ratio (may correlate with BP or DVR)
• Standardized uptake value (SUV) image

With metabolite corrected plasma input

• Kinetic model fit to produce volume of distribution image (sensitive to noise, smoothing recommended)
• Logan analysis to produce volume of distribution image (sensitive to noise, smoothing recommended)
• Late-scan tissue/plasma ratio: may correlate with the volume of distribution
• Reversible one-tissue model can be used to compute K₁ and k₂ or K₁/k₂ images using either basis function approach (with imgbfk2) or multilinear model approach (with imglhk1).

Models for irreversible tracer uptake

Outcome is the K_i, FUR, k₃, or a related parameter, representing metabolic rate, or enzyme or transporter activity.

Without blood sampling

• Patlak analysis to produce parametric K_i image
• Standardized uptake value (SUV) image
• TRTM can be applied to compute k₃ image using program imgtrtm if a positive reference region TAC is available.

With metabolite corrected plasma input

• Patlak analysis to produce parametric K_i image (e.g. glucose uptake)
• Fractional uptake rate (FUR) image
• Late-scan tissue/plasma ratio: may correlate with Ki
• 2- or 3-CM fit to produce K_i image
• 3-CM fit to produce K₁ and k₃ image, or λ*k₃ image with multilinear fit or with Fowler & Logan method

Perfusion (blood flow)

Without blood sampling

• Early-scan tissue/reference tissue ratio (may correlate with the patterns of blood flow)

With arterial blood sampling

• Autoradiography (ARG) method for bolus [¹⁵O]H₂O-PET studies
• Compartment model for bolus [¹⁵O]H₂O-PET studies using either basis function approach (with imgbfh2o) or multilinear model approach (with imgflow).

Miscellaneous utility software for image model calculation

• Correction of vascular blood radioactivity in dynamic image
• Subtraction of reference region TAC from a dynamic image
• Thresholding and reducing image noise.
• "Clustering" dynamic images, based on method suggested by M'hamed Bentourkia (2001).
• Simple arithmetic calculation for ECAT sinogram and image files
• Create SIF file
• Correlation (linear regression) between two parametric images (same head, different analysis method), or
• residual image (same patient, before and after stimulus): imgcorrl with option -resid

See also:

• Cardiac image analysis system
• Instruction by tracer
• Input function
• Correlation coefficient constrained parametric images
• Model calculations for sinograms
• Tools for processing image data
• Parametric images in presentations and reports
• Multilinear models

---

Literature

Bentourkia M. A flexible image segmentation prior to parametric estimation. Comput Med Imaging Graph. 2001; 25: 501-506. doi: 10.1016/S0895-6111(01)00016-7.

Feng DD, Wen L, Eberl S. Techniques for parametric imaging. In: Feng DD (ed.): Biomedical Information Technology. Elsevier, 2008, pp 137-163. ISBN: 9780080550725. doi: 10.1016/B978-012373583-6.50010-4.

Herholz K. Non-stationary spatial filtering and accelerated curve fitting for parametric imaging with dynamic PET. Eur J Nucl Med. 1988; 14: 477-484. doi: 10.1007/BF00252392.

Zhou Y, Huang SC, Bergsneider M, Wong DF. Improved parametric image generation using spatial-temporal analysis of dynamic PET studies. Neuroimage 2002; 15(3): 697-707. doi: 10.1006/nimg.2001.1021.

Tags: Image, Modeling, Analysis, NNLS, SRTM

---

Updated at: 2023-06-16
Created at: 2008-01-18
Written by: Vesa Oikonen, Kaisa Liukko