Skip to main content
EJNMMI Physics
Fig. 5 | EJNMMI Physics

Fig. 5

From: Combining deep learning with a kinetic model to predict dynamic PET images and generate parametric images

Fig. 5

Example of generated parametric Ki images obtained from different planes (the transverse plane, coronal plane and sagittal plane). The Ki images of both the ground truth and our proposed method were obtained by fitting the last 13 frames of the Patlak plot’s data points with linear regression. The main differences between both methods are that the data points from the last 30 min were generated by the proposed network rather than being real. The Ki images of the method without the kinetic model in the lower right were generated directly rather than by fitting a Patlak plot. The images in the upper right were obtained by fitting only the first 30 min of frames

Back to article page