Dynamic FDG-PET demonstration of functional brain abnormalities

Abstract

Positron emission tomography with fluorine-18 fluorodeoxyglucose (¹⁸ F-FDG-PET) has been used over 3 decades to map patterns of brain glucose metabolism to evaluate normal brain function or demonstrate abnormalities of metabolism in brain disorders. Traditional PET maps patterns of absolute tracer uptake but has demonstrated shortcomings in disorders such as brain neoplasm or focal epilepsy in the ability to resolve normally from pathological tissue. In this review, we describe an alternative process of metabolic mapping, dynamic PET. This new technology quantifies the dynamics of tracer uptake and decays with the goal of improving the functional mapping of the desired metabolic activity in the target organ. We discuss technical implementation and findings of initial pilot studies in brain tumor treatment and epilepsy surgery.

Figure 1

Differences between static and dynamic PET. In static PET, the absolute amount of radiation is mapped voxel‐by‐voxel during an acquisition window starting at a fixed time after completion of tracer injection. In this example, voxels A, B, and C end up with the same absolute radiation counts, measured as the standardized uptake value (SUV). In dynamic PET, tracer is captured in time windows, and analysis of kinetics (in this case, measurement of peak tracer) facilitates differentiation among voxels despite having similar absolute SUVs in the static window. [Colour figure can be viewed at wileyonlinelibrary.com]

Figure 2

Dynamic PET (A) is preceded by a high‐resolution post‐contrast T1‐weighted (B) MPRAGE MRI (256 pixels × 256 pixels × 192 slices) using a Siemens 3 T scanner for co‐registration. PET images are obtained with a Siemens Biograph time of flight (TOF) mCT scanner. Dynamic acquisition consists of an intravenous ~10 mCi FDG tracer injection over 10 s at the start of a 60‐min scan in list‐mode format. Subsequent processing is performed with custom tools developed in Matlab (Mathworks Inc., Natick, MA) and performed using MRtrix functions. Motion correction for the 60‐min acquisition (C) is performed by averaging the first 14 frames of PET data (400 pixels × 400 pixels × 111 slices × 38‐time frames) to create a reference for a rigid body transform across subsequent frames. Motion corrected PET frames are resliced and co‐registered with T1‐weighted MRI using non‐rigid transform to generate a transformation matrix used, in turn, to generate a co‐registered dynamic PET. Next, the MRI is co‐registered with a high‐resolution T1‐weighted MRI template provided by the Montreal Neurological Institute (MNI). (D and E) using a non‐rigid transform, and a transformation matrix is generated. The total 164 regions of the Destrieux atlas defined on the same MR brain template are binned to generate regions of interest. The transformation matrix is inverted and applied to all regions of interest to move them from the MNI template into the patient MRI. The image is calibrated with the model‐corrected blood input function (MCIF) (F). Each voxel of the dynamic PET volume is then independently fed into a 4 parameter 3‐compartment model or graphical Patlak model together with the MCIF. By analyzing millions of voxels across the entire brain PET volume, a parametric kinetic map (G) is computed. On a practical note, such iterative computing is only feasible with the use of parallel computing techniques. The above regions generated in patient MR space are then applied to the computed parametric PET maps in Matlab to obtain kinetic or z‐score anatomic maps (H). [Colour figure can be viewed at wileyonlinelibrary.com]

Figure 3

Dynamic FDG‐PET in brain tumor treatment. (A) MRI provides anatomic structure of the tumor which, (B) when co‐registered upon dynamic FDG‐PET with the locations and shapes of tumors, provides local measures of glucose kinetics within each tumor. Preliminary work established that tumor‐site‐associated glucose kinetics can help create statistical models that better differentiate tumor progression from radiation‐induced inflammation.

Figure 4

T1‐weighted MRI (left side images) and co‐registered dynamic FDG‐PET maps of hypometabolism by z score (right images) from epilepsy patients whose standard, static FDG‐PET images were normal. (A) This patient underwent a previous right anterior frontal lobotomy with no improvement in seizure frequency. iD‐PET demonstrated relative right hippocampal hypometabolism; subsequent intracranial monitoring confirmed seizure onset in the right hippocampus. (B) This patient had evidence of bilateral mesial temporal lobe epilepsy. iD‐PET disclosed worse right hippocampal hypometabolism; this patient declined further surgical intervention. (C) Patient with normal MRI and poorly localizing seizures with worse right hippocampal hypometabolism. This patient underwent thermal ablation of the right hippocampus and has long‐term seizure remission.