Darrell Dennis Burckhardt

Validation of Software Gating: A Practical Technology for Respiratory Motion Correction in PET

Purpose To assess the performance of hardware- and software-gating technologies in terms of qualitative and quantitative characteristics of respiratory motion in positron emission tomography (PET) imaging. Materials and Methods Between 2010 and 2013, 219 fluorine 18 fluorodeoxyglucose PET examinations were performed in 116 patients for assessment of pulmonary nodules. All patients provided informed consent in this institutional review board-approved study. Acquisitions were reconstructed as respiratory-gated images by using hardware-derived respiratory triggers and software-derived signal (via an automated postprocessing method). Asymmetry was evaluated in the joint distribution of reader preference, and linear mixed models were used to evaluate differences in outcomes according to gating type. Results In blind reviews of reconstructed gated images, software was selected as superior 16.9% of the time (111 of 657 image sets; 95% confidence interval [CI]: 14.0%, 19.8%), and hardware was selected as superior 6.2% of the time (41 of 657 image sets; 95% CI: 4.4%, 8.1%). Of the image sets, 76.9% (505 of 657; 95% CI: 73.6%, 80.1%) were judged as having indistinguishable motion quality. Quantitative analysis demonstrated that the two gating strategies exhibited similar performance, and the performance of both was significantly different from that of nongated images. The mean increase ± standard deviation in lesion maximum standardized uptake value was 42.2% ± 38.9 between nongated and software-gated images, and lesion full width at half maximum values decreased by 9.9% ± 9.6. Conclusion Compared with vendor-supplied respiratory-gating hardware methods, software gating performed favorably, both qualitatively and quantitatively. Fully automated gating is a feasible approach to motion correction of PET images. (©) RSNA, 2016 Online supplemental material is available for this article.

Automatic registration of cardiac PET/CT for attenuation correction

Misalignments of images in cardiac Positron Emission Tomography (PET)-CT imaging may lead to erroneous Attenuation Correction (AC) and mis-diagnosis. Such misalignment may be corrected manually prior to reconstruction and clinical assessment; however this step is laborious and may be subject to operator variability. The aim of this study is to assess the performance of an algorithm to automatically align CT to PET prior to AC. We conclude that automatic registration is a viable option for the task of aligning cardiac CT and PET for AC, with a consistency comparable to that of using manual alignment.

Frequency based gating: An alternative, conformal, approach to 4D PET data utilization

PURPOSE Respiratory gating is a strategy for overcoming image degradation caused by patient motion in Positron Emission Tomography (PET) imaging. Traditional methods for sorting data, namely, phase-based gating or amplitude-based gating, come with an inherent trade-off between resolution improvements and added noise present in the subjugated data. If the goal of motion correction in PET is realigned from creating 4D images that attempt to mimic nongated images, towards ideal utilization of the information available, then new paths for data management emerge. In this work, the authors examine the application of a method in a new class of frequency based data subjugation algorithms, termed gating +. This strategy utilizes data driven information to locally adapt signal to its optimal segregation, thereby creating a new approach to 4D data utilization PET. METHODS 189 (18)F-fluorodeoxyglucose (FDG) PET scans were acquired at a single bed position centered on the thorax region. 4D gated image sets were reconstructed using data driven gating. The gating+ signal optimization algorithm, previously presented in small animal PET images and simulations, was used to segregate data in frequency space to generate optimized 4D images in the population-the first application and analysis of gating+ in human PET scans. The nongated, phase gated, and gating+ representations of the data were compared using FDG uptake analysis in the identified lesions and noise measurements from background regions. RESULTS Optimized processing required less than 1 min per scan on a standard PC (plus standard reconstruction time), and yielded entire 4D optimized volumes plus motion maps. Optimized scans had noise characteristics similar to nongated images, yet also contained much of the resolution and motion information found in the gated images. The average SUVmax increase in the lesion sample between gated/nongated and gating+/nongated (±SD in population) was 35.8% ± 34.6% and 28.6% ± 27.9%, respectively. The average percent standard deviation (%SD ± SD in population) in liver volumes of interest (VOIs) across the sample for the nongated, gated, and gating+ scans was 6.7% ± 2.4%, 13.6% ± 3.3%, and 7.1% ± 2.5%, respectively. In all cases, the noise in the gating+ liver VOIs was closer to the nongated measurements than to the gated. CONCLUSIONS The gating+ algorithm introduces the notion of conforming 4D data segregation to the local information and statistics that support it. By segregating data in frequency space, the authors are able to generate low noise motion information rich image sets, derived solely from selective use of raw data. Their work shows that the gating+ algorithm can be robustly applied in populations, and across varying qualities of motion and scans statistics, and be integrated as part of a fully automated motion correction workflow. Furthermore, the idea of smart signal utilization underpins a new concept of low risk or even risk-free motion correction application in PET.

SU‐FF‐T‐217: Evaluation and Validation of GATE‐Based Absorbed Dose Calculation for 3D Patient‐Specific Internal Dosimetry

Purpose: The objective of this study is to validate the use of GATE Monte Carlo simulations to determine patient‐specific dosimetry using quantitative multi‐modality imaging.Method and Materials: Data acquired with the Symbia ™ hybrid SPECT/CT system are reconstructed with Flash‐3D™ and corrected for scatter, collimator detector response and attenuation using a CT attenuation correction method. The voxel‐based CT model is used with GATE to compute 3D dose distributions from the SPECT data. Validation is performed using both simulated SPECT data and anthropomorphic phantoms. Results: Initial validation studies show that GATE, while time consuming, is suitable in dose calculation. Conclusions: The approach presented can be used for radionuclide multi‐modality dosimetry leading to patient‐specific dose calculations for treatment planning.

Real-time MRI for assessment of PET/CT attenuation correction protocols

A geometrical mismatch between emission and transmission occurs in cardiac PET/CT because of respiratory and cardiac motions. Proposed solutions to this problem include fast CT during free breathing, with or without image alignment, and slow CT. To study this problem, including the variability of human breathing patterns and compliance with instructions, we performed real-time FISP MRI measurements of two free-breathing volunteer subjects. We have developed a method for simulating PET and CT coronal images from these image sequences, and for locating the left cardiac free wall semi-automatically. The PET geometry represents an average over the whole MR examination. The simulated CT geometry represented 28.8 mm axial extent and speeds of 83 mm/sec (fast) and 12 mm/sec (slow), representing a Senation-64 CT scanner running at fast and slow settings. We considered 400 start times for each CT geometry. The free wall was located in each CT image and was compared with the location expected in PET. The error in a given CT scan depends on the geometry, i.e. fast or slow scan, and also on the state of breathing at the time of the scan. A more accurate result can be realized by selecting the best-aligned of two or three fast CT scans. The average errors in fast, slow, best of 2, and best of 3 fast CT scans are respectively 5, 4, 4, and 3 mm. The errors in these scans have 98% likelihood of being less than, respectively, 14, 9, 9, and 7 mm.