3335

Validation of an Image Derived Input Function Method for 15O-water PET/MR Brain Scans
Mohammad Mehdi Khalighi1, Audrey Peiwen Fan1, Mathias Engström2, Lieuwe Appel3, Gunnar Antoni4, Frederick T Chin1,5, Eva Kumlien6, Greg Zaharchuk1,5, and Mark Lubberink7

1Radiology, Stanford, Stanford, CA, United States, 2Applied Science lab, GE Healthcare, Uppsala, Sweden, 3PET Centre, Medical Imaging Centre, Uppsala University Hospital, Uppsala, Sweden, 4Medicinal Chemistry, Uppsala University, Uppsala, Sweden, 5Molecular Imaging Program, Stanford, Stanford, CA, United States, 6Neuroscience, Uppsala University, Uppsala, Sweden, 7Medical Physics, Uppsala University, Uppsala, Sweden

Synopsis

A recently introduced image-derived input function (IDIF) method on a PET/MR scanner addresses the spill-in and spill-over artifacts on the PET images by measuring the true carotid artery volume by an MR-angiogram. This study validates the IDIF method for quantification of cerebral blood flow (CBF) from 15O-water PET, using arterial blood sampling as the gold standard in 20 subjects. CBF measured by IDIF and BSIF were correlated (R2= 0.5) in the gray-matter and whole-brain, with average difference of only 3%.

Purpose

Recently a new image derive input function (IDIF) method [1] has been introduced on a PET/MRI scanner, which directly measures and corrects for spill-in and spill-out artifacts, using simultaneous MR and PET angiograms. In this study we have compared this new IDIF method with the arterial blood sampling input function (BSIF) method, which is considered as a gold standard, in 20 subjects who went under a PET/MR brain scan with 15O-water. We then performed kinetic modeling with each input function to estimate and compare gray-matter and whole-brain cerebral blood flow (CBF) from each input function method.

Methods

Data from 20 subjects (11 male and 9 female subjects; age: 38 ± 12.5 years; weight: 79.5 ± 14.5 kg; 10 focal epilepsy patients and 10 healthy controls) included in a PET/MR study on epilepsy were used in the present work. The study was approved by the Regional Ethical Review Board in Uppsala and all subjects provided written consent prior to inclusion in the study. Subjects were injected with 400 ±64 MBq of 15O-water using a power injector simultaneous with the start of a 10 min list-mode PET scan on a time-of-flight (TOF) enabled PET/MR scanner [2,3] (SIGNA, GE Healthcare, Waukesha, WI). For each patient, blood was continuously drawn from the radial artery at the wrist and the tracer concentration in the arterial blood was measured using a Twilite two detector (Swisstrace) to estimate the BSIF.

T1-weighted and MR angiographic (MRA) images were acquired for each subject, with the following parameters: a sagittal 3D GRE with 8.6 ms repetition time (TR), 3.2 ms echo time (TE), 450 ms inversion time (TI), flip angle 12°, 25×23.8 cm field of view (FOV), matrix 256×256, 178 slices with 1 mm thickness, and scan duration of 4:30 min; non-contrast time-of-flight (TOF) with 120 slices, TR=20 ms, TE=2 ms, flip angle 15°, FOV=22×19.4 cm, matrix 256×256, slice thickness 1.2 mm, and scan duration of 2:29 min. The coverage of the MRA was from above the circle of Willis to the mid-cervical region (~70) mm.

The PET list file was started right after the injection and the first 20 seconds were skipped to account for tracer arrival to the carotid arteries. The PET images were dynamically reconstructed with thirty 1-s frames, thirty 3-s frames, twelve 5-s frames, twelve 10-s frames, followed by eight 30-sec frames. The reconstruction was done with TOF-OSEM, 2 iterations, 28 subsets, using attenuation, scatter, random, and dead-time corrections. A PET angiogram (PETA) was reconstructed over the first pass of arterial input function (AIF) [1]. Using MRA and PETA images, two masks were formed to measure the volume of arteries and the extent of spill-over artifacts, respectively, and the IDIF was calculated by dividing the total number of counts (obtained from PETA) by the arterial volume (obtained from MRA) after taking the spill-in into account (Figure 1).

Results

Figure 2 shows a visual comparison between the IDIF and BSIF input functions in 4 subjects. Top row shows IDIF and BSIF for 2 subjects with the best visual agreement between the two input functions. In these cases, the peak amplitude, peak time, FWHM, and the tail of input functions are closely matched. The bottom row shows 2 typical subjects where BSIF input functions are generally noisier than IDIF input functions. Figure 3-a shows the average of IDIF and BSIF input functions over all 20 subjects. Although the peak amplitude of BSIF is lower than IDIF input function, the area of the first pass shown by dashed lines are almost identical (<0.1% difference). Figure 3-b shows the average of IDIF and BSIF input functions with their corresponding standard deviation. As expected, BSIF input functions have a higher standard deviation due to the noisy nature of the sampler data. Figure 4 shows the correlation between CBF measurements for gray matter and whole brain using IDIF and BSIF input functions as well as their corresponding Bland-Altman plots. The two input functions provided correlated CBF measurements with an R2 value of 0.5 and the Bland-Altman plots show that all the data points are within mean ±1.96 standard deviation window.

Conclusion

The image-derived input function estimation (IDIF) was validated by blood sampled input function (BSIF), which is considered to be the gold standard. The area under the first pass of the IDIF is the same as that of the BSIF after averaging over 20 subjects and there was a strong correlation (R2= 0.5, slope = 0.9) between CBF measured by BSIF and IDIF for gray matter and whole brain in 20 subjects. Based on this comparison, we suggest that the IDIF method is a promising method for measuring CBF.

Acknowledgements

GE Healthcare, Uppsala University, Swedish Research Council and Lucas foundation.

References

[1] Khalighi MM, Deller TW, Fan AP, Gulaka PK, Shen B, Singh P et al. Image-derived input function estimation on a TOF-enabled PET/MR for cerebral blood flow mapping. Journal of Cerebral Blood Flow & Metabolism. 2018; 38 (1):126-135

[2] Levin CS, Maramraju HS, Khalighi MM, Deller TW, Delso G, Jansen F. Design Features and Mutual Compatibility Studies of the Time-of-Flight PET Capable GE SIGNA PET/MR System. IEEE TMI. 2016; 35(8):1907-1914.

[3] Grant A, Deller TW, Khalighi MM, Maramraju SH, Delso G, Levin CS. NEMA NU 2-2012 performance studies for the SiPM-based ToF-PET component of the GE SIGNA PET/MR system, Med. Phys. 2016; 43:2334.

Figures

Figure 1: IDIF method summary. Top row shows the MR an d PET angiograms with their corresponding masks. Bottom row shows the 1-compartmental kinetic modeling along with the corresponding equations [1], used for CBF (i.e. K1) estimation.

Figure 2: IDIF and BSIF input functions in 4 subjects are shown. The top row shows 2 cases with best agreement between IDIF and BSIF input functions. The bottom row shows comparisons between IDIF and BSIF input functions in 2 typical subjects.

Figure 3: (a) Average of IDIF and BSIF input functions over all 20 subjects. The area of the both IDIF and BSIF input functions within the dashed lines is the same. (b) The average of IDIF and BSIF input functions for all 20 subjects is shown with their corresponding standard deviation.

Figure 4: Top row showsthe correlation between CBF measurements for gray matter and whole brain using IDIF and BSIF input functions and the bottom row shows their Bland-Altman plot with all data points within mean ±1.96 standard deviation window.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)
3335