Clinical visual readings of brain region-specific hypometabolism in cognitive impairment patients is independent of attenuation correction method for integrated PET/MR

1977 Objectives PET/MR may be used in the evaluation of cognitively impaired patients. There are known quantitative differences between PET images obtained on PET/MR scanners when reconstructed with Dixon-MR, CT-based or atlas-based attenuation correction (AC) maps. This study seeks to assess the impact, if any, of these three-different AC methods on the blinded visual interpretation of regional hypometabolism in patients with cognitive impairment. Methods Forty-five minutes following injection of 10 mCi of FDG, 15 patients with cognitive impairment underwent brain PET/CT. PET/MR scanning with a 10 minute PET acquisition and Dixon MR imaging was subsequently performed on a Siemens Biograph mMR scanner under an IRB-approved protocol, at approximately two hours post-injection. A manufacturer-provided non-product offline reconstruction tool was used to reconstruct PET data obtained from PET/MR with AC based on the patient’s own CT images, a Dixon-MR derived AC map and an atlas-based AC map that combined Dixon-MR with a segmentation of bony skull structures. Two nuclear medicine physicians blindly scored brain regions (frontal, temporal, parietal, occipital, precuneus) as normal versus hypometabolic using 2D and 3D images generated by MIM software. Abnormal regions were scored as mild, moderate, or severely hypometabolic (score of 0, 1, 2 or 3 respectively). The hypometabolism scores obtained using the different methods of AC were compared and reader agreement assessed. All statistical tests were conducted at the two-sided 5% significance level using SAS 9.3 (SAS Institute, Cary, NC). Results Regional hypometabolism versus normal metabolism was correctly classified (accuracy) for 150 regions in 15 patients by two readers on atlas- and Dixon-based AC map PET reconstructions (versus CT reference AC) for 94% (90 - 96% c.i.) and 93% (89 - 96% c.i.) of all regions. The averaged sensitivity/specificity for detection of any regional hypometabolism was 95%/94% and 90%/91% for atlas-based and Dixon-based AC maps, respectively, compared to the reference standard CT images. The mean absolute error of regional hypometabolism scores for atlas- and Dixon-based PET reconstructions (versus CT) was 0.25 +/- 0.44 and 0.21 +/- 0.42 . There were no statistically significant differences between the visual assessments. Intra-reader agreement for detection of regional hypometabolism was high, with similar outcome assessments when using atlas- and Dixon-corrected PET data in 93% and 93% of scored regions, respectively. The simple kappa coefficient to assess reader agreement in terms of hypometabolism versus normal regions was 0.82 for atlas- and 0.84 for Dixon-based AC. The weighted kappa coefficient to assess reader agreement in terms of the hypometabolism score was 0.75 for atlas- and 0.77 for Dixon-AC. Conclusions Despite the more accurate FDG SUV quantification with CT-based and atlas-based attenuation correction in brain PET/MR compared to Dixon AC, there were no measureable differences between the three AC methods with respect to visual identification of regional hypometabolism in the evaluation of cognitively impaired patients.