30 Healthy Elderly Controls and 62 MCI patients: SUVr comparison between PMOD 3.2 and PNEURO 3.5 analysis.
(2016) SNMMI-A-538-2016 — Location: San Diego, USA (11.June.2016)
Files
SNMMI2016_OP513515.pdf
Open Access - Adobe PDF
- 916.88 KB
Details
- Authors
- Abstract
- OBJECTIVES: Although visual analysis of Amyloid PET has been considered sufficient to discriminate Alzheimer’s patients from healthy elderly subjects, early stage of the degenerative process (MCI stade), exhibiting fewer amyloid deposits, may be more challenging to interpret. In this situation, a robust quantitative analysis may not only be helpful to reinforce the visual analysis but also to allow longitudinal evaluation of the amyloid load in this population. METHODS: As part of a academic study evaluating the impact of biomarkers to stratify MCI patients, 30 healthy controls (70±5.5 y) and 62 (70.8±8.0 y) patients presenting Mild Cognitive Impairement (MCI) were imaged 90 min PI with F18-Flutemetamol scan (GE Healthcare; target ID 185 MBq). SUVr (neocortex/cerebellum) values computed with step-by-step analysis based on several PMOD 3.2 tools (see previously published data) were compared to those obtained with a more recent evolution of the PMOD software (PNEURO 3.5) implementing several useful improvements for clinical translation (semi-automatic workflow, more accurate MRI segmentation for VOI delineation). However, those methodological changes could also potentially influence the results, reason why a front-to-front comparison was performed. RESULTS: PMOD 3.2 and PNEURO 3.5 SUVr (Mean±SD/Median/PC90) of the control group were respectively 1.40±0.23/1.33/1.69 and 1.29±0.16/1.26/1.49, both significantly different from the MCI group analysis (1.67±0.39/1.57(P=0.002) and 1.55±0.35/1.42 (p<0.0001)). Despite a good correlation (r2: 0.95), pooled SUVr PNEURO 3.5 values were significantly lower than the previously computed PMOD 3.2 values (P< 0.0001; bias(d): -0.12, SD of bias: 0.09; 95% limits of agreement: from -0.29 to 0.06). The SUVmean of the cerebellar region of reference (Mean±SD/Median) was not statistically different between controls and MCI: 0.59±0.15/0.59 vs 0.63±0.15/0.61 (P=0.28) for PMOD 3.2 and 0.63±0.14/0.64 vs 0.67±0.14/0.65 (P=0.42) for PNEURO 3.5. However, a significant difference was observed (P<0.0001) between the two methods of analysis, with slightly higher cerebellar SUVmean values using PNEURO 3.5 (bias: 0.04; SD of bias: 0.04; 95% limits of agreement: from -0.03 to 0.12). PC90 SUVr of the control group used as discriminating cutoff to separate positive from negative amyloid scans, perfectly correlates with the binary visual analysis scans in all cases, whatever the method used. Finally, SUVr workflow reproducibility using PNEURO 3.5 was also tested between two independent observers for n=24 scans selected ad random, with excellent results (bias: 0.002; SD of bias:0.012; 95% limits of Agreement, from -0.021 to 0.026 (-1.22 to 1.53%)). CONCLUSION: SUVr computation based of PNEURO 3.5 seems a easy way to quantify the neocortical amyloid load in MCI patients, providing similar results than the visual scan reading but with the advantage to be independent of the physician’s degree of expertise.
- Affiliations
Citations
Lhommel, R., Gerard, T., Hanseeuw, B., & Ivanoiu, A. (2016). 30 Healthy Elderly Controls and 62 MCI patients: SUVr comparison between PMOD 3.2 and PNEURO 3.5 analysis. The Journal of Nuclear Medicine, 57(supplement 2), OP513. https://hdl.handle.net/2078.5/236487 (Original work published 2016)