Manduukhai Badarchin, Youichi Otomi, Takayoshi Shinya, Hideki Otsuka, Yukiko Takaoka, Tomoki Matsushita, Tomoyasu Matsubara, Koji Fujita, Yukiko Tomioka, Masahito Nakataki, Yuishin Izumi, Shusuke Numata and Masafumi Harada : Comparative evaluation of CortexID and VIZCalc software in brain amyloid PET: a retrospective study of 116 cases, Annals of Nuclear Medicine, 2025.
(Summary)
Quantitative analysis of amyloid positron emission tomography (PET) is increasingly applied in clinical and research settings; however, its consistency across software platforms remains uncertain. This study aimed to compare standardized uptake value ratio (SUVr) measurements obtained from CortexID Suite and VIZCalc, to evaluate their concordance with expert visual assessment, and to assess the concordance of Centiloid values derived from VIZCalc with the visual reference. We retrospectively analyzed 116 patients who underwent 18F-flutemetamol PET at a single institution. SUVr values were calculated using both CortexID Suite and VIZCalc, while Centiloid values were derived from VIZCalc only. Visual assessments were performed by two nuclear medicine physicians. Correlations among indices were examined using Pearson's correlation. Agreement between SUVr values was assessed with Bland-Altman analysis. Agreement with the non-independent visual reference was evaluated using receiver operating characteristic (ROC) analysis, and areas under the curves (AUCs) were compared with DeLong's test. SUVr values from CortexID and VIZCalc were strongly correlated (r = 0.986, p < 0.001), with a small mean difference of + 0.0397. Both platforms showed high concordance with the non-blinded visual assessment (AUC: 0.991 for CortexID; 0.989 for VIZCalc). Centiloid values also showed high agreement with the visual reference (AUC: 0.994) and were strongly correlated with SUVr values (r = 0.975 for CortexID; r = 0.965 for VIZCalc, p < 0.001). No significant difference was observed between platforms (p = 0.84). CortexID Suite and VIZCalc demonstrated high concordance with the non-blinded visual assessment and showed consistent quantitative trends. Both platforms can be reliably applied for amyloid burden quantification, provided that software-specific characteristics are appropriately considered.