Cheon Miju, Yi Hyunkyung, Ha Sang-Won, Kang Min Ju, Jeong Da-Eun, Abdelhafez Yasser G, Nardo Lorenzo
Department of Nuclear Medicine, Veterans Health Service Medical Center, Seoul 05368, Republic of Korea.
Department of Neurology, Veterans Health Service Medical Center, Seoul 05368, Republic of Korea.
Diagnostics (Basel). 2025 Aug 13;15(16):2028. doi: 10.3390/diagnostics15162028.
: Quantitative analysis of amyloid PET imaging plays a crucial role in diagnosing Alzheimer's disease (AD), particularly in cases where visual interpretation is equivocal. Multiple commercial software tools are available for this purpose, yet differences in their quantification and diagnostic performance remain understudied, especially for Neurophet SCALE PET. : We retrospectively analyzed F-florbetaben PET/CT scans from 129 patients with cognitive impairment, comprising 39 patients with AD and 90 with non-AD diagnoses, using three software tools: MIMneuro, CortexID Suite, and Neurophet SCALE PET. Standardized uptake value ratios (SUVRs) were obtained for six brain regions known for amyloid accumulation. Diagnostic accuracy was evaluated using ROC curve analysis, while inter-software correlations and reliability were assessed via Pearson correlation coefficients and intraclass correlation coefficients (ICC). : All three software programs significantly distinguished AD from non-AD patients in most brain regions. MIMneuro and Neurophet SCALE PET demonstrated the highest diagnostic performance, with MIMneuro achieving an AUC of 1.000 in the anterior cingulate gyrus. While MIMneuro and Neurophet SCALE PET showed moderate-to-strong SUVR correlations (r = 0.715-0.865), CortexID Suite showed limited correlation with the other tools. Inter-software reliability was moderate only in selected regions (ICC ≈ 0.5), indicating potential variability in SUVR measurements across platforms. : MIMneuro, CortexID Suite, and Neurophet SCALE PET are effective for the semi-quantitative analysis of amyloid PET and can aid in the diagnosis of AD. However, clinicians should be cautious when interpreting SUVRs across different software tools due to limited inter-software consistency. Standardization efforts or consistent use of a single platform are recommended to avoid diagnostic discrepancies.
淀粉样蛋白PET成像的定量分析在阿尔茨海默病（AD）的诊断中起着关键作用，特别是在视觉解读不明确的情况下。为此有多种商业软件工具可用，但它们在定量和诊断性能方面的差异仍未得到充分研究，尤其是对于Neurophet SCALE PET。我们回顾性分析了129例认知障碍患者的¹⁸F-氟贝他班PET/CT扫描，其中包括39例AD患者和90例非AD诊断患者，使用了三种软件工具：MIMneuro、CortexID Suite和Neurophet SCALE PET。获得了六个已知有淀粉样蛋白积累的脑区的标准化摄取值比率（SUVRs）。使用ROC曲线分析评估诊断准确性，同时通过Pearson相关系数和组内相关系数（ICC）评估软件间的相关性和可靠性。所有三个软件程序在大多数脑区都能显著区分AD患者和非AD患者。MIMneuro和Neurophet SCALE PET表现出最高的诊断性能，MIMneuro在前扣带回的AUC达到1.000。虽然MIMneuro和Neurophet SCALE PET显示出中度到强的SUVR相关性（r = 0.715 - 0.865），但CortexID Suite与其他工具的相关性有限。软件间的可靠性仅在选定区域为中度（ICC≈0.5），表明跨平台的SUVR测量可能存在变异性。MIMneuro、CortexID Suite和Neurophet SCALE PET对淀粉样蛋白PET的半定量分析有效，有助于AD的诊断。然而，由于软件间一致性有限，临床医生在解释不同软件工具的SUVR时应谨慎。建议进行标准化工作或始终使用单一平台以避免诊断差异。
