Identifying Mild Alzheimer's Disease With First 30-Min 11C-PiB PET Scan
Chushu Shen,Zhenguo Wang,Hongzhao Chen,Yan Bai,Xiaochen Li,Dong Liang,Xin Liu,Hairong Zheng,Meiyun Wang,Yongfeng Yang,Haifeng Wang,Tao Sun
DOI: https://doi.org/10.3389/fnagi.2022.785495
IF: 4.8
2022-04-05
Frontiers in Aging Neuroscience
Abstract:Introduction 11 C-labeled Pittsburgh compound B ( 11 C-PiB) PET imaging can provide information for the diagnosis of Alzheimer's disease (AD) by quantifying the binding of PiB to β-amyloid deposition in the brain. Quantification index, such as standardized uptake value ratio (SUVR) and distribution volume ratio (DVR), has been exploited to effectively distinguish between healthy and subjects with AD. However, these measures require a long wait/scan time, as well as the selection of an optimal reference region. In this study, we propose an alternate measure named amyloid quantification index (AQI), which can be obtained with the first 30-min scan without the selection of the reference region. Methods 11 C-labeled Pittsburgh compound B PET scan data were obtained from the public dataset “OASIS-3”. A total of 60 mild subjects with AD and 60 healthy controls were included, with 50 used for training and 10 used for testing in each group. The proposed measure AQI combines information of clearance rate and mid-phase PIB retention in featured brain regions from the first 30-min scan. For each subject in the training set, AQI, SUVR, and DVR were calculated and used for classification by the logistic regression classifier. The receiver operating characteristic (ROC) analysis was performed to evaluate the performance of these measures. Accuracy, sensitivity, and specificity were reported. The Kruskal–Wallis test and effect size were also performed and evaluated for all measures. Then, the performance of three measures was further validated on the testing set using the same method. The correlations between these measures and clinical MMSE and CDR-SOB scores were analyzed. Results The Kruskal–Wallis test suggested that AQI, SUVR, and DVR can all differentiate between the healthy and subjects with mild AD ( p < 0.001). For the training set, ROC analysis showed that AQI achieved the best classification performance with an accuracy rate of 0.93, higher than 0.88 for SUVR and 0.89 for DVR. The effect size of AQI, SUVR, and DVR were 2.35, 2.12, and 2.06, respectively, indicating that AQI was the most effective among these measures. For the testing set, all three measures achieved less superior performance, while AQI still performed the best with the highest accuracy of 0.85. Some false-negative cases with below-threshold SUVR and DVR values were correctly identified using AQI. All three measures showed significant and comparable correlations with clinical scores ( p < 0.01). Conclusion Amyloid quantification index combines early-phase kinetic information and a certain degree of β-amyloid deposition, and can provide a better differentiating performance using the data from the first 30-min dynamic scan. Moreover, it was shown that clinically indistinguishable AD cases regarding PiB retention potentially can be correctly identified.
neurosciences,geriatrics & gerontology,gerontology