Synaptic density change in patients with MSA and its role in MSA diagnosis: A Positron Emission Tomography Study with 18F-SynVesT-1 and Machine Learning Analysis

Introduction: Multiple system atrophy (MSA) is a rare fatal neurological disorder of unknown etiology and difficult early diagnosis. Synaptic loss is a prominent and early feature of many neurodegenerative diseases and positron emission tomography (PET) imaging of synaptic vesicle glycoprotein 2A (SV2A) has emerged as a promising biomarker of synaptic density.

Methods: All patients and HC underwent ¹⁸F-SynVesT-1 PET and 3D-T1-weighted structural MR images. T1 images were used for PET spatial normalization. Standardized uptake value ratio (SUVR) relative to centrum semiovale was computed and regions of interest (ROIs) were defined in MNI standard space using AAL3 template. PD. SPM12 was used for group comparisons. Support vector machine-recursive feature elimination (SVM-RFE) was used to select the optimal ALL3 ROIs and 4 ML (Linear-SVM, logistic regression, K-NearestNeighbor, Gaussian Naive Bayes) based algorithms were applied to construct MSA subtype diagnosis and differential diagnosis signatures. Accuracy, sensitivity, specificity, and receiver operating curves were used to assess the performance of these signatures.

Results: We observed widespread and significant reductions of ¹⁸F-SynVesT-1 uptake in the Cerebellum in both MSA-C and MSA-P compared to HC participants, especially in MSA-C subjects (Figure 1 A and B). A significant reduction of ¹⁸F-SynVesT-1 uptake in the posterior putamen was also observed in MSA-P compared with HC participants(Figure 1 B). Compared with MSA-C, MSA-P has more cerebellar synaptic density reserved and more bilateral posterior putamen synaptic density lost (Figure 1 C). SVM-based MSA-C and MSA-P diagnosis and differential diagnosis signatures have the best performance. The accuracy, sensitivity, specificity, and area under curve of MSA-C and MSA-P diagnosis signatures were 1.0-1.0-1.0-1.0, 0.97-0.95-0.99-0.99, respectively (Figure 1 D and E). The accuracy, sensitivity, specificity, and area under curve of the differential diagnosis between MSA-C and MSA-P were 0.93,0.93,0.95, and 0.98 (Figure 1 F).

Conclusions: Our findings first demonstrate that MSA-C and MSA-P have significantly different ¹⁸F-SynVesT-1 uptake patterns. SVM-based MSA subtype diagnosis and differential diagnosis signatures hold promise as an in vivo biomarker for MSA and as an outcome measure for trials of disease-modifying therapies, particularly those targeted at the preservation and restoration of synapses.

• Download figure
• Open in new tab
• Download powerpoint