The Parkinson’s Disease Progression Neuroimaging Initiative

The Parkinson’s Disease Progressive Neuroimaging Initiative (PDPNI) is a longitudinal observational clinical study. In PDPNI, the clinical and imaging data of patients diagnosed with Parkinsonian syndromes and Idiopathic rapid eye movement sleep behavior disorder (RBD) were longitudinally followed every two years, aiming to identify progression biomarkers of Parkinsonian syndromes through functional imaging modalities including FDG-PET, DAT-PET imaging, ASL MRI, and fMRI, as well as the treatment conditions, clinical symptoms, and clinical assessment results of patients. From February 2012 to March 2019, 224 subjects (including 48 healthy subjects and 176 patients with confirmed PDS) have been enrolled in PDPNI. The detailed clinical information and clinical assessment scores of all subjects were collected by neurologists from Huashan Hospital, Fudan University. All subjects enrolled in PDPNI were scanned with 18F-FDG PET, 11C-CFT PET, and MRI scan sequence. All data were collected in strict accordance with standardized data collection protocols.

Data-driven identification of diagnostically useful extrastriatal signal in dopamine transporter SPECT using explainable AI

This study used explainable artificial intelligence for data-driven identification of extrastriatal brain regions that can contribute to the interpretation of dopamine transporter SPECT with 123I-FP-CIT in parkinsonian syndromes. A total of 1306 123I-FP-CIT-SPECT were included retrospectively. Binary classification as ‘reduced’ or ‘normal’ striatal 123I-FP-CIT uptake by an experienced reader served as standard-of-truth. A custom-made 3-dimensional convolutional neural network (CNN) was trained for classification of the SPECT images with 1006 randomly selected images in three different settings: “full image”, “striatum only” (3-dimensional region covering the striata cropped from the full image), “without striatum” (full image with striatal region removed). The remaining 300 SPECT images were used to test the CNN classification performance. Layer-wise relevance propagation (LRP) was used for voxelwise quantification of the relevance for the CNN-based classification in this test set. Overall accuracy of CNN-based classification was 97.0%, 95.7%, and 69.3% in the “full image”, “striatum only”, and “without striatum” setting. Prominent contributions in the LRP-based relevance maps beyond the striatal signal were detected in insula, amygdala, ventromedial prefrontal cortex, thalamus, anterior temporal cortex, superior frontal lobe, and pons, suggesting that 123I-FP-CIT uptake in these brain regions provides clinically useful information for the differentiation of neurodegenerative and non-neurodegenerative parkinsonian syndromes.

Atypical Parkinsonian Syndromes

The cardinal characteristics of parkinsonism are represented in the mnemonic TRAP: tremor at rest, rigidity, akinesia and bradykinesia, and postural instability. The parkinsonian phenotype encompasses a broad range of clinical and pathologic disorders; the most common (about 55% of cases) is idiopathic (sporadic) Parkinson disease. Rapid disease progression, poor initial response to dopaminergic therapy, or the early presence of certain other signs may suggest an atypical parkinsonian syndrome, sometimes called parkinsonism-plus syndrome.

Cerebrospinal fluid levels of 5-hydroxyindoleacetic acid in Parkinson’s disease and atypical parkinsonian syndromes

Background: Various cerebrospinal fluid (CSF) biomarkers are studied in Parkinson’s disease (PD) and atypical parkinsonian syndromes (APS). Several studies found reduced 5-hydroxyindoleacetic acid (5-HIAA), the main serotonin metabolite, in PD. There is little evidence regarding its levels in APS. Methods: We measured 5-HIAA in the CSF of 90 PD patients, 16 MSA patients, 26 progressive supranuclear palsy (PSP) patients, 11 corticobasal degeneration (CBD) patients, and 31 controls. We also compared the values in depressed and non-depressed patients. Results: There was a statistically significant difference in CSF 5-HIAA in PD and MSA compared to the control group (median in PD 15.8 µg/l, in MSA 13.6 µg/l vs. 24.3 µg/l in controls; P=0.0008 in PD, P=0.006 in MSA). There was no statistically significant difference in CSF 5-HIAA in PSP and CBD compared to the control group (median in PSP 22.7 µg/l, in CBD 18.7 µg/l vs. 24.3 µg/l in controls; P= 1 in both PSP and CBD). CSF 5-HIAA levels were lower in PD patients with depression compared to PD patients without depression (median 8.34 vs. 18.48, P<0.0001). Conclusions: CSF 5-HIAA is decreased in PD and MSA. The CSF 5-HIAA levels in PSP and CBS did not differ from those of the control group. There was a tendency toward lower CSF 5-HIAA in MSA than in PD, however, the results did not reach statistical significance. These results may be explained by more severe damage of the serotonergic system in synucleinopathies (PD, MSA) than in tauopathies (PSP, CBS).

Explainable AI to improve acceptance of convolutional neural networks for automatic classification of dopamine transporter SPECT in the diagnosis of clinically uncertain parkinsonian syndromes

Purpose Deep convolutional neural networks (CNN) provide high accuracy for automatic classification of dopamine transporter (DAT) SPECT images. However, CNN are inherently black-box in nature lacking any kind of explanation for their decisions. This limits their acceptance for clinical use. This study tested layer-wise relevance propagation (LRP) to explain CNN-based classification of DAT-SPECT in patients with clinically uncertain parkinsonian syndromes. Methods The study retrospectively included 1296 clinical DAT-SPECT with visual binary interpretation as “normal” or “reduced” by two experienced readers as standard-of-truth. A custom-made CNN was trained with 1008 randomly selected DAT-SPECT. The remaining 288 DAT-SPECT were used to assess classification performance of the CNN and to test LRP for explanation of the CNN-based classification. Results Overall accuracy, sensitivity, and specificity of the CNN were 95.8%, 92.8%, and 98.7%, respectively. LRP provided relevance maps that were easy to interpret in each individual DAT-SPECT. In particular, the putamen in the hemisphere most affected by nigrostriatal degeneration was the most relevant brain region for CNN-based classification in all reduced DAT-SPECT. Some misclassified DAT-SPECT showed an “inconsistent” relevance map more typical for the true class label. Conclusion LRP is useful to provide explanation of CNN-based decisions in individual DAT-SPECT and, therefore, can be recommended to support CNN-based classification of DAT-SPECT in clinical routine. Total computation time of 3 s is compatible with busy clinical workflow. The utility of “inconsistent” relevance maps to identify misclassified cases requires further investigation.