Dataset Information

Stability of AI-Enabled Diagnosis of Parkinson's Disease: A Study Targeting Substantia Nigra in Quantitative Susceptibility Mapping Imaging.

ABSTRACT: Purpose: Parkinson's disease (PD) diagnosis algorithms based on quantitative susceptibility mapping (QSM) and image algorithms rely on substantia nigra (SN) labeling. However, the difference between SN labels from different experts (or segmentation algorithms) will have a negative impact on downstream diagnostic tasks, such as the decrease of the accuracy of the algorithm or different diagnostic results for the same sample. In this article, we quantify the accuracy of the algorithm on different label sets and then improve the convolutional neural network (CNN) model to obtain a high-precision and highly robust diagnosis algorithm. Methods: The logistic regression model and CNN model were first compared for classification between PD patients and healthy controls (HC), given different sets of SN labeling. Then, based on the CNN model with better performance, we further proposed a novel "gated pooling" operation and integrated it with deep learning to attain a joint framework for image segmentation and classification. Results: The experimental results show that, with different sets of SN labeling that mimic different experts, the CNN model can maintain a stable classification accuracy at around 86.4%, while the conventional logistic regression model yields a large fluctuation ranging from 78.9 to 67.9%. Furthermore, the "gated pooling" operation, after being integrated for joint image segmentation and classification, can improve the diagnosis accuracy to 86.9% consistently, which is statistically better than the baseline. Conclusion: The CNN model, compared with the conventional logistic regression model using radiomics features, has better stability in PD diagnosis. Furthermore, the joint end-to-end CNN model is shown to be suitable for PD diagnosis from the perspectives of accuracy, stability, and convenience in actual use.

SUBMITTER: Xiao B

PROVIDER: S-EPMC8650720 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Introduction: The loss of nigrosome-1, which is also referred to as the swallow tail sign (STS) in T2*-weighted iron-sensitive magnetic resonance imaging (MRI), has recently emerged as a new biomarker for idiopathic Parkinson's disease (IPD). However, consistent recognition of the STS is difficult due to individual variations and different imaging parameters. Radiomics might have the potential to overcome these shortcomings. Therefore, we chose to explore whether radiomic features of nigrosome-1 of substantia nigra (SN) based on quantitative susceptibility mapping (QSM) could help to differentiate IPD patients from healthy controls (HCs). Methods: Three-dimensional multi-echo gradient-recalled echo images (0.86 × 0.86 × 1.00 mm3) were obtained at 3.0-T MRI for QSM in 87 IPD patients and 77 HCs. Regions of interest (ROIs) of the SN below the red nucleus were manually drawn on both sides, and subsequently, volumes of interest (VOIs) were segmented (these ROIs included four 1-mm slices). Then, 105 radiomic features (including 18 first-order features, 13 shape features, and 74 texture features) of bilateral VOIs in the two groups were extracted. Forty features were selected according to the ensemble feature selection method, which combined analysis of variance, random forest, and recursive feature elimination. The selected features were further utilized to distinguish IPD patients from HC using the SVM classifier with 10 rounds of 3-fold cross-validation. Finally, the representative features were analyzed using an unpaired t-test with Bonferroni correction and correlated with the UPDRS-III scores. Results: The classification results from SVM were as follows: area under curve (AUC): 0.96 ± 0.02; accuracy: 0.88 ± 0.03; sensitivity: 0.89 ± 0.06; and specificity: 0.87 ± 0.07. Five representative features were selected to show their detailed difference between IPD patients and HCs: 10th percentile and median in IPD patients were higher than those in HCs (all p < 0.00125), while Gray Level Run Length Matrix (GLRLM)-Long Run Low Gray Level Emphasis, Gray Level Size Zone Matrix (GLSZM)-Gray Level Non-Uniformity, and volume (all p < 0.00125) in IPD patients were lower than those in HCs. The 10th percentile was positively correlated with UPDRS-III score (r = 0.35, p = 0.001). Conclusion: Radiomic features of the nigrosome-1 region of SN based on QSM could be useful in the diagnosis of IPD and could serve as a surrogate marker for the STS.

Project description:BACKGROUND AND PURPOSE:To test whether nigrosome-1 imaging using high-resolution quantitative susceptibility mapping (QSM) combined with histogram analysis can improve the diagnostic accuracy in early-stage idiopathic Parkinson's disease (IPD) patients. METHODS:Three-dimensional multiecho gradient-recalled echo images (0.5×0.5×1.0 mm³) were obtained at 3 T for QSM in 38 patients with IPD and 25 healthy subjects. To segment the substantia nigra (SN), regions of interest (ROIs) were semiautomatically drawn at the location below the red nucleus, and the normal-appearing nigrosome-1 was determined by manual correction. QSM histograms were obtained within the ROI. The segmented SN regions on the right and left that had higher mean susceptibility values and fewer voxels with susceptibility values lower than 60, 65, 70, 75, and 80 ppb were chosen for comparisons between the IPD patients and healthy subjects. These results were compared with those of the visual assessments of nigrosome-1 in susceptibility map-weighted imaging (SMWI) by analyzing receiver operating characteristics curves. RESULTS:The proportion of voxels with susceptibility values lower than 70 ppb showed the best diagnostic performance, with its value differing significantly between the IPD patients (median=0, interquartile range=0-0.23) and healthy subjects (median=10.67, interquartile range=5.98-21.57) (p<0.0001). The number of voxels with susceptibility values lower than 60, 65, 70, 75, and 80 ppb showed worse diagnostic performances but were still significantly better than that of the mean susceptibility value (p=0.0249, 0.0192, 0.0183, 0.0191, and 0.0186, respectively), which also differed significantly between the two groups: 125.81±16.27 ppb (mean±standard deviation) in IPD versus 98.41±11.70 ppb in healthy subjects (p<0.0001). Additionally, using the proportion of voxels with susceptibility values lower than 70 ppb provided significantly better diagnostic performance than did visual assessments of SMWI (p=0.0143). CONCLUSIONS:High-spatial-resolution QSM combined with histogram analysis at 3 T can improve the diagnostic accuracy of early-stage IPD.

Project description:BACKGROUND:Susceptibility MRI may capture Parkinson's disease-related pathology. This study delineated longitudinal changes in different substantia nigra regions. METHODS:Seventy-two PD patients and 62 controls were studied at both baseline and after 18 months with MRI. R2* and quantitative susceptibility mapping values from the substantia nigra pars compacta and substantia nigra pars reticulata were calculated. Mixed-effects models compared controls with PD or PD subgroups having different disease durations: early (<1 year), middle (<5 years, middle-stage PD), and late (>5 years, late-stage PD). Pearson's correlation assessed associations between imaging and clinical measures. RESULTS:At baseline, R2* and quantitative susceptibility mapping were higher in both the substantia nigra pars compacta and substantia nigra pars reticulata in all PD patients (group effect, P???0.003). Longitudinally, the substantia nigra pars compacta R2* showed a faster increase in PD compared with controls (time?×?group, P?=?0.002), whereas quantitative susceptibility mapping did not (P?=?0.668). The substantia nigra pars reticulata R2* and quantitative susceptibility mapping did not differ between PD and controls (time?×?group, P???0.084), although both decreased longitudinally (time effect, P???0.004). Baseline substantia nigra pars compacta R2* was higher in all PD subgroups (group, P???0.006), but showed a significantly faster increase only in later-stage PD (time?×?group, P?<?0.0001) that correlated with changes in nonmotor symptoms (r?=?0.746, P?=?0.002). Baseline substantia nigra pars reticulata quantitative susceptibility mapping was higher in middle-stage PD and later-stage PD (group, P???0.002), but showed a longitudinal decrease (time?×?group, P?=?0.004) only in later-stage PD that correlated with changes in motor signs (r?=?0.837, P?<?0.001). CONCLUSION:Susceptibility MRI revealed distinct patterns of PD progression in the substantia nigra pars compacta and substantia nigra pars reticulata. The different patterns are particularly clear in later-stage patients. These findings may resolve past controversies and have implications in the pathophysiological processes during PD progression. © 2018 International Parkinson and Movement Disorder Society.

Project description:BackgroundParkinson's disease (PD) is the most common neurodegenerative disorder involving the motor system. Although not being the only region involved in PD, affection of the substantia nigra and its projections is responsible for some of the most debilitating features of the disease. To further advance a comprehensive understanding of nigral pathology, we conducted a tissue based comparative proteome study of healthy and diseased human substantia nigra.ResultsThe gross number of differentially regulated proteins in PD was 221. In total, we identified 37 proteins, of which 16 were differentially expressed. Identified differential proteins comprised elements of iron metabolism (H-ferritin) and glutathione-related redox metabolism (GST M3, GST P1, GST O1), including novel redox proteins (SH3BGRL). Additionally, many glial or related proteins were found to be differentially regulated in PD (GFAP, GMFB, galectin-1, sorcin), as well as proteins belonging to metabolic pathways sparsely described in PD, such as adenosyl homocysteinase (methylation), aldehyde dehydrogenase 1 and cellular retinol-binding protein 1 (aldehyde metabolism). Further differentially regulated proteins included annexin V, beta-tubulin cofactor A, coactosin-like protein and V-type ATPase subunit 1. Proteins that were similarly expressed in healthy or diseased substantia nigra comprised housekeeping proteins such as COX5A, Rho GDI alpha, actin gamma 1, creatin-kinase B, lactate dehydrogenase B, disulfide isomerase ER-60, Rab GDI beta, methyl glyoxalase 1 (AGE metabolism) and glutamine synthetase. Interestingly, also DJ-1 and UCH-L1 were expressed similarly. Furthermore, proteins believed to serve as internal standards were found to be expressed in a constant manner, such as 14-3-3 epsilon and hCRMP-2, thus lending further validity to our results.ConclusionUsing an approach encompassing high sensitivity and high resolution, we show that alterations of SN in PD include many more proteins than previously thought. The results point towards a heterogeneous aetiopathogenesis of the disease, including alterations of GSH-related proteins as well as alterations of proteins involved in retinoid metabolism, and they indicate that proteins involved in familial PD may not be differentially regulated in idiopathic Parkinson's disease.

Project description:Parkinson's disease (PD) is a progressive neurodegenerative disease whose etiology is attributed to development of Lewy bodies and degeneration of dopaminergic neurons in the substantia nigra (SN). Currently, there are no definitive diagnostic indicators for PD. In this study, we aimed to identify potential diagnostic biomarkers for PD and analyzed the impact of immune cell infiltrations on disease pathogenesis. The PD expression profile data for human SN tissue, GSE7621, GSE20141, GSE20159, GSE20163 and GSE20164 were downloaded from the Gene Expression Omnibus (GEO) database for use in the training model. After normalization and merging, we identified differentially expressed genes (DEGs) using the Robust rank aggregation (RRA) analysis. Simultaneously, DEGs after batch correction were identified. Gene interactions were determined through venn Diagram analysis. Functional analyses and protein-protein interaction (PPI) networks were used to the identify hub genes, which were visualized through Cytoscape. A Lasso Cox regression model was employed to identify the potential diagnostic genes. The GSE20292 dataset was used for validation. The proportion of infiltrating immune cells in the samples were determined via the CIBERSORT method. Sixty-two DEGs were screened in this study. They were found to be enriched in nerve conduction, dopamine (DA) metabolism, and DA biosynthesis Gene Ontology (GO) terms. The PPI network and Lasso Cox regression analysis revealed seven potential diagnostic genes, namely SLC18A2, TAC1, PCDH8, KIAA0319, PDE6H, AXIN1, and AGTR1, were subsequently validated in peripheral blood samples obtained from healthy control (HC) and PD patients, as well as in the GSE20292 dataset. The results revealed the exceptional sensitivity and specificity of these genes in PD diagnosis and monitoring. Moreover, PD patients exhibited a higher number of plasma cells, compared to HC individuals. The SLC18A2, TAC1, PCDH8, KIAA0319, PDE6H, AXIN1, and AGTR1 are potential diagnostic biomarkers for PD. Our findings also reveal the essential roles of immune cell infiltration in both disease onset and trajectory.

Dataset Information

Stability of AI-Enabled Diagnosis of Parkinson's Disease: A Study Targeting Substantia Nigra in Quantitative Susceptibility Mapping Imaging.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets