Estimating uncertainty in read-out patterns: Application to controls-based denoising and voxel-based morphometry patterns in neurodegenerative and neuropsychiatric diseases.

Quantifying pathology-related patterns in patient data with pattern expression score (PES) is a standard approach in medical image analysis. In order to estimate the PES error, we here propose to express the uncertainty contained in read-out patterns in terms of the expected squared Euclidean distance between the read-out pattern and the unknown "true" pattern (squared standard error of the read-out pattern, SE2 ). Using SE2 , we predicted and optimized the net benefit (NBe) of the recently suggested method controls-based denoising (CODE) by weighting patterns of nonpathological variance (NPV). Multi-center MRI (1192 patients with various neurodegenerative and neuropsychiatric diseases, 1832 healthy controls) were analysed with voxel-based morphometry. For each pathology, accounting for SE2 , NBe correctly predicted classification improvement and allowed to optimize NPV pattern weights. Using these weights, CODE improved classification performances in all but one analyses, for example, for prediction of conversion to Alzheimer's disease (AUC 0.81 vs. 0.75, p = .01), diagnosis of autism (AUC 0.66 vs. 0.60, p < .001), and of major depressive disorder (AUC 0.62 vs. 0.50, p = .03). We conclude that the degree of uncertainty in a read-out pattern should generally be reported in PES-based analyses and suggest using weighted CODE as a complement to PES-based analyses.

Controls-based denoising, a new approach for medical image analysis, improves prediction of conversion to Alzheimer's disease with FDG-PET.

OBJECTIVE: The pattern expression score (PES), i.e., the degree to which a pathology-related pattern is present, is frequently used in FDG-brain-PET analysis and has been shown to be a powerful predictor of conversion to Alzheimer's disease (AD) in mild cognitive impairment (MCI). Since, inevitably, the PES is affected by non-pathological variability, our aim was to improve classification with the simple, yet novel approach to identify patterns of non-pathological variance in a separate control sample using principal component analysis and removing them from patient data (controls-based denoising, CODE) before calculating the PES. METHODS: Multi-center FDG-PET from 220 MCI patients (64 non-converter, follow-up ≥ 4 years; 156 AD converter, time-to-conversion ≤ 4 years) were obtained from the ADNI database. Patterns of non-pathological variance were determined from 262 healthy controls. An AD pattern was calculated from AD patients and controls. We predicted AD conversion based on PES only and on PES combined with neuropsychological features and ApoE4 genotype. We compared classification performance achieved with and without CODE and with a standard machine learning approach (support vector machine). RESULTS: Our model predicts that CODE improves the signal-to-noise ratio of AD-PES by a factor of 1.5. PES-based prediction of AD conversion improved from AUC 0.80 to 0.88 (p= 0.001, DeLong's method), sensitivity 69 to 83%, specificity 81% to 88% and Matthews correlation coefficient (MCC) 0.45 to 0.66. Best classification (0.93 AUC) was obtained when combining the denoised PES with clinical features. CONCLUSIONS: CODE, applied in its basic form, significantly improved prediction of conversion based on PES. The achieved classification performance was higher than with a standard machine learning algorithm, which was trained on patients, explainable by the fact that CODE used additional information (large sample of healthy controls). We conclude that the proposed, novel method is a powerful tool for improving medical image analysis that offers a wide spectrum of biomedical applications, even beyond image analysis.

Hypermetabolism in the cerebellum and brainstem and cortical hypometabolism are independently associated with cognitive impairment in Parkinson's disease.

PURPOSE: Cognitive impairment (CI) in Parkinson's disease (PD) is associated with a widespread reduction in cortical glucose metabolism and relative increases in the cerebellum and brainstem as measured using 18F-fluorodesoxyglucose (FDG) PET. We separately analysed CI-related hypermetabolism and hypometabolism in comparison with neuropsychological test performance and investigated whether increased FDG uptake is a true feature of the disease or a normalization effect. METHODS: The study included 29 subjects (12 patients with PD, 10 patients with PD dementia and 7 healthy controls") who underwent FDG PET and comprehensive neuropsychological testing. Test performance across various cognitive domains was summarized in a cognitive staging score. Metabolic indices reflecting associated changes in regional cerebral glucose metabolism (rCGM) were calculated: index(-) for CI-related hypometabolism, and index(+) for CI-related hypermetabolism. We tested whether index(+) offered additional value in predicting the severity of CI in multiple regression analysis. RESULTS: At higher stages of CI, increased rCGM was found in the posterior cerebellar vermis and pons, associated with impaired attention, executive function and memory. Reduced rCGM was found in various cortical regions in agreement with the literature. In multiple regression analysis, both indices independently predicted the severity of CI with a whole-model R2 of 0.68 (index(-), p = 0.0006; index(+), p = 0.013), confirmed by alternative analyses combining different reference tissues in the multiple regression. CONCLUSION: We found CI-related hypermetabolism in cerebellar regions that are known to be involved in several cognitive functions and in the pons. These alterations may represent compensatory activation of cognitive networks including cerebropontocerebellar tracts.

Home-Based Physical Behavior in Late Stage Parkinson Disease Dementia: Differences between Cognitive Subtypes.

BACKGROUND: For the early diagnosis of Parkinson disease dementia (PDD), objective home-based tools are needed to quantify even mild stages of dysfunction of the activities of daily living (ADL). OBJECTIVES: In this pilot study, home-based physical behavior was assessed to examine whether it is possible to distinguish mild cognitive impairment (PD-MCI) from PDD. METHODS: Fifty-five patients with mild to severe Parkinson disease (PD) participated in this cross-sectional study. Based on comprehensive neuropsychological testing, PD patients were classified as cognitively nonimpaired (PD-NC), PD-MCI or PDD. For physical behavior assessments, patients wore the accelerometer DynaPort® (McRoberts) for 3 days. Ordinal logistic regression models with continuous Y were applied to correct results for motor impairment and depressive symptoms. RESULTS: After excluding 7 patients due to insufficient wearing time, 48 patients with a mean of 2 recorded days were analyzed (17 PD-NC, 22 PD-MCI, 9 PDD). ADL-impaired PDD patients showed fewer sedentary bouts than non-ADL-impaired PD-MCI (p = 0.01, odds ratio [OR] = 8.9, 95% confidence interval [CI] = 1.8-45.2) and PD-NC (p = 0.01, OR = 10.3, CI = 1.6-67.3) patients, as well as a longer sedentary bout length (PD-NC: p = 0.02, OR = 0.1, CI = 0.02-0.65; PD-MCI: p = 0.02, OR = 0.14, CI = 0.03-0.69). These differences were mainly caused by fewer (PD-NC: p = 0.02, OR = 9.6, CI = 1.5-62.4; PD-MCI: p = 0.01, OR = 8.5, CI = 1.5-37.3) but longer sitting bouts (PD-NC: p = 0.03, OR = 0.12, CI = 0.02-0.80; PD-MCI: p = 0.04, OR = 0.19, CI = 0.04-0.93). Tests assessing executive function, visuoconstruction and attention correlated significantly with specific activity parameters (ρ ≥ 0.3; p < 0.05). CONCLUSION: Objective assessment of physical behavior, in particular the detection of sedentary bouts, is a promising contributor to the discrimination between PD-MCI and PDD.

Uncertainty estimation and explainability in deep learning-based age estimation of the human brain: Results from the German National Cohort MRI study.

Brain ageing is a complex neurobiological process associated with morphological changes that can be assessed on MRI scans. Recently, Deep learning (DL)-based approaches have been proposed for the prediction of chronological brain age from MR images yielding high accuracy. These approaches, however, usually do not address quantification of uncertainty and, therefore, intrinsic physiological variability. Considering uncertainty is essential for the interpretation of the difference between predicted and chronological age. In addition, DL-based models lack in explainability compared to classical approaches like voxel-based morphometry. In this study, we aim to address both, modeling uncertainty and providing visual explanations to explore physiological patterns in brain ageing. T1-weighted brain MRI datasets of 10691 participants of the German National Cohort Study, drawn from the general population, were included in this study (chronological age from 20 to 72 years). A regression model based on a 3D Convolutional Neural Network taking into account aleatoric noise was implemented for global as well as regional brain age estimation. We observed high overall accuracy of global brain age estimation with a mean absolute error of 3.2 ±â€¯2.5 years and mean uncertainty of 2.9 ±â€¯0.6 years. Regional brain age estimation revealed higher estimation accuracy and lower uncertainty in central compared to peripheral brain regions. Visual explanations illustrating the importance of brain sub-regions were generated using Grad-CAM: the derived saliency maps showed a high relevance of the lateral and third ventricles, the insular lobe as well as parts of the basal ganglia and the internal capsule.

Central Insulin Modulates Dopamine Signaling in the Human Striatum.

OBJECTIVE: Activity in the dopaminergic pathways of the brain is highly sensitive to body weight and metabolic states. Animal studies show that dopamine neurons are important targets for the metabolic hormone insulin with abolished effects in the insulin-resistant state, leading to increases in body weight and food intake. In humans, the influence of central acting insulin on dopamine and effects of their interplay are still elusive. RESEARCH DESIGN AND METHODS: We investigated whether central administered insulin influences dopaminergic activity in striatal regions and whole-brain neural activity. Using a positron emission tomography (PET)/magnetic resonance imaging (MRI) hybrid scanner, we simultaneously performed [11C]-raclopride-PET and resting-state functional MRI in 10 healthy normal-weight men after application of intranasal insulin or placebo on 2 separate days in a randomized, placebo-controlled, blinded, crossover trial. RESULTS: In response to central insulin compared with placebo administration, we observed greater [11C]-raclopride binding potential in the bilateral ventral and dorsal striatum. This suggests an insulin-induced reduction in synaptic dopamine levels. Resting-state striatal activity was lower 15 and 30 minutes after nasal insulin compared with placebo. Functional connectivity of the mesocorticolimbic circuitry associated with differences in dopamine levels: individuals with a stronger insulin-induced effect on dopamine levels showed a stronger increase in functional connectivity 45 minutes after intranasal insulin. CONCLUSIONS: This study indicates that central insulin modulates dopaminergic tone in the striatum, which may affect regional brain activity and connectivity. Our results deepen the understanding of the insulin-dopamine interaction and the complex network that underlies the regulation of whole-body metabolism.