We performed a systematic review of studies focusing on the automatic prediction of the progression of mild cognitive impairment to Alzheimer's disease (AD) dementia, and a quantitative analysis of the methodological choices impacting performance. This review included 172 articles, from which 234 experiments were extracted. For each of them, we reported the used data set, the feature types, the algorithm type, performance and potential methodological issues. The impact of these characteristics on the performance was evaluated using a multivariate mixed effect linear regressions. We found that using cognitive, fluorodeoxyglucose-positron emission tomography or potentially electroencephalography and magnetoencephalography variables significantly improved predictive performance compared to not including them, whereas including other modalities, in particular T1 magnetic resonance imaging, did not show a significant effect. The good performance of cognitive assessments questions the wide use of imaging for predicting the progression to AD and advocates for exploring further fine domain-specific cognitive assessments. We also identified several methodological issues, including the absence of a test set, or its use for feature selection or parameter tuning in nearly a fourth of the papers. Other issues, found in 15% of the studies, cast doubts on the relevance of the method to clinical practice. We also highlight that shortterm predictions are likely not to be better than predicting that subjects stay stable over
Background Machine learning algorithms using magnetic resonance imaging (MRI) data can accurately discriminate parkinsonian syndromes. Validation in patients recruited in routine clinical practice is missing. Objective The aim of this study was to assess the accuracy of a machine learning algorithm trained on a research cohort and tested on an independent clinical replication cohort for the categorization of parkinsonian syndromes. Methods Three hundred twenty‐two subjects, including 94 healthy control subjects, 119 patients with Parkinson's disease (PD), 51 patients with progressive supranuclear palsy (PSP) with Richardson's syndrome, 35 with multiple system atrophy (MSA) of the parkinsonian variant (MSA‐P), and 23 with MSA of the cerebellar variant (MSA‐C), were recruited. They were divided into a training cohort (n = 179) scanned in a research environment and a replication cohort (n = 143) examined in clinical practice on different MRI systems. Volumes and diffusion tensor imaging (DTI) metrics in 13 brain regions were used as input for a supervised machine learning algorithm. To harmonize data across scanners and reduce scanner‐dependent effects, we tested two types of normalizations using patient data or healthy control data. Results In the replication cohort, high accuracies were achieved using volumetry in the classification of PD–PSP, PD–MSA‐C, PSP–MSA‐C, and PD‐atypical parkinsonism (balanced accuracies: 0.840–0.983, area under the receiver operating characteristic curves: 0.907–0.995). Performances were lower for the classification of PD–MSA‐P, MSA‐C–MSA‐P (balanced accuracies: 0.765–0.784, area under the receiver operating characteristic curve: 0.839–0.871) and PD–PSP–MSA (balanced accuracies: 0.773). Performance using DTI was improved when normalizing by controls, but remained lower than that using volumetry alone or combined with DTI. Conclusions A machine learning approach based on volumetry enabled accurate classification of subjects with early‐stage parkinsonism, examined on different MRI systems, as part of their clinical assessment. © 2020 International Parkinson and Movement Disorder Society
We ranked third in the Predictive Analytics Competition (PAC) 2019 challenge by achieving a mean absolute error (MAE) of 3.33 years in predicting age from T1-weighted MRI brain images. Our approach combined seven algorithms that allow generating predictions when the number of features exceeds the number of observations, in particular, two versions of best linear unbiased predictor (BLUP), support vector machine (SVM), two shallow convolutional neural networks (CNNs), and the famous ResNet and Inception V1. Ensemble learning was derived from estimating weights via linear regression in a hold-out subset of the training sample. We further evaluated and identified factors that could influence prediction accuracy: choice of algorithm, ensemble learning, and features used as input/MRI image processing. Our prediction error was correlated with age, and absolute error was greater for older participants, suggesting to increase the training sample for this subgroup. Our results may be used to guide researchers to build age predictors on healthy individuals, which can be used in research and in the clinics as non-specific predictors of disease status.
Purpose of review: To review recent findings and research directions on impulse control disorders and related behaviors (ICDRBs) in Parkinson's disease (PD). Recent findings:Longitudinal studies found that prevalence increases during PD progression, incident ICDRBs being around 10% per year in patients treated with dopaminergic therapies.Screening tools and severity scales already developed have been validated and are available in several countries and languages. Main clinical risk factors include young age, male gender, type, doses and duration of the dopaminergic therapy, PD motor severity and dyskinesia, depression, anxiety, apathy, sleep disorders and impulsivity traits. Genetic factors are suspected by a high estimated heritability, but individual genes and variants remain to be replicated.Management of ICDRBs is centered on dopamine agonists decrease, with the risk to develop withdrawal symptoms. Cognitive behavioral therapy and subthalamic nucleus deep brain stimulation also improve ICDRBs. In the perspective of precision medicine, new individual prediction models of these disorders have been proposed, but they need further independent replication. Summary: Regular monitoring of ICDRB during the course of PD is needed, particularly in subject at high risk of developing these complications. Precision medicine will require appropriate use of machine learning to be reached in the clinical setting.
Parkinson’s disease is one of the most common age-related neurodegenerative disorders. Although predominantly a motor disorder, cognitive impairment and dementia are important features of Parkinson’s disease, particularly in the later stages of the disease. However, the rate of cognitive decline varies among Parkinson’s disease patients, and the genetic basis for this heterogeneity is incompletely understood. To explore the genetic factors associated with rate of progression to Parkinson’s disease dementia, we performed a genome-wide survival meta-analysis of 3,923 clinically diagnosed Parkinson’s disease cases of European ancestry from four longitudinal cohorts. In total, 6.7% of individuals with Parkinson’s disease developed dementia during study follow-up, on average 4.4 ± 2.4 years from disease diagnosis. We have identified the APOE ε4 allele as a major risk factor for the conversion to Parkinson’s disease dementia [hazards ratio = 2.41 (1.94–3.00), P = 2.32 × 10−15], as well as a new locus within the ApoE and APP receptor LRP1B gene [hazards ratio = 3.23 (2.17–4.81), P = 7.07 × 10−09]. In a candidate gene analysis, GBA variants were also identified to be associated with higher risk of progression to dementia [hazards ratio = 2.02 (1.21–3.32), P = 0.007]. CSF biomarker analysis also implicated the amyloid pathway in Parkinson’s disease dementia, with significantly reduced levels of amyloid β42 (P = 0.0012) in Parkinson’s disease dementia compared to Parkinson’s disease without dementia. These results identify a new candidate gene associated with faster conversion to dementia in Parkinson's disease and suggest that amyloid-targeting therapy may have a role in preventing Parkinson’s disease dementia.
Background Tourette disorder (TD), hallmarks of which are motor and vocal tics, has been related to functional abnormalities in large-scale brain networks. Using a fully data driven approach in a prospective, case–control study, we tested the hypothesis that functional connectivity of these networks carries a neural signature of TD. Our aim was to investigate (i) the brain networks that distinguish adult patients with TD from controls, and (ii) the effects of antipsychotic medication on these networks. Methods Using a multivariate analysis based on support vector machine (SVM), we developed a predictive model of resting state functional connectivity in 48 patients and 51 controls, and identified brain networks that were most affected by disease and pharmacological treatments. We also performed standard univariate analyses to identify differences in specific connections across groups. Results SVM was able to identify TD with 67% accuracy (p = 0.004), based on the connectivity in widespread networks involving the striatum, fronto-parietal cortical areas and the cerebellum. Medicated and unmedicated patients were discriminated with 69% accuracy (p = 0.019), based on the connectivity among striatum, insular and cerebellar networks. Univariate approaches revealed differences in functional connectivity within the striatum in patients v. controls, and between the caudate and insular cortex in medicated v. unmedicated TD. Conclusions SVM was able to identify a neuronal network that distinguishes patients with TD from control, as well as medicated and unmedicated patients with TD, holding a promise to identify imaging-based biomarkers of TD for clinical use and evaluation of the effects of treatment.
Objective: To study the association between impulse control disorders (ICDs) in Parkinsons disease (PD) and genetic risk scores (GRS) for 40 known or putative risk factors (e.g. depression, personality traits). Background: In absence of published genome-wide association studies (GWAS), little is known about the genetics of ICDs in PD. GRS of related phenotypes, for which large GWAS are available, may help shed light on the genetic contributors of ICDs in PD. Methods: We searched for GWAS on European ancestry populations with summary statistics publicly available for a broad range of phenotypes, including other psychiatric disorders, personality traits, and simple phenotypes. We separately tested their predictive ability in two of the largest PD cohorts with clinical and genetic available: the Parkinsons Progression Markers Initiative database (N = 368, 33% female, age range = [33 − 84]) and the Drug Interaction With Genes in Parkinsons Disease study (N = 373, 40% female, age range = [29 − 85]). Results: We considered 40 known or putative risk factors for ICDs in PD for which large GWAS had been published. After Bonferroni correction for multiple comparisons, no GRS or the combination of the 40 GRS were significantly associated with ICDs from the analyses in each cohort separately and from the meta-analysis. Conclusion: Albeit unsuccessful, our approach will gain power in the coming years with increasing availability of genotypes in clinical cohorts of PD, but also from future increase in GWAS sample sizes of the phenotypes we considered. Our approach may be applied to other complex disorders, for which GWAS are not available or limited.
In order to reach precision medicine and improve patients’ quality of life, machine learning is increasingly used in medicine. Brain disorders are often complex and heterogeneous, and several modalities such as demographic, clinical, imaging, genetics and environmental data have been studied to improve their understanding. Deep learning, a subpart of machine learning, provides complex algorithms that can learn from such various data. It has become state of the art in numerous fields, including computer vision and natural language processing, and is also growingly applied in medicine. In this article, we review the use of deep learning for brain disorders. More specifically, we identify the main applications, the concerned disorders and the types of architectures and data used. Finally, we provide guidelines to bridge the gap between research studies and clinical routine.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.