Application of deep canonically correlated sparse autoencoder for the classification of schizophrenia

Li, Gang; Han, De-Peng; Wang, Chao; Hu, Wei; Calhoun, Vince D.; Wang, Yu Ping

doi:10.1016/j.cmpb.2019.105073

Cited by 56 publications

(29 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Preferably even, the model should be tested on a separate, external dataset as this provides most evidence of model generalization, but this is often not feasible. Still, several studies only report accuracies based on a single train/test split ( Li G. et al, 2018 , Srinivasagopalan et al, 2019 , Kuang et al, 2014 , Kuang and He, 2014 , Li et al, 2019 ), therefore reporting an overly optimistic outcome and complicating comparisons with other studies. As the best practice for model generalizability is to use an independently collected dataset as test set, it is good practice to report leave site out validation as each site is an independent dataset.…”

Section: Discussionmentioning

confidence: 99%

Deep learning applications for the classification of psychiatric disorders using neuroimaging data: Systematic review and meta-analysis

Quaak

Mortel

Thomas

et al. 2021

NeuroImage: Clinical

View full text Add to dashboard Cite

Highlights Deep learning methods are a promising tool for classification of individual psychiatric patients. Classification performance is better for schizophrenia than autism spectrum disorders than ADHD. Research is mainly focused on large publicly available datasets. There is large heterogeneity in choice of preprocessing, features, models and evaluation. Studies often lack sufficient performance measures for quantitative comparisons to conventional machine learning methods.

show abstract

Section: Discussionmentioning

confidence: 99%

Deep learning applications for the classification of psychiatric disorders using neuroimaging data: Systematic review and meta-analysis

Quaak

Mortel

Thomas

et al. 2021

NeuroImage: Clinical

View full text Add to dashboard Cite

show abstract

“…Other ML methods [e.g., deep autoencoder (DeepAE), a likelihood-free inference framework] consider a framework in which the information of the input variables is compressed and subsequently reconstruct the input data, minimizing the loss function. In this sense, the deep learning (DL) approach is a class of neural networks and has been an active area of ML research, emerging as a powerful tool in genetics and genomics studies, e.g., schizophrenia classification through datasets of SNP and functional magnetic resonance imaging (Li et al, 2020), gene expression prediction from SNP genotypes (Xie et al, 2017), MADS-box gene classification system for angiosperms (Chen et al, 2019), and RNA secondary structure prediction (Zhang et al, 2019) and to predict quantitative phenotypes from SNPs (Liu et al, 2019). Unlike the traditional artificial neural network, DL algorithms consider many hidden layers during the network training (Xie et al, 2016).…”

Section: Introductionmentioning

confidence: 99%

A Deep Learning Approach to Population Structure Inference in Inbred Lines of Maize

et al. 2020

View full text Add to dashboard Cite

Analysis of population genetic variation and structure is a common practice for genome-wide studies, including association mapping, ecology, and evolution studies in several crop species. In this study, machine learning (ML) clustering methods, K-means (KM), and hierarchical clustering (HC), in combination with non-linear and linear dimensionality reduction techniques, deep autoencoder (DeepAE) and principal component analysis (PCA), were used to infer population structure and individual assignment of maize inbred lines, i.e., dent field corn (n = 97) and popcorn (n = 86). The results revealed that the HC method in combination with DeepAE-based data preprocessing (DeepAE-HC) was the most effective method to assign individuals to clusters (with 96% of correct individual assignments), whereas DeepAE-KM, PCA-HC, and PCA-KM were assigned correctly 92, 89, and 81% of the lines, respectively. These findings were consistent with both Silhouette Coefficient (SC) and Davies–Bouldin validation indexes. Notably, DeepAE-HC also had better accuracy than the Bayesian clustering method implemented in InStruct. The results of this study showed that deep learning (DL)-based dimensional reduction combined with ML clustering methods is a useful tool to determine genetically differentiated groups and to assign individuals into subpopulations in genome-wide studies without having to consider previous genetic assumptions.

show abstract

“…Preferably even, the model should be tested on a separate, external dataset as this provides most evidence of model generalization, but this is often not feasible. Still, several studies only report accuracies based on a single train/test split 37,44,52,53,66 , therefore reporting an overly optimistic outcome and complicating comparisons with other studies. As the best practice for model generalizability is to use an independently collected dataset as test set, it is good practice to report leave site out validation as each site is an independent dataset.…”

Section: Discussionmentioning

confidence: 99%

Deep learning applications for the classification of psychiatric disorders using neuroimaging data: systematic review and meta-analysis

Quaak

Mortel

Thomas

et al. 2020

Preprint

View full text Add to dashboard Cite

Deep learning (DL) methods have been increasingly applied to neuroimaging data to identify patients with psychiatric and neurological disorders. This review provides an overview of the different DL applications within psychiatry and compares DL model accuracy to conventional machine learning (ML). Fifty-three articles were included for qualitative analysis, primarily investigating autism spectrum disorder (ASD; n=22), schizophrenia (SZ; n=22) and attention-deficit/hyperactivity disorder (ADHD; n=9). Thirty-two of the thirty-five studies that directly compared DL to ML reported a higher accuracy for DL. Only sixteen studies could be included in a meta-regression to quantitatively compare DL and ML performance. This showed a higher odds ratio for DL models, though the comparison attained significance only for ASD. These results suggest that deep learning of neuroimaging data is a promising tool for the classification of individual psychiatric patients. The current evaluation is limited by minimal reporting of performance measures to enable quantitative comparisons, and the restriction to ADHD, SZ and ASD as current research focusses on large publicly available datasets.

show abstract

Application of deep canonically correlated sparse autoencoder for the classification of schizophrenia

Cited by 56 publications

References 31 publications

Deep learning applications for the classification of psychiatric disorders using neuroimaging data: Systematic review and meta-analysis

Deep learning applications for the classification of psychiatric disorders using neuroimaging data: Systematic review and meta-analysis

A Deep Learning Approach to Population Structure Inference in Inbred Lines of Maize

Deep learning applications for the classification of psychiatric disorders using neuroimaging data: systematic review and meta-analysis

Contact Info

Product

Resources

About