CoChat
Sign Up
Cited by 368Artificial Intelligence in HealthcareMachine Learning in HealthcareArtificial Intelligence in Healthcare and Education

The DOI matches its Crossref record; 137 of the 139 references it cites were found in a registry or an index.

Deep learning in mental health outcome research: a scoping review

Open access
Su, Chang, Xu, Zhenxing, Pathak, Jyotishman, Wang, Fei·2020·Translational Psychiatry
DOI 10.1038/s41398-020-0780-3PMC PMC7293215
Open this paper in CoChat — free accountOpen PDF

Abstract

Mental illnesses, such as depression, are highly prevalent and have been shown to impact an individual’s physical health. Recently, artificial intelligence (AI) methods have been introduced to assist mental health providers, including psychiatrists and psychologists, for decision-making based on patients’ historical data (e.g., medical records, behavioral data, social media usage, etc.). Deep learning (DL), as one of the most recent generation of AI technologies, has demonstrated superior performance in many real-world applications ranging from computer vision to healthcare. The goal of this study is to review existing research on applications of DL algorithms in mental health outcome research. Specifically, we first briefly overview the state-of-the-art DL techniques. Then we review the literature relevant to DL applications in mental health outcomes. According to the application scenarios, we categorize these relevant articles into four groups: diagnosis and prognosis based on clinical data, analysis of genetics and genomics data for understanding mental health conditions, vocal and visual expression data analysis for disease detection, and estimation of risk of mental illness using social media data. Finally, we discuss challenges in using DL algorithms to improve our understanding of mental health conditions and suggest several promising directions for their applications in improving mental health diagnosis and treatment.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41398-020-0780-3

Cites

137 of 137 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Examples of deep neural networks. a Deep feedforward neural network (DFNN). It is the basic design of DL models. Commonly, a DFNN contains multiple hidden layers. b A recurrent neural network (RNN) is presented to process sequence data. To encode history information, each recurrent neuron receives the input element and the state vector of the predecessor neuron, and yields a hidden state fed to the successor neuron. For example, not only the individual information but also the dependence of the elements of the sequence x1 → x2 → x3 → x4 → x5 is encoded by the RNN architecture. c Convolutional neural network (CNN). Between input layer (e.g., input neuroimage) and output layer, a CNN commonly contains three types of layers: the convolutional layer that is to generate feature maps by sliding convolutional kernels in the previous layer; the pooling layer is used to reduce dimensionality of previous convolutional layer; and the fully connected layer is to make prediction. For the illustrative purpose, this example only has one layer of each type; yet, a real-world CNN would have multiple convolutional and pooling layers (usually in an interpolated manner) and one fully connected layer. d Autoencoder consists of two components: the encoder, which learns to compress the input data into a latent representation layer by layer, whereas the decoder, inverse to the encoder, learns to reconstruct the data at the output layer. The learned compressed representations can be fed to the downstream predictive model.
Fig. 1 Examples of deep neural networks. a Deep feedforward neural network (DFNN). It is the basic design of DL models. Commonly, a DFNN contains multiple hidden layers. b A recurrent neural network (RNN) is presented to process sequence data. To encode history information, each recurrent neuron receives the input element and the state vector of the predecessor neuron, and yields a hidden state fed to the successor neuron. For example, not only the individual information but also the dependence of the elements of the sequence x1 → x2 → x3 → x4 → x5 is encoded by the RNN architecture. c Convolutional neural network (CNN). Between input layer (e.g., input neuroimage) and output layer, a CNN commonly contains three types of layers: the convolutional layer that is to generate feature maps by sliding convolutional kernels in the previous layer; the pooling layer is used to reduce dimensionality of previous convolutional layer; and the fully connected layer is to make prediction. For the illustrative purpose, this example only has one layer of each type; yet, a real-world CNN would have multiple convolutional and pooling layers (usually in an interpolated manner) and one fully connected layer. d Autoencoder consists of two components: the encoder, which learns to compress the input data into a latent representation layer by layer, whereas the decoder, inverse to the encoder, learns to reconstruct the data at the output layer. The learned compressed representations can be fed to the downstream predictive model.CC BY · PubMed Central
Fig. 2 Technical details of neural networks. a An illustration of basic unit of neural networks, i.e., artificial neuron. Each input xi is associated with a weight wi. The weighted sum of all inputs Σwixi is fed to a nonlinear activation function f to generate the output yj of the j-th neuron, i.e., yj = f(Σwixi). b Illustrations of the widely used nonlinear activation function.
Fig. 2 Technical details of neural networks. a An illustration of basic unit of neural networks, i.e., artificial neuron. Each input xi is associated with a weight wi. The weighted sum of all inputs Σwixi is fed to a nonlinear activation function f to generate the output yj of the j-th neuron, i.e., yj = f(Σwixi). b Illustrations of the widely used nonlinear activation function.CC BY · PubMed Central
Fig. 3 PRISMA flow diagram: deep learning in mental health outcome research. In total, 57 studies, in terms of clinical data analysis, genetic data analysis, vocal and visual expression data analysis, and social media data analysis, which met our eligibility criteria, were included in this review.
Fig. 3 PRISMA flow diagram: deep learning in mental health outcome research. In total, 57 studies, in terms of clinical data analysis, genetic data analysis, vocal and visual expression data analysis, and social media data analysis, which met our eligibility criteria, were included in this review.CC BY · PubMed Central
Fig. 4 An illustration of the multimodal deep neural network. One can model each modality with a specific network and combine them using the final fully-connected layers. In this way, parameters of the entire neural network can be jointly learned in a typical backpropagation manner.
Fig. 4 An illustration of the multimodal deep neural network. One can model each modality with a specific network and combine them using the final fully-connected layers. In this way, parameters of the entire neural network can be jointly learned in a typical backpropagation manner.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 139 references

Research works printed but not linked: 7

  1. Biship, C. M. Pattern Recognition and Machine Learning (Information Science and Statistics) (Springer-Verlag, Berlin, 2007).
  2. McGuffin, P., Owen, M. J. & Gottesman, I. I. Psychiatric Genetics and Genomics (Oxford Univ. Press, New York, 2004).
  3. Yue, T. & Wang, H. Deep learning for genomics: a concise overview. Preprint at arXiv:1802.00810 (2018). · arXiv
  4. Salakhutdinov, R. & Hinton, G. Deep Boltzmann machines. In Proc. 12th Int. Conference on Artificial Intelligence and Statistics 448–455 (Clearwater, Florida, USA, 2009).
  5. Sadeque, F., Xu, D. & Bethard, S. UArizona at the CLEF eRisk 2017 pilot task: linear and recurrent models for early depression detection. CEUR Workshop Proc. 1866 (2017).
  6. Yosinski, J., Clune, J., Bengio, Y. & Lipson, H. How transferable are features in deep neural networks? In Proc. Advances in Neural Information Processing Systems 3320–3328 (Montreal, Canada, 2014).
  7. Srivastava, N., Mansimov, E. & Salakhudinov, R. Unsupervised learning of video representations using LSTMs. In Proc. Int. Conference on Machine Learning 843–852 (Lille, France, 2015).

Cited by

Sorted by citations

Related

See all

Metadata

Authors
Su, C., Xu, Z., Pathak, J., & Wang, F.
Affiliations
Weill Cornell Medicine (US)
Corresponding author
Wang, Fei
ORCID
Venue
Translational Psychiatry
Publisher
Springer Science and Business Media LLC
Published
2020-04-22
Volume
10
Issue
1
Pages
116-116
ISSN
2158-3188 (ISSN-L)
DOI
10.1038/s41398-020-0780-3
PMC
PMC7293215
OpenAlex
W3034674374
Cited by
368
References
139
Citations by year
2026: 35 · 2025: 83 · 2024: 74 · 2023: 73 · 2022: 60 · 2021: 38 · 2020: 5
FWCI
6.73
Type
journal article
Language
en
Open access
Licence
Locations
  • Translational Psychiatry (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
Funders
  • National Science Foundation
  • U.S. Department of Health & Human Services | NIH | National Institute of Mental Health
Topics
  • Health Sciences › Health Professions › Health Information Management › Artificial Intelligence in Healthcare
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
Keywords
deep learning, mental health outcomes, depression, artificial intelligence in psychiatry, medical data analysis, social media data, diagnosis and prognosis, clinical decision support
MeSH terms
Deep Learning, Algorithms, Artificial Intelligence, Humans, Mental Health, Outcome Assessment, Health Care

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review