Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 40 (3), 944-954

Using Deep Autoencoders to Identify Abnormal Brain Structural Patterns in Neuropsychiatric Disorders: A Large-Scale Multi-Sample Study


Using Deep Autoencoders to Identify Abnormal Brain Structural Patterns in Neuropsychiatric Disorders: A Large-Scale Multi-Sample Study

Walter H L Pinaya et al. Hum Brain Mapp.


Machine learning is becoming an increasingly popular approach for investigating spatially distributed and subtle neuroanatomical alterations in brain-based disorders. However, some machine learning models have been criticized for requiring a large number of cases in each experimental group, and for resembling a "black box" that provides little or no insight into the nature of the data. In this article, we propose an alternative conceptual and practical approach for investigating brain-based disorders which aim to overcome these limitations. We used an artificial neural network known as "deep autoencoder" to create a normative model using structural magnetic resonance imaging data from 1,113 healthy people. We then used this model to estimate total and regional neuroanatomical deviation in individual patients with schizophrenia and autism spectrum disorder using two independent data sets (n = 263). We report that the model was able to generate different values of total neuroanatomical deviation for each disease under investigation relative to their control group (p < .005). Furthermore, the model revealed distinct patterns of neuroanatomical deviations for the two diseases, consistent with the existing neuroimaging literature. We conclude that the deep autoencoder provides a flexible and promising framework for assessing total and regional neuroanatomical deviations in neuropsychiatric populations.

Keywords: autism spectrum disorder; computational psychiatry; deep autoencoder; deep learning; schizophrenia; structural MRI.


Figure 1
Figure 1
The semi‐supervised deep autoencoder structure. During the training, the deep autoencoder learns to reconstruct the input data and to predict the observed variables y, in this case, the subject's age and sex
Figure 2
Figure 2
(a) The mean learning curve of the best structure (100–75–100) along the 10‐fold cross‐validation. (b) The mean absolute error curve of age prediction of the best configuration along the 10‐fold cross‐validation. (c) The balanced accuracy curve of sex prediction of the best configuration along the 10‐fold cross‐validation [Color figure can be viewed at]
Figure 3
Figure 3
Boxplot of the deviation metric (mean squared reconstruction error) from the patients with schizophrenia group and the healthy controls subjects (NUSDAST data set) and from patients with autism spectrum disorder and the corresponding healthy control group (ABIDE data set). ASD = autism spectrum disorder; HC = healthy controls; SCZ = schizophrenia [Color figure can be viewed at]

Similar articles

See all similar articles

Cited by 1 article


    1. Abadi M, Agarwal A, Barham P, Brevdo E, Chen Z, Citro C, Corrado GS, Davis A, Dean J, Devin M (2016). Tensorflow: Large‐scale machine learning on heterogeneous distributed systems. arXiv Preprint, arXiv160304467.
    1. Abou‐Saleh M. T. (2006). Neuroimaging in psychiatry: An update. Journal of Psychosomatic Research, 61, 289–293. - PubMed
    1. Bhalla M., & Mahmood H. (2015). Assessing accuracy of automated segmentation methods for brain lateral ventricles in MRI data. Queen's Science Undergraduate Research Journal, 1, 25–30.
    1. Cauda F., Geda E., Sacco K., D'agata F., Duca S., Geminiani G., & Keller R. (2011). Grey matter abnormality in autism spectrum disorder: An activation likelihood estimation meta‐analysis study. Journal of Neurology, Neurosurgery, and Psychiatry, 82, 1304–1313. - PubMed
    1. Cheung B, Livezey JA, Bansal AK, Olshausen BA (2014): Discovering hidden factors of variation in deep networks. arXiv.1412.6583: 1–10.

Publication types