Structural MRI-Based Schizophrenia Classification Using
Autoencoders and 3D Convolutional Neural Networks in
Combination with Various Pre-Processing Techniques

J 2022

Structural MRI-Based Schizophrenia Classification Using Autoencoders and 3D Convolutional Neural Networks in Combination with Various Pre-Processing Techniques

VYŠKOVSKÝ, Roman, Daniel SCHWARZ, Vendula CHUROVÁ and Tomáš KAŠPÁREK

Basic information

Original name

Structural MRI-Based Schizophrenia Classification Using Autoencoders and 3D Convolutional Neural Networks in Combination with Various Pre-Processing Techniques

Authors

VYŠKOVSKÝ, Roman (203 Czech Republic, guarantor, belonging to the institution), Daniel SCHWARZ (203 Czech Republic, belonging to the institution), Vendula CHUROVÁ (203 Czech Republic, belonging to the institution) and Tomáš KAŠPÁREK (203 Czech Republic, belonging to the institution)

Edition

Brain Sciences, Basel, MDPI, 2022, 2076-3425

Other information

Language

English

Type of outcome

Článek v odborném periodiku

Field of Study

30103 Neurosciences

Country of publisher

Switzerland

Confidentiality degree

není předmětem státního či obchodního tajemství

References:

URL

Impact factor

Impact factor: 3.300

RIV identification code

RIV/00216224:14110/22:00129674

Organization unit

Faculty of Medicine

DOI

http://dx.doi.org/10.3390/brainsci12050615

UT WoS

000803514900001

Keywords in English

schizophrenia; classification; 3D CNN; autoencoders; voxel-based morphometry; deformation-based morphometry; deep learning

Abstract

V originále

Schizophrenia is a severe neuropsychiatric disease whose diagnosis, unfortunately, lacks an objective diagnostic tool supporting a thorough psychiatric examination of the patient. We took advantage of today's computational abilities, structural magnetic resonance imaging, and modern machine learning methods, such as stacked autoencoders (SAE) and 3D convolutional neural networks (3D CNN), to teach them to classify 52 patients with schizophrenia and 52 healthy controls. The main aim of this study was to explore whether complex feature extraction methods can help improve the accuracy of deep learning-based classifiers compared to minimally preprocessed data. Our experiments employed three commonly used preprocessing steps to extract three different feature types. They included voxel-based morphometry, deformation-based morphometry, and simple spatial normalization of brain tissue. In addition to classifier models, features and their combination, other model parameters such as network depth, number of neurons, number of convolutional filters, and input data size were also investigated. Autoencoders were trained on feature pools of 1000 and 5000 voxels selected by Mann-Whitney tests, and 3D CNNs were trained on whole images. The most successful model architecture (autoencoders) achieved the highest average accuracy of 69.62% (sensitivity 68.85%, specificity 70.38%). The results of all experiments were statistically compared (the Mann-Whitney test). In conclusion, SAE outperformed 3D CNN, while preprocessing using VBM helped SAE improve the results.

Links

NV17-33136A, research and development project

Name: Neurominer: odhalování skrytých vzorů v datech ze zobrazování mozku

Citovat

VYŠKOVSKÝ, Roman, Daniel SCHWARZ, Vendula CHUROVÁ and Tomáš KAŠPÁREK. Structural MRI-Based Schizophrenia Classification Using Autoencoders and 3D Convolutional Neural Networks in Combination with Various Pre-Processing Techniques. Brain Sciences. Basel: MDPI, 2022, vol. 12, No 5, p. 1-16. ISSN 2076-3425. Available from: https://dx.doi.org/10.3390/brainsci12050615.

@article{1883020,
   author = {Vyškovský, Roman and Schwarz, Daniel and Churová, Vendula and Kašpárek, Tomáš},
   article_location = {Basel},
   article_number = {5},
   doi = {http://dx.doi.org/10.3390/brainsci12050615},
   keywords = {schizophrenia; classification; 3D CNN; autoencoders; voxel-based morphometry; deformation-based morphometry; deep learning},
   language = {eng},
   issn = {2076-3425},
   journal = {Brain Sciences},
   title = {Structural MRI-Based Schizophrenia Classification Using Autoencoders and 3D Convolutional Neural Networks in Combination with Various Pre-Processing Techniques},
   url = {https://www.mdpi.com/2076-3425/12/5/615},
   volume = {12},
   year = {2022}
}

TY  - JOUR
ID  - 1883020
AU  - Vyškovský, Roman - Schwarz, Daniel - Churová, Vendula - Kašpárek, Tomáš
PY  - 2022
TI  - Structural MRI-Based Schizophrenia Classification Using Autoencoders and 3D Convolutional Neural Networks in Combination with Various Pre-Processing Techniques
JF  - Brain Sciences
VL  - 12
IS  - 5
SP  - 1-16
EP  - 1-16
PB  - MDPI
SN  - 20763425
KW  - schizophrenia
KW  - classification
KW  - 3D CNN
KW  - autoencoders
KW  - voxel-based morphometry
KW  - deformation-based morphometry
KW  - deep learning
UR  - https://www.mdpi.com/2076-3425/12/5/615
N2  - Schizophrenia is a severe neuropsychiatric disease whose diagnosis, unfortunately, lacks an objective diagnostic tool supporting a thorough psychiatric examination of the patient. We took advantage of today's computational abilities, structural magnetic resonance imaging, and modern machine learning methods, such as stacked autoencoders (SAE) and 3D convolutional neural networks (3D CNN), to teach them to classify 52 patients with schizophrenia and 52 healthy controls. The main aim of this study was to explore whether complex feature extraction methods can help improve the accuracy of deep learning-based classifiers compared to minimally preprocessed data. Our experiments employed three commonly used preprocessing steps to extract three different feature types. They included voxel-based morphometry, deformation-based morphometry, and simple spatial normalization of brain tissue. In addition to classifier models, features and their combination, other model parameters such as network depth, number of neurons, number of convolutional filters, and input data size were also investigated. Autoencoders were trained on feature pools of 1000 and 5000 voxels selected by Mann-Whitney tests, and 3D CNNs were trained on whole images. The most successful model architecture (autoencoders) achieved the highest average accuracy of 69.62% (sensitivity 68.85%, specificity 70.38%). The results of all experiments were statistically compared (the Mann-Whitney test). In conclusion, SAE outperformed 3D CNN, while preprocessing using VBM helped SAE improve the results.
ER  -

VYŠKOVSKÝ, Roman, Daniel SCHWARZ, Vendula CHUROVÁ and Tomáš KAŠPÁREK. Structural MRI-Based Schizophrenia Classification Using Autoencoders and 3D Convolutional Neural Networks in Combination with Various Pre-Processing Techniques. \textit{Brain Sciences}. Basel: MDPI, 2022, vol.~12, No~5, p.~1-16. ISSN~2076-3425. Available from: https://dx.doi.org/10.3390/brainsci12050615.

Detailed Information on Publication Record