With the privacy concern of mental patients’ records as it is
protected with federal privacy legislation, this paper proposes an
optimized federated learning model for schizophrenia detection
from functional magnetic resonance imaging (fMRI) and
structural magnetic resonance imaging (sMRI) outputs. As the
diagnosis of schizophrenia has no biological indicator, this study
investigated the human brain's functional and structural defense
for the disorder. fMRI and sMRI have an effective contribution to
the diagnosis of schizophrenia and differentiate it from a healthy
one. The proposed models predict schizophrenic states among The
10th annual MLSP competition data. To improve the classification
of traditional models on magnetic resonance data, meta-heuristic
models are proposed to improve the classification accuracy and the
generalization ability. The features selected by the swarm
intelligence algorithms are used as the most influential factors in
the creation of machine learning algorithms and the evaluation of
the proposed hybrid models. The proposed federated learning
models and a hybrid K-NN model reached 100% of accuracy, and
area under curve metrices.
|
