RETRACTED ARTICLE: MRI feature engineering and SVM framework for schizophrenia recognition.

Area of Science:

• Neuroimaging
• Machine Learning
• Psychiatry

Background:

• Traditional schizophrenia diagnosis relies on subjective clinical evaluation, lacking objective quantitative data.
• Existing machine learning methods for neuroimaging face challenges with high-dimensional, small-sample MRI data, including low feature extraction automation and poor model generalization.
• Early diagnosis of schizophrenia is critical for improving patient prognosis and reducing societal burden.

Purpose of the Study:

• To develop an objective, quantitative framework for early schizophrenia recognition using MRI data.
• To address limitations in current machine learning approaches for neuroimaging data in schizophrenia diagnosis.
• To enhance the accuracy and generalizability of schizophrenia detection models.

Main Methods:

• Proposed a framework combining MRI feature engineering and Support Vector Machines (SVM) for schizophrenia recognition.
• Implemented preprocessing steps (skull stripping, data registration) to reduce inter-individual structural differences.
• Extracted macroscopic statistical features, optimized using feature masking for key regions of interest, and analyzed with SVM.

Main Results:

• Achieved an average classification accuracy of 95.00% on the COBRE dataset using five-fold cross-validation.
• Demonstrated superior performance compared to six mainstream machine learning algorithms across multiple metrics.
• Validated the effectiveness of the proposed MRI feature engineering and SVM framework.

Conclusions:

• The study presents an objective and innovative approach for the auxiliary diagnosis of schizophrenia.
• The findings provide strong support for the application of this method in early schizophrenia intervention practices.
• This framework offers a more reliable and automated method for schizophrenia detection compared to traditional approaches.