100 MRI IMAGE CLASSIFICATION AND SCHIZOPHRENIA RECOGNITION BASED ON IMPROVED ACNET

Abstract

Background Schizophrenia is a complex neurodevelopmental disorder with significant heterogeneity, complex etiology, and an increasing prevalence worldwide. It faces significant challenges in clinical diagnosis and treatment, with missed and misdiagnosed cases occurring from time to time. Improving the diagnostic accuracy of schizophrenia is a prerequisite for early intervention and an important aspect of improving patients’ quality of life and reducing mortality rates. Studies have shown that the pathogenesis of schizophrenia may be related to changes in brain structure and tissue, and the use of neuroimaging technology can effectively provide auxiliary interventions for clinical diagnosis. With the development of machine learning algorithms in recent years, achieving automated medical diagnosis has become a current research hotspot. Considering the neglect of global features in traditional convolutional neural networks for extracting neuroimaging features, research proposes to improve them by using an adaptive asymmetric convolutional network (ACNet) to improve the accuracy of Magnetic Resonance Imaging (MRI) image classification and provide clinical intervention for the recognition of schizophrenia. Methods The study collected MRI image data of schizophrenia patients and healthy control subjects from public datasets and a tertiary hospital. The data was preprocessed using Matlab tools, including image slicing, weighted averaging, and extraction of texture features from gray level co-occurrence matrix. Afterwards, based on the convolutional neural network structure, sparse interaction mechanism and attention mechanism are introduced to enhance feature extraction ability and classification performance. Train the ACNet model using preprocessed MRI image data, and evaluate diagnostic results and classification accuracy using metrics such as accuracy and area under the curve. Results In Table 1, the improved model can effectively improve the classification accuracy of MRI images and the recognition accuracy of schizophrenia, with accuracy rates exceeding 94%, and the application effect is significant. Discussion Improving the MRI image classification method of ACNet provides a new approach for computer-aided diagnosis of schizophrenia. The model has shown significant accuracy in feature extraction and classification performance, with values exceeding 90%. MRI, as a non-invasive brain imaging technique, is widely used in brain examinations due to its ability to achieve arbitrary slicing in multiple directions. The core symptoms of schizophrenia include hallucinations, delusions, disordered thinking, and emotional apathy. These symptoms typically last for at least six months and severely affect the patient’s daily life and social functioning. The application of MRI in the identification of schizophrenia is becoming increasingly widespread. With the help of machine learning and deep learning techniques, MRI images can be automatically analyzed and classified to assist doctors in making more accurate diagnoses. The application of the improved method proposed in the study in MRI images can provide strong support for early intervention and precise treatment of the disease, help to better understand the neuropathological mechanism of schizophrenia, and improve the quality of life of patients.