Semi-Supervised Support Vector Machine for Digital Twins Based Brain Image Fusion.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Neuroscience

Background:

• Brain image analysis is crucial for diagnosis and forecasting.
• Digital Twins (DTs) offer a virtual environment for complex data modeling.
• Semi-Supervised Support Vector Machines (S3VMs) can leverage limited labeled data.

Purpose of the Study:

• To develop and evaluate a brain image fusion Digital Twin (DT) model for feature recognition, diagnosis, and forecasting.
• To explore the performance of S3VMs in conjunction with an improved AlexNet for brain image analysis.
• To establish a robust diagnostic and predictive framework for brain imaging data.

Main Methods:

• Integration of S3VMs with an improved AlexNet model for brain image analysis.
• Mapping real-space brain Magnetic Resonance Imaging (MRI) data to a virtual space using DTs.
• Comparative analysis against state-of-the-art models including LSTM, CNN, RNN, AlexNet, and MLP.

Main Results:

• The proposed model achieved 92.52% accuracy in feature recognition and extraction, surpassing other models by at least 2.76%.
• Achieved low Root Mean Square Error (RMSE) of 4.91% and Mean Absolute Error (MAE) of 5.59%.
• Demonstrated superior performance in segmentation and fusion metrics, including a Jaccard coefficient of 79.55% and Dice Similarity Coefficient (DSC) of 75.58%.

Conclusions:

• The developed brain image fusion DT model offers high accuracy and efficiency for feature recognition and digital diagnosis.
• The improved AlexNet component shows significant acceleration efficiency for processing large brain image datasets.
• The model provides a strong experimental basis for advancing brain image analysis and clinical decision-making.