Linear regressive weighted Gaussian kernel liquid neural network for brain tumor disease prediction using time series...

Area of Science:

• Artificial Intelligence
• Medical Diagnostics
• Computational Neuroscience

Background:

• Brain tumors, abnormal cell growths, pose significant diagnostic challenges.
• Early detection is critical for improving patient survival and treatment efficacy.
• Existing machine learning models struggle with accuracy and efficiency in brain tumor prediction.

Purpose of the Study:

• To introduce a novel Linear Regressive Weighted Gaussian Kernel Liquid Neural Network (LRWGKLNN) model for brain tumor prediction.
• To enhance the accuracy and reduce the time complexity of brain tumor disease detection using time-series data.
• To address limitations in conventional machine and deep learning models for early brain tumor diagnosis.

Main Methods:

• Data acquisition and preprocessing, including linear regression for missing value imputation and Generalized Extreme Studentized Deviation for outlier removal.
• Feature selection using the Cosine Congruence Weighted Majority Algorithm to minimize prediction time.
• Classification using the Gaussian Kernelized Liquid Neural Network with selected features from time-series data.

Main Results:

• The LRWGKLNN model demonstrated improved performance with higher accuracy, precision, recall, specificity, and F1 scores.
• Achieved a 16% reduction in time consumption for feature selection compared to existing deep learning methods.
• Experimental evaluation confirmed the model's effectiveness on various performance metrics.

Conclusions:

• The proposed LRWGKLNN model offers a significant advancement in accurate and efficient brain tumor prediction.
• This novel approach enhances early diagnosis capabilities, potentially improving patient outcomes.
• The model's efficiency in feature selection and classification makes it a promising tool for clinical application.