Improved Weighted Quantum Firefly Optimization With Vanilla Vision Transformer and Big Data for Precision Diagnosis and Biomarker Identification in Neurodegenerative Disorders

Abstract

Due to their varied symptoms and cumulative nature, neurodegenerative illnesses present significant challenges to rapid identification and biomarker discovery. To address these issues, this work introduces an advanced system that integrates large-scale information analytics, a Vanilla Vision Transformer (VViT), and Improved Weighted Quantum Firefly Optimization (IWQFO) to enhance panoptic categorization in neuroimaging. The VViT effectively captures both local and global information through self-attention mechanisms, while the IWQFO method improves hyperparameter optimization, leading to shorter convergence times and enhanced global search capabilities. By leveraging large volumes of data, the model can generalize across diverse patient demographics and imaging techniques. Experimental evaluations were conducted using benchmark neuroimaging databases. The proposed architecture outperformed existing CNN-based models and more recent transformer-based techniques, achieving a Dice Similarity Coefficient (DSC) of 94.7%, an Intersection over Union (IoU) of 92.3%, and an accuracy of 95.1%. Compared to existing optimization techniques, convergence time was reduced by 18% through IWQFO-based hyperparameter tuning. Ablation studies confirmed the efficiency of each component, demonstrating that big data integration enhances model stability and that the VViT plays a crucial role in detecting subtle neurodegeneration patterns. The proposed approach offers a promising tool for the early detection and effective treatment of neurological conditions, thanks to its higher segmentation precision, faster convergence, and improved diagnostic accuracy.