An Advanced Framework for the Early Detection of Neurodegenerative Disorders Using a Latent Encoder Coupled Generative Adversarial Network Optimized

Abstract

Neurodegenerative disorders (NDs) are progressive conditions that affect the central nervous system, leading to gradual neuronal loss and cognitive decline. The cerebrospinal fluid (CSF) biomarkers and clinical data are useful indicators of identifying these disorders, and this gives information regarding the changes in biochemicals and the health condition of the patient. Nevertheless, diagnosis occurs early because it has vague initial symptoms, diverse clinical manifestations, and the shortcomings of conventional diagnostic tools. To overcome these challenges, this research presents an advanced framework for the early detection of NDs using a Latent Encoder Coupled Generative Adversarial Network optimized with Ruppell's Fox Optimizer (LECouGAN-RFO). The methodology begins with data collection, including CSF biomarker measurements (amyloid- $\beta$, tau, phosphorylated tau) obtained from lumbar puncture samples and clinical information such as age, gender, cognitive scores, medical history, and symptoms, sourced from Alzheimer's disease Neuroimaging Initiative (ADNI) and Parkinson's Progression Markers Initiative (PPMI). Preprocessing is performed using a Modified Square Root Sage-Husa Adaptive Kalman Filter (MSR-SHAKF) to handle missing values, normalize biomarker levels, and reduce noise. The cleaned data is then fed into a Dual Aggregation Transformer (DAT) for effective multimodal feature extraction. The extracted features are classified using Latent Encoder Coupled Generative Adversarial Network (LECouGAN), and the loss function is optimized via Ruppell's Fox Optimizer (RFO) to enhance convergence and performance. The experimental findings based on ADNI and PPMI data- sets show that the proposed framework has higher accuracy (99.8%, 99.7%) compared with existing approaches. The ROC curve and accuracy plots show that the proposed framework has been well trained with less overfitting and has high accuracy for the problem. The proposed framework is useful for the early diagnosis of Alzheimer's and Parkinson's diseases.