Yassine Labyed

The subtle, non-linear degradation of speech is a promising but challenging biomarker for Alzheimer's Disease (AD). This research investigates the modeling capacity of quantum-inspired neural networks for this complex classification task. We present a comprehensive system that processes spontaneous speech through an advanced pipeline of noise reduction, speaker isolation, and feature extraction (MFCCs with deltas) to generate spatio-temporal representations of speech. The core of this study is a direct comparison between a traditional Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model and a novel CNN-Quantum LSTM (CNN-QLSTM) architecture. The QLSTM replaces the classical LSTM cell with a variational quantum circuit, leveraging principles of superposition and entanglement to create a richer representational space for temporal data. Our central hypothesis is that the quantum-inspired framework can more efficiently capture the long-range, intricate dependencies inherent in neuropathological speech patterns, potentially leading to higher classification accuracy and a more robust diagnostic model. Both models are packaged within a reproducible Docker container and exposed via a Flask API to validate their real-world performance and benchmark their computational trade-offs

Enhanced data visualization by developing a dashboard for key performance indicators (KPIs) using Express and Angular .Improved data accessibility and decision-making efficiency by 65% through better visualization and reporting.