English Abstract
Alzheimer’s disease (AD) presents a significant global health challenge, necessitating accurate and early prediction methods for effective intervention and treatment planning. However, traditional models designed for prediction classification face several challenges, including handling complex data, which neglects many data points for the diagnosis. In this thesis, novel approaches for the prediction of AD are proposed, which are composed of two models. In the first model, we introduced a framework that integrates Neural Process (NPs) with a transformer encoder to model the complex temporal dependencies inherent in longitudinal health data, where our model learns to capture subtle patterns and variations indicative of disease progression. The novelty of our approach lies in the fusion of NPs, renowned for their ability to model stochastic processes, with transformer architectures, which are known for their capacity to capture long-range dependencies. This combination enables our model to effectively adapt to individual patient trajectories and generalize across diverse populations. While the second model involves integration of the previous one which is NPs and a transformer encoder, with the Normalizing Flow (NF), this model is able to combine the effect of the previous one while the addition of the NF made the model able to transform the Gaussian distributions from simple to complex distributions, allowing them to model a wide range of data distributions. We trained our proposed models with the Alzheimer’s Disease Prediction of Longitudinal Evolution dataset (TADPOLE), which contains three classes: Cognitively Normal (CN), Mild Cognitive Impairment (MCI), and AD. The experimental results for the first model demonstrate that the proposed model enhances the prediction of these models in terms of mAUC, Recall, and Precision by 0.937±0.014, 0.920± 0.010, and 0.923 ±0.009, respectively. The addition of NF to the previous model enhances these terms to become mAUC, Precision, and Recall, 0.965 ± 0.006, 0.929 ± 0.007, and 0.929 ± 0.006, respectively. These findings prove the efficacy of the proposed framework in accurately predicting the progression of AD.
