High Dimensional Survival Analysis Model for Predicting Diagnosis of alzheimer’s disease over Specific Period of time

Abstract Survival analysis is a statistical tool for predicting the period until an event such as death or disease diagnosis. It requires the availability of censored data demonstrating that the event of interest did not occur during the study period. The model faces issues of overfitting if it has more features than data. However, not all elements are necessary for addressing the problem, and incorporating non-essential aspects might occasionally degrade learning performance. As a result, constructing an accurate survival model using electronic health records is difficult. With this rationale, a hybrid approach for high-dimensional survival analysis is developed. First, labeled (Alzheimer’s Disease) and unlabeled (Normal Cognitive) instances are used to offer better representation with lower dimensions from clinical features, and then we actively train the survival model by labeling the censored data. The effectiveness of this strategy was evaluated using a c-index study. The results show that our method outperforms baseline models by a wide margin.