PROJECT TITLE :

Stroke Risk Prediction with Hybrid Deep Transfer Learning Framework

ABSTRACT:

There is currently no treatment that is proven to be effective for stroke, despite the fact that it has become the leading cause of death and long-term disability in the world. Deep learning-based approaches have the potential to outperform the existing models for predicting the risk of stroke; however, in order to function properly, they require large amounts of data that have been labeled. Stroke data is typically broken up into manageable chunks before being sent to various hospitals. This is done to comply with the stringent privacy protection policy that is prevalent in health-care systems. In addition to this, there is a significant imbalance between the positive and negative examples of such data. Transfer learning is useful for solving problems with small amounts of data because it draws on the prior knowledge of a correlated domain. This is especially true when more than one source of data is available. In this work, we propose a novel scheme to exploit the knowledge structure from multiple correlated sources by using Hybrid Deep Transfer Learning-based Stroke Risk Prediction (HDTL-SRP) (i.e., external stroke data, chronic diseases data, such as hypertension and diabetes). The proposed framework has been put through rigorous testing in both simulated and real-world settings, and it has been shown to perform better than the most recent generation of stroke risk prediction models. In addition to this, it demonstrates the potential for deployment in the real world across multiple hospitals with the assistance of 5 G/B5G infrastructures.