A Super-Resolution MRI Pipeline for Predictive Stroke Risk Evaluation Using ESRGAN and DQN Models

View at publisher Summarise with VIRAT
Authors: Athim R, Akilan B, Gnanasoundhram J
Cited by: 0
References: 15

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Stroke is one of the most common causes of death and lasting disability and the discovery of the factors that are putting people at risk at an early age is important. This project is an implementation of a system in MATLAB that in turn combines the Enhanced Super-Resolution Generative Adversarial Network (ESRGAN) with a Deep QNetwork (DQN) predictive model in stroke-susceptibility analysis based on high-definition MRI scan data. ESRGAN improves the quality of the MRI, and finer elements of structure are sharpened, including the tissues of the brain, blood vessels, and any possible lesions. Attributes extracted around these regions are then fed into the DQN model to be predicted so as to be able to predict stroke probability with accuracy. A realistic Kaggle brain MRI dataset was tested on the system and comparison was made between the classical methods of machine learning and the deep learning models, showing that deep learning models exhibit better precision and earlier detection of a pathology. The solution has practical real-life implementation, which can be identified by at-risk people at earlier stages of age and aid in the execution of preventive and treatment approaches on a personal level. This approach is different to the traditional methods because it uses super-resolution enhancement and reinforcement learning.

Fields of Study & Subjects

Brain Tumor Detection and Classification Artificial Intelligence in Healthcare Acute Ischemic Stroke Management

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
1435–1440
Publication date
21 January 2026
Language
EN