Brain Stroke Classification Using Wavelet Transform and MLP Neural Networks on DWI MRI Images

TL;DR

This paper introduces a lightweight, efficient method for classifying brain stroke types from DWI MRI images using Wavelet Transform for feature extraction and an MLP neural network, achieving high accuracy with low computational cost.

Contribution

It proposes a novel combination of Wavelet Transform and a compact MLP for stroke classification, reducing computational complexity compared to CNN-based methods.

Findings

Achieved 86% accuracy with Haar wavelet.

Demonstrated a balance between accuracy and efficiency.

Provided a practical solution for resource-limited clinical settings.

Abstract

This paper presents a lightweight framework for classifying brain stroke types from Diffusion-Weighted Imaging (DWI) MRI scans, employing a Multi-Layer Perceptron (MLP) neural network with Wavelet Transform for feature extraction. Accurate and timely stroke detection is critical for effective treatment and improved patient outcomes in neuroimaging. While Convolutional Neural Networks (CNNs) are widely used for medical image analysis, their computational complexity often hinders deployment in resource-constrained clinical settings. In contrast, our approach combines Wavelet Transform with a compact MLP to achieve efficient and accurate stroke classification. Using the "Brain Stroke MRI Images" dataset, our method yields classification accuracies of 82.0% with the "db4" wavelet (level 3 decomposition) and 86.00% with the "Haar" wavelet (level 2 decomposition). This analysis highlights a balance between diagnostic accuracy and computational efficiency, offering a practical solution for automated stroke diagnosis. Future research will focus on enhancing model robustness and integrating additional MRI modalities for comprehensive stroke assessment.