An Intelligent Medical Model for Classification of Brain Tumours and Stroke Lesions Using Machine Learning in Healthcare for Resource-Constrained Devices

Ahed Abugabah, Zayed University

Document Type

Article

Source of Publication

SN Computer Science

Publication Date

6-1-2025

Abstract

Brain tumors and stroke lesions are major health problems and need to be diagnosed on time and accurately to improve patient outcomes. Nevertheless, resource-constrained devices like portable medical equipment and embedded systems are limited by computational resources that prevent the deployment of sophisticated deep-learning models. We address the problem of efficient and accurate brain tumor and stroke lesion classification on such devices. To realize this, we present a light weight deep learning framework using MobileNetV3 for feature extraction and a hybrid optimization algorithm based on Bat Algorithm (BA) and Differential Evolution (DE) to fine-tune the hyper parameters. The framework relies on quantization techniques and a compressed representation of medical images to reduce memory and computational overhead. Brain tumor Magnetic Resonance Imaging (MRI) scan datasets (BraTS 2021), stroke lesion datasets (ISLES 2022), and others were preprocessed with normalization and data augmentation to be robust. To evaluate the performance of the model under limited conditions, it was deployed on Raspberry Pi 4 and other edge devices. The proposed framework achieves an accuracy of 96.3% for brain tumor classification and 94.8% for stroke lesison detection, with an inference time of 2.4 s per image on Raspberry Pi 4, which outperforms state-of-the-art methods in both accuracy and computational efficiency. Furthermore, the hybrid BA-DE optimization reduced model size by 28% without significant loss of accuracy. This study demonstrates that the proposed lightweight framework effectively balances accuracy and computational efficiency, making it suitable for real-time applications in resource-constrained healthcare environments. The results highlight its potential to empower low-resource medical facilities with advanced diagnostic capabilities.

ISSN

2662-995X

Publisher

Springer Science and Business Media LLC

Volume

6

Issue

5

Disciplines

Computer Sciences

Keywords

Brain tumor, Deep learning, Healthcare, Machine learning, Resource constrained devices, Stroke lesion

Scopus ID

05002872865

Indexed in Scopus

yes

Open Access

no

Link to Full Text

Share

COinS