Malaria Detection Using Residual Network
Kannan Arun, Sebastian Terence, Titus, Jude Immaculate, Angeline Lydia · 2024
The millions of people are at risk of contracting malaria and dying from it, making it a persistent global health concern. For treatment and prevention to be successful, timely and accurate diagnosis is essential. Accessibility and efficiency issues plague traditional techniques such as microscopic evaluation and quick diagnostic testing. This study investigates the use of cutting-edge technologies, such as deep learning and machine learning, to improve malaria detection. This paper presents the XResNet-152 architecture, which is a potent instrument renowned for its efficacy in image classification applications. This model demonstrates the effectiveness of novel techniques, such as batch normalization and weight start, to accurately detect malaria. This study assesses the performance of several deep learning architectures, including ResNet variations, and compares them across training epochs. With a dataset of 27,558 images, the XResNet-152 model demonstrates its resilience and ongoing development by reaching a peak accuracy of 98% after 100 epochs. A thorough examination that includes a confusion matrix demonstrates how well the model classifies instances in various classes. The results highlight the XResNet-152 model's potential as a trustworthy and effective tool for malaria detection, making a major contribution to international healthcare initiatives. To improve the model's relevance in real-world situations, more research could concentrate on deployable and optimization techniques. Particularly in areas where outbreaks are likely, the effectiveness of this model signifies a significant advancement in the management of malaria.