A Machine Learning Model for Prediction of Malaria from Microscopic Blood Cell Images

Abstract

Machine learning (ML), a pivotal subset of artificial intelligence (AI), has been instrumental in developing systems that learn from and enhance their performance based on data. This technology has facilitated the creation of predictive models for a range of diseases, including pneumonia, heart disease, and COVID-19, allowing for diagnosis based on specific data parameters. Traditionally, malaria detection has relied on the microscopic examination of blood cells, a method that is not only time-consuming but also prone to misdiagnosis, leading potentially to the incorrect treatment of patients. This study introduces a machine-learning model designed to identify patterns in microscopic images of blood cells to differentiate between healthy and malaria-infected cells. Utilizing publicly available datasets from Kaggle for both training and testing, we evaluated the performance of K-Nearest Neighbor (KNN) and Convolutional Neural Network (CNN) algorithms. Our findings show that the CNN model significantly outperformed the KNN model, achieving an accuracy of 94%, and requiring less memory and computational time. Therefore, the CNN model is recommended for further development and use. The implementation of this model represents a significant step forward in achieving the sustainable development goals related to health and wellness, offering a more reliable, efficient, and scalable tool for malaria diagnosis.