Automated Malaria Detection Using Convolutional Neural Networks and Machine Learning

ABSTRACT: Millions of people suffer from malaria, considered one of the most dangerous parasitic diseases threatening human life, especially in tropical and subtropical regions. There are many challenges in using traditional diagnostic methods such as blood smear checking, which can be achieved by using a microscope due to the inaccuracy of manual analysis and reliance on individual skills. Therefore, utilizing machine learning or deep learning algorithms for automating malaria detection provides encouraging solutions to enhance accuracy, minimize diagnostic time, and empower scalability. This paper employs a Convolutional Neural Network for automating malaria detection and classification using a dataset from the National Institute of Health (NIH) that is available publicly. The proposed model achieves an accuracy of 97.5%, and excellent results are performed regarding sensitivity and specificity when compared with machine learning algorithms such as Support Vector Machine (SVM) and Decision Tree. Furthermore, the results are validated using cross-validation techniques and compared with the existing methods. Our proposed CNN model can be deployed and potentially helps professionals with real-time malaria diagnosis and classification.