Development of Deep Learning Models for Rapid and Accurate Malaria Parasite Detection in Blood Smear Image

Abstract

This study aims to develop an efficient model using deep learning techniques for the rapid and accurate diagnosis of malaria in blood smear images. The analysis follows the Cross-Industry Standard Process for Data Mining (CRISP-DM), employing forward selection and backward elimination feature selection techniques combined with four classification methods: k-Nearest Neighbors, Neural Network, Deep Learning, and Random Forest. The performance of these models is compared using metrics such as accuracy, overall performance, speed, and specificity. The results indicate that the deep learning-based feature selection technique achieves an accuracy of 97.75%, an overall performance of 97.73%, a speed of 97.19%, and a specificity of 98.31%. This technique is suitable for developing a malaria patient screening model, reducing delays in patient care in remote and underserved areas. It supports the formulation of treatment policies by multidisciplinary teams and enables rapid and accurate malaria diagnosis using artificial intelligence techniques. This can ensure timely and effective treatment for malaria patients, ultimately reducing malaria-related mortality in the future.