Malaria Parasite Detection Using CNN-Based Ensemble Technique on Blood Smear Images

The parasite genus Plasmodium found in red blood cells of an infected person causes malaria, a disease that is transmitted by female Anopheles mosquitoes. The gold standard for diagnosing malaria is by reviewing blood smear under a mi-croscope. By aiding with triage and disease diagnosis, computer-aided diagnostic (CADx) methods and tools incorporating ma-chine learning (ML) algorithms on microscopic blood smear images possess the capability to lessen therapeutic encumbrance. On blood smear images, prominent pre-processing techniques have been performed, and two benchmark architectures have been applied along with ensemble technique. The ensemble approach was used to enhance the accuracy and performance of our suggested method. The NIH Malaria Dataset is used for all experimental evaluations. According to the results, the proposed approach has a 96.71 percent accuracy rate for identifying malaria in microscopic blood smears, with precision of 97.44 percent, specificity of 97.48 percent, F1-score of 96.69 percent, and Cohen's Kappa coefficient of 99.39 percent. Our study supports that convolutional neural networks (CNNs) can be used to quickly diagnose red blood cells (RBCs) with malaria parasite infection from segmented microscopic blood smear images, which is advantageous in regions with a shortage of medical personnel.