Malaria Parasite Detection using Convolutional Neural Networks

Research Supervisor
Journal Title
Journal Title
Journal ISSN
Volume Title
Malaria is a parasitic disease spread by Plasmodium parasites that can be lethal. Deep learning algorithms can be used to perform this analysis automatically. We argued for a Convolutional Neural Network-based automated malaria parasite identification approach (CNN). We tested numerous well-known CNN architectures, including ResNet18, MobileNetV3, DenseNet121, and AlexNet. Then we created our own model and compared it to other well-known models. This analysis might be done automatically using modern deep learning tech niques. With the construction of an autonomous, precise, and efficient model, the demand for trained staff can be considerably decreased. For the diagnosis of malaria using cell pictures, we suggested a totally automated Convolutional Neural Network (CNN) based model.This report makes two important contributions. To begin, we assess the efficacy of various existing deep learning models for malaria detection. Second, we offer a customized CNN model that outperforms all deep learning models that have been observed. Before training the model, it uses filtering and image augmentation techniques to emphasize features of red blood cells. The tailored CNN model is generalized and prevents over-fitting thanks to image augmentation techniques. The suggested method is 96.82 percent accurate in detecting malaria from microscopic blood smears, according to the findings of all experimental evaluations on the benchmark NIH Malaria Dataset.With a 96.21 percent accuracy, our deep learning-based model can recognize malarial parasites in cell pictures.
Department Name
Electrical and Computer Engineering
Printed Thesis