Deep learning driven automated malaria parasite detection in thin blood smears
Aryan Verma, Sejal Mansoori, Adithya Srivastava, Priyanka Rathee, Nagendra Singh · 2023
Malaria has remained a recurrent challenge in subtropical and tropical areas across the globe. Malaria is caused by plasmodium parasites and can become fatal for the infected person; hence, its precise diagnosis within a short time frame becomes crucial in ending or controlling it. Malaria is analyzed by conventional microscopy, but this methodology has its draw-backs. It is time-consuming and requires an efficient skill set that is often unavailable in many health facilities in several malaria-endemic countries. Hence, tools like computer-aided diagnosis are momentous in analyzing cell images in the modern era of technology. In this paper, we employ deep neural networks to detect malaria presence in the cell images. Both uninfected and parasitized image samples are used to train the models. We use the holdout Validation technique with 15% and 20% validation data to train and compare four variant transfer learned models, VGG19, Resnet50, InceptionV1, and InceptionV3. We employ data augmentation techniques and early stopping callbacks for training the final model Inception V3, which receives 91.40% accuracy, which is better than most of the existing studies in the domain.