Malaria Parasite Detection Using Deep Neural Networks

Malaria, a contagious disease that has the potential to be fatal, affects millions of individuals globally. Early identification of malaria is one of the main difficulties in its efficient treatment. Microscopy is one of the older diagnostic techniques for malaria, although it has drawbacks such as limited sensitivity and specificity. Deep learning algorithms have recently demonstrated encouraging outcomes in the detection of malaria from photographs of infected and uninfected cells. To choose the most effective method for detecting malaria, we assess the performance of five alternative deep learning models in this study. The dataset used in this study contains many images of both infected and uninfected cells, the difference between these images is clearly visible when inspected visually. This research used the dataset from Kaggle that has two classes which are uninfected and parasitized images of blood cells. The study compared the performance of five different deep learning models in detecting malaria from these images: VGG-19, Inception-v3, ResNet-101, Inception Resnet-v2, and a custom CNN model. The results of the study indicated that Inception Resnet-v2 was the most accurate model, achieving an accuracy rate of 95.4%.