Implementation of Convolutional Neural Network Malarial Cells Detection

This paper proposes a Convolutional Neural Network (CNN) approach to analyze and detect the malarial parasite-infected blood smear cells. Malaria is a fatal illness solely transmits through the bites of infected female mosquitoes of Anopheles.. Recent studies show that in 2020, there were 241 million cases of malaria worldwide, which resulted in the death of nearly 6,27,000 people. The diagnostic process must be automated to avoid human participation during the automated diagnosis because a delayed or inaccurate diagnosis causes most of these deaths. To enhance diagnostic reliability, deep-learning technologies and CNN, such as medical image processing techniques, are employed to assess parasitemia in microscopic blood slides. In this research, we propose a supervised learning-based Visual Geometry Group (VGG-19) that performs accurate classification to detect malaria-infected cells. The dataset comprises 27,560 images of segmented blood cells, equally divided into parasitized (infected) and uninfected cells, which were utilized for VGG-19 architecture. The first step is to define the image processing methods that can be used to analyze the dataset for training the model. The next stage discusses the techniques for training deep neural networks and the data augmentation techniques used to increase the size of the dataset and enhance the model’s performance. Finally, the accuracy of classification outcomes is compared from deep CNN using the same datasets for the testing, training and validating phases. Our trained model uses blood smear samples to predict the presence of malarial-infected cells and achieves a 97% accuracy rate.