Malaria Cell Image Classification Using Convolutional Neural Networks (CNNs)
Drishti Agarwal, K. Sashanka, Sajal Madan, Akshay Kumar, Preeti Nagrath, Rachna Jain · Lecture notes on data engineering and communications technologies · 2021
This study provides an insight into malaria as a disease. Malaria is a disease caused due to plasmodium parasite. It requires a type of mosquito as its host. Hence, the bite of the mosquito leads to malaria. The impact it carries on the health of people around the world is extremely large and cannot be curtailed or controlled without quick and efficient diagnostics and treatment. Subsequent topics dwell on the constraints of detection of the malaria parasite. These constraints may include problems with the feasibility of certain types of tests, or not having access to a diagnostics center or problems with transportation of necessary infrastructure. We also must understand that traditional prognosis methods are very tedious and hence always have a chance for human error or oversight leading to devastating consequences. The ease or simplification of diagnosis of malaria upon the use of machine learning and deep learning is undeniable; hence, in our project, we aim to create a model, using CNN, which using feature extraction, can predict whether a sample image of a Red Blood Cell provided to the model is parasitized or unhealthy. This model has a primary goal of detecting malaria in Red Blood Cells from blood smears with the least number of losses. This allows for the most minimal number of malaria-infected cell to be mistakenly passed off as healthy cells. There will be a further comparison between the custom CNN model, VGG19 model with no fine-tuning, VGG19 model fine-tuned, and a ResNet50 model. All of these are models which have been pre-trained on a vast number of images previously with a set of weights termed as Imagenet.