Deep Learning Based Model for Malaria Disease Detection Using Convolution Neural Network
Archana Saini, Kalpna Guleria, Shagun Sharma
Abstract
The parasite plasmodium causes malaria, a blood disease that is spread by the bite of a female Anopheles mosquito. Microscopists frequently compare thick and thin blood smears to determine parasitemia and make medical diagnoses. However, the accuracy of these tests depends on how large the smear is and the level of skill used to identify and count infected & parasitized cells. However, large-scale diagnoses made during such an examination might be of poor quality. Machine learning and deep learning techniques are used in image-analysis-based computer-aided diagnosis methods on microscopic images of the smears to identify the presence of malaria. In this work, convolutional neural networks (CNNs), which is a technique of DL have been used for end-to-end feature extraction and classification with highly scalable and superior results. This model has been trained and tested on a dataset contained from an open-source repository called “Kaggle”. In order to get better disease screening, we evaluate the effectiveness of CNN-based DL models as feature extractors for classifying parasitized and uninfected cells. Furthermore, the model has been tested at a learning rate of 0.00001 and 0.001 resulting in accuracy values of 93.70% and 94.42%, respectively. In addition to the implementation, the model can be further improved by applying augmentation techniques to the collected images for efficient and effective results.