Malaria Disease Recognition through Adaptive Deep Learning Models of Convolutional Neural Network

Abstract

Malaria is a disease from blood generated by the Plasmodium parasites spread through the bite of female Anopheles mosquito. Medical personnel diagnose and checks for malaria infection and compute parasitemia through blood smears. Yet, their accuracy and validity in classifying and detecting malaria infection relies on the quality of the smear. This kind of examination may result in inaccurate and poor validation of result especially in a massive scale of diagnoses. Convolutional Neural Networks (CNN) is a division of Deep Neural Networks that is commonly used to analyzed images through learning of image patterns. Deep learning techniques applied to malaria screening will be useful diagnostic aid. In this study, the author was able to evaluate the performance of several CNN architectures in detecting and classifying malaria infected disease and not infected with the malaria disease. The author experimentally determines the ideal CNN model for extracting features from the cell image data. Results shows that ResNet, GoogleNet and VGGNet models are the best CNN models achieving an accuracy rate ranging from 90% to 96% for malaria disease detection in an adaptive deep learning.