Malaria Disease Diagnosis from a Blood Smear Samples using the Deep Learning MobileNet Models
Sammy V. Militante, Renante A. Diamante · 2021
Malaria is caused by a bite of a female Malaria mosquito known as Anopheles mosquitoes and is life-threatening to human lives. Infected mosquitoes carry parasites and transmit them to a human person. An estimated 229 million reported cases of malaria-affected worldwide, and it claims more than 400 thousand lives and mostly children below five years of age, according to the report of the World Health Organization. Early prevention and treatment of malaria can cut disease spread and avoid deaths. The common techniques implemented by the medical authorities are either microscopy or rapid diagnostic test. Deep learning is a popular method used in solving classification problems. One such method is the Convolutional Neural Networks (CNN) that captures raw images in the pixel that learns to extract features from learned images and classifies the input image. In this study, the researchers were able to implement three versions MobileNet models of CNN. A total of 27,558 malaria parasitized and uninfected images were used. These datasets were divided into trained data of 22,046 images, test data of 4,134 images, and validation data of 1,378 images. The trained models generated an accuracy rate between 94% to 96.5%, having the MobileNetV3 with the highest accuracy rate earned of 96.5%. In contrast, the lowest recorded accuracy rate was the MobileNetV2 with a rate of The models were able to predict malaria parasitized and malaria uninfected images.