Enhancing An Image Blood Staining Malaria Diagnosis Using Convolution Neural Network On Raspberry Pi

Abstract

: Malaria is a common disease in Sub-Saharan Africa, the disease is caused by a class of parasites called protozoan, and it is transmitted by female Anopheles mosquitoes to humans. Plasmodium ovale, plasmodium vivax, plasmodium knowlesi, plasmodium falciparum, and plasmodiummalariae. T he five known plasmodium species that cause malaria in humans. The microscopic diagnosis has always been a gold standard but today, computational tools like deep learning are used in malaria prediction. The deep learning model use images to diagnose infection. The model was trained using the Kaggle dataset with 27,560 images with equal instances of primary images,used to validate primary images from the microscope were annotated using Roboflow. A total of 27 primary images were collected. The model gave accuracy and precision of 85% and Recall of 96% both on the personal computer and Raspberry Pi 4. This research provides a prototype for enhancing malaria diagnosis from images by deploying a deep learning model - a convolution neural network, on a Raspberry Pi. This research has proven the possibility of classifying malaria images as parasitized or unparasitized by deploying a deep-learning model on the Raspberry Pi. This study demonstrates that Raspberry Pi can be utilized for diagnosis and overcome the constraint of requiring high computer hardware specifications to operate a deep learning model. The result obtained 90% accuracy in the detection of parasites in the Red Blood Smear.