Malaria Diagnostic System using Deep Neural Network (DNN)                                                  

Rosemary Chika Nweze ¹,
Unoh, Inyang Nkanu ²,  
Victoria Ukamaka Maduahonwu³,
Ogah, Sunday Ogbonnaya ⁴,
 Confidence Ifeoma Odoh ⁵

_{^1,3,5 Computer Science Department, State University of Medical & Applied Sciences, Igbo-Eno, Enugu State, Nigeria,}
^{2_{ICT Unit, Ministry of Agriculture, Calabar, Cross River State, Nigeria,}
4}Department of Computer Science & Mathematics, Evangel University, Akaeze, Ebonyi State, Nigeria

Abstract

Malaria is one of the major public health challenges caused by the plasmodium parasite. Various methods being used for the diagnosis of this disease ranges from microscopy which remains the gold standard for laboratory confirmation of malaria parasite in a blood sample. The process is time consuming, labor intensive, requires trained personnel, and depends on the quality of the microscope thus the need for a computer aided approach. An open source dataset was obtained from National Library of Science to train the model. The model was trained on a visual geometry group 16 convolutional neural network pre-trained model. We used deep neural network framework and some deep learning libraries such as keras, tensorflow, matplotlib, seaborn, sklearn and numpy based on python programming language. CRISP-DM methodology and very deep learning feature representation was used in the analysis. We deduced a model whose sensitivity score is higher than that of our predecessor. The model achieved 98% sensitivity score based on metrics validatory library which outperformed 93% achieved by the previous authors. We developed a robust and user friendly desktop application that classifies malaria blood samples as infected or uninfected with percentage accuracy.

Keywords: Malaria, visual geometry group 16 convolutional neural network, deep neural network, sensitivity score

Malaria-Diagnostic-System-using-Deep-Neural-Network-DNN