Automated Malaria Detection and staging by blood image using Deep Learning

Abstract

Malaria is a life-threatening yet preventable and treatable disease that continues to pose a major public health challenge, especially in tropical and developing regions. According to Organization, millions of cases are reported every year, leading to significant morbidity and mortality. Early and accurate detection of malaria is crucial for effective treatment and prevention of severe complications. This project aims to develop a computer vision–based diagnostic system for automated malaria detection and staging using blood smear images. The system utilizes deep learning techniques to identify infected red blood cells and classify the stage of infection. The dataset consists of blood smear images, where approximately 80% of the data is used for training and 20% for testing to ensure reliable model performance. Firstly, a Convolutional Neural Network (CNN) model is developed using a customized architecture rather than pre-trained models. The CNN includes layers such as Conv2D, pooling, batch normalization, dropout, flatten, and fully connected dense layers to extract and learn important features from blood images. Secondly, advanced techniques and model tuning are applied to improve classification accuracy and robustness. The system classifies images into different categories such as infected and uninfected cells, and further stages of infection if applicable. The model achieves high accuracy in the range of 0.95 to 0.98, demonstrating its effectiveness in detecting malaria from microscopic images. Performance evaluation is conducted using metrics such as confusion matrix, accuracy, precision, and recall. Finally, the system is integrated with a user-friendly interface using Flask, allowing users to upload images and receive real-time predictions