Malaria Categorization Using an Autonomous CNN System in Conjunction with Data Augmentation
Poonam Shourie, Vatsala Anand, Rahul Singh Chauhan, Ankur Choudhary, Sheifali Gupta · 2023
A major worldwide health issue continues to be malaria, a potentially fatal infectious illness spread by mosquitoes and caused by parasites called Plasmodium. An early and precise diagnosis is essential for the condition to be treated and controlled effectively. Deep learning (DL) methods, in specific Convolutional Neural Networks (CNNs), have freshly established promising results in a variety of picture identification applications, including the detection of malaria. The development of an automated malaria classification system employing CNNs and data augmentation is the main goal of this study. A sizable number of blood smear pictures, some labeled as “Parasitized” (infested with Plasmodium parasites), are present in the dataset used for training and assessment. In order to increase model generalization and solve the lack of labeled data, data augmentation structures including rotation, flipping, and scaling are used to extract more varied samples from the existing pictures. The findings show that the CNN model, in conjunction with data augmentation, achieves good accuracy and achievement in differentiating among “Parasitized” and “Uninfected” blood smear pictures. The danger of overfitting is decreased by the use of data augmentation, which also aids in the model's capacity to simplify effectively to novel data. By advancing automated malaria detection, this study might help medical professionals in places with a shortage of resources and knowledge. To maximize their beneficial effects on malaria management and control efforts, such models must be carefully integrated into the current healthcare infrastructures. This demands careful consideration of ethical and legal issues.