Malaria Cell Classification Using CNN: A Deep Learning Approach
Muhammad Qasim Ali, Rand Kouatly, Meerah Karunanithi, Ahmed Hassan, Syed Arslan Abbas Rizvi, Nafeesath Parappurath Puthiyapurayil · 2024
Plasmodium malaria is a parasitic protozoan that causes malaria in humans. Computer-aided detection of Plasmodium is a research area attracting significant interest. In the current paper, the results of different machine learning, neural network, and deep learning approaches are speculated to evaluate their performance on cell images from digital microscopy concerning the identification of Plasmodium. A publicly available dataset containing 27,558 individual images is utilized, with half of these images containing Plasmodium and the remaining consisting of uninfected cells. The dataset is randomly split into 80-20% groups for training and testing. Color normalization is performed, and all images are spatially reshaped to a predefined size that matches the chosen architecture configurations. A fast CNN model is developed to classify cells in specific cell images. Additionally, classification and ranking of algorithms for transfer learning are conducted using existing networks such as AlexNet, ResNet, VGG-16, and DenseNet. Alongside the evaluation of the bag-of-features model performance, a Support Vector Machine is also employed for classification. A generalized probability trait of an image is assigned only after thoroughly evaluating the probability values given by each CNN model developed for this purpose in this paper. The algorithm achieves an overall accuracy of 96.7% on the test dataset.