A comparative analysis of deep learning architectures on high variation malaria parasite classification dataset. (April 2021)

• Record Type: Journal Article
• Authors: Rahman, Aimon; Zunair, Hasib; Reme, Tamanna Rahman; Rahman, M. Sohel; Mahdy, M.R.C.
• Is Part Of: Tissue & cell. Volume 69(2021)
• Publication Date: 2021-04
• Subjects: Adversarial training -- Transfer learning -- Microscopy data -- Malaria detection; Cytology -- Periodicals; 571.5
• DOI: 10.1016/j.tice.2020.101473
• Languages: English
• ISSNs: 0040-8166

Abstract

Highlights: Proposed malaria dataset exhibits higher variation in uninfected class than the existing public dataset. Transfer learning on medical images outperforms transfer learning on the natural image domain. Conditional image synthesis can address the problem of malaria data imbalance. Training on a high variation dataset yields better performance on data from a different domain. Abstract: Malaria, one of the leading causes of death in underdeveloped countries, is primarily diagnosed using microscopy. Computer-aided diagnosis of malaria is a challenging task owing to the fine-grained variability in the appearance of some uninfected and infected class. In this paper, we transform a malaria parasite object detection dataset into a classification dataset, making it the largest malaria classification dataset (63, 645 cells), and evaluate the performance of several state-of-the-art deep neural network architectures pretrained on both natural and medical images on this new dataset. We provide detailed insights into the variation of the dataset and qualitative analysis of the results produced by the best models. We also evaluate the models using an independent test set to demonstrate the model's ability to generalize in different domains. Finally, we demonstrate the effect of conditional image synthesis on malaria parasite detection. We provide detailed insights into the influence of synthetic images for the class imbalance problem in the malaria diagnosis context.