Dynamic Routing Among Capsule Neural Networks for Malaria Parasite Classification: An Empirical Investigation
Gifsy Madhu · Zenodo (CERN European Organization for Nuclear Research) · 2021
Malaria is a serious parasite disease spread by female Anopheles mosquitos. If the parasite remains untreated for an extended period, it can cause serious damage, and delaying treatment can lead to coma and death. Malaria cannot be detected through a clinical diagnosis, it must be identified or ruled out by a microscopic study of blood smears. Experts examine blood smears under a microscope to help identify the parasites. The manual examination is a time consuming and challenging process. The proposed empirical study works on dynamic routing among capsule neural networks for automated malaria parasite identification and classification in blood cell images. The experimental results confirm that the capsule network model designed in this present work renders the accurate classification of blood cell images as opposed to different deep learning networks. This achievement may be attributed to the utilization of swish activation function and this aspect is missing in other studies. Also, it is worthy to note the present methodology depicts 99.72% accuracy rate in the detection of malaria parasites in thin blood cell images. From the findings of the present study, we conclude that the proposed capsule network methodology is a promising tool for malaria classification. Also, it is interesting to note when it comes to rapid malaria diagnosis, the proposed method outperforms conventional methodologies.