Comparative Analysis of Deep Learning Models for Malaria Detection
Harshita Dooja Poojary, Sumithra T. V · 2022 IEEE 3rd Global Conference for Advancement in Technology (GCAT) · 2022
The blood morphology of a human provides information on changes in the human body and underlying maladies. Disease like malaria, caused by parasites of the genus Plasmodium transmitted through infected Anopheles mosquitoes is an endemic across large areas of Asia, Africa, Central and South America, Eastern Europe and the South Pacific. Conventional diagnosis of malaria involves microscopic examination of the blood smear of a patient which is stained with the nucleic acid. Recent decades have experienced rapid development of Artificial Intelligence in the healthcare industry due to the influx of large datasets provided by healthcare practitioners and the advancement of powerful computer hardware. In particular, the development of deep learning and machine learning is considered to be effective in computer-aided detection and diagnosis, image analysis, drug discovery and health monitoring. This work analyses and provides a detailed review of the implementation of state-of-the-art Deep learning networks in detecting malaria infections. In the present work, the CNN, transfer learning, CNN-KNN and Vision Transformer networks are being analyzed and evaluated against various metrics like accuracy and precision. The network experiments for malaria detection using deep learning on real-world datasets have proven to show significant results despite the limit on training data.