Insights into AI-Driven malaria diagnosis : A systematic review with implications for Plasmodium knowlesi

Jia Ying, Tiong and Khairunnisa, Hasikin and Romano, Ngui and Paul C. S., Divis and Chu Kiong, Loo and Khin Wee, Lai and Fei Wen, Cheong and Wan Yusoff, Wan Sulaiman (2025) Insights into AI-Driven malaria diagnosis : A systematic review with implications for Plasmodium knowlesi. Acta Tropica, 271 (107842). pp. 1-23. ISSN 0001-706X

PDF Insights into AI-Driven malaria diagnosis - Copy.pdf Download (391kB)

Abstract

Plasmodium knowlesi has emerged as a significant zoonotic malaria threat, particularly in Southeast Asia, where its incidence continues to rise. Timely and accurate diagnosis of its blood stages is critical for effective diagnosis and treatment, as disease severity and transmission dynamics vary across different stages. Microscopic examination is the gold standard for malaria diagnosis; however, it is labour-intensive and requires professional interpretation. This makes it prone to variability and possible misclassification, especially among morphologically identical Plasmodium species. Recent advancements in artificial intelligence (AI)-driven approaches, particularly deep learning, offer significant potential to assist microscopists in automating blood-stage identification, reducing diagnostic variability, and improving efficiency without replacing expert validation. However, the research on AI-based classification of P. knowlesi blood stages remains limited. This systematic review critically evaluates the datasets, preprocessing methods, and deep learning techniques used for Plasmodium blood-stage classification with a specific focus on P. knowlesi. Unlike previous reviews that primarily address species classification, this study provides an in-depth comparative analysis of AI-driven blood-stage identification, emphasizing the effectiveness of convolutional neural networks (CNNs), transfer learning, ensemble learning, and object detection models such as YOLO and Faster R-CNN. Additionally, this review highlights key challenges, including limited annotated datasets, class imbalance, and interpretability concerns that persist. Addressing these gaps through enhanced dataset curation, domain adaptation strategies, and explainable AI approaches will be crucial in advancing AI-driven P. knowlesi diagnostics.

Actions (For repository members only: login required)

View Item