Bridging the Gap in Malaria Diagnostics: An Attention-Centric YOLO Framework with Species-Specific Augmentation for Tiny Parasite Detection in Low-Resource Settings

Ahmed Tahiru Issah, Carine Mukamakuz
Proceedings of The Second AAAI Bridge Program on AI for Medicine and Healthcare, PMLR 317:141-149, 2026.

Abstract

Malaria remains a major health challenge in Africa, with accurate species identification critical for prompt treatment and control, especially in resource-limited settings like Rwanda. This study systematically benchmarks state-of-the-art detection architectures for automated multi-species malaria parasite detection from high-resolution Giemsa-stained blood smear images, addressing the persistent problems of class imbalance and the difficulty of detecting small Plasmodium falciparum ring forms. We evaluate and optimize YOLO-SPAM, YOLO-Para, and YOLOv12 models, applying a novel species-specific augmentation protocol with copy-paste and noise injection. YOLOv12, trained with these protocols, achieves outstanding overall performance (0.878 mAP50, 0.677 mAP50-95) and demonstrates significant improvement in detecting small, clinically relevant P. falciparum parasites over non-attention YOLO methods). Comparative experiments reveal that targeted data augmentation and strategic model selection can overcome significant class imbalance, achieving reliable multi-species differentiation and precise localization even for morphologically subtle classes. Our findings validate the practical promise of advanced object detection models, such as YOLOv12, for malaria diagnosis workflows and support their future deployment in real-world microscopy labs in Rwanda and similar endemic regions.

Cite this Paper

BibTeX
@InProceedings{pmlr-v317-issah26a, title = {Bridging the Gap in Malaria Diagnostics: An Attention-Centric YOLO Framework with Species-Specific Augmentation for Tiny Parasite Detection in Low-Resource Settings}, author = {Issah, Ahmed Tahiru and Mukamakuz, Carine}, booktitle = {Proceedings of The Second AAAI Bridge Program on AI for Medicine and Healthcare}, pages = {141--149}, year = {2026}, editor = {Wu, Junde and Pan, Jiazhen and Zhu, Jiayuan and Luo, Luyang and Li, Yitong and Xu, Min and Jin, Yueming and Rueckert, Daniel}, volume = {317}, series = {Proceedings of Machine Learning Research}, month = {20--21 Jan}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v317/main/assets/issah26a/issah26a.pdf}, url = {https://proceedings.mlr.press/v317/issah26a.html}, abstract = {Malaria remains a major health challenge in Africa, with accurate species identification critical for prompt treatment and control, especially in resource-limited settings like Rwanda. This study systematically benchmarks state-of-the-art detection architectures for automated multi-species malaria parasite detection from high-resolution Giemsa-stained blood smear images, addressing the persistent problems of class imbalance and the difficulty of detecting small Plasmodium falciparum ring forms. We evaluate and optimize YOLO-SPAM, YOLO-Para, and YOLOv12 models, applying a novel species-specific augmentation protocol with copy-paste and noise injection. YOLOv12, trained with these protocols, achieves outstanding overall performance (0.878 mAP50, 0.677 mAP50-95) and demonstrates significant improvement in detecting small, clinically relevant P. falciparum parasites over non-attention YOLO methods). Comparative experiments reveal that targeted data augmentation and strategic model selection can overcome significant class imbalance, achieving reliable multi-species differentiation and precise localization even for morphologically subtle classes. Our findings validate the practical promise of advanced object detection models, such as YOLOv12, for malaria diagnosis workflows and support their future deployment in real-world microscopy labs in Rwanda and similar endemic regions.} }
Endnote
%0 Conference Paper %T Bridging the Gap in Malaria Diagnostics: An Attention-Centric YOLO Framework with Species-Specific Augmentation for Tiny Parasite Detection in Low-Resource Settings %A Ahmed Tahiru Issah %A Carine Mukamakuz %B Proceedings of The Second AAAI Bridge Program on AI for Medicine and Healthcare %C Proceedings of Machine Learning Research %D 2026 %E Junde Wu %E Jiazhen Pan %E Jiayuan Zhu %E Luyang Luo %E Yitong Li %E Min Xu %E Yueming Jin %E Daniel Rueckert %F pmlr-v317-issah26a %I PMLR %P 141--149 %U https://proceedings.mlr.press/v317/issah26a.html %V 317 %X Malaria remains a major health challenge in Africa, with accurate species identification critical for prompt treatment and control, especially in resource-limited settings like Rwanda. This study systematically benchmarks state-of-the-art detection architectures for automated multi-species malaria parasite detection from high-resolution Giemsa-stained blood smear images, addressing the persistent problems of class imbalance and the difficulty of detecting small Plasmodium falciparum ring forms. We evaluate and optimize YOLO-SPAM, YOLO-Para, and YOLOv12 models, applying a novel species-specific augmentation protocol with copy-paste and noise injection. YOLOv12, trained with these protocols, achieves outstanding overall performance (0.878 mAP50, 0.677 mAP50-95) and demonstrates significant improvement in detecting small, clinically relevant P. falciparum parasites over non-attention YOLO methods). Comparative experiments reveal that targeted data augmentation and strategic model selection can overcome significant class imbalance, achieving reliable multi-species differentiation and precise localization even for morphologically subtle classes. Our findings validate the practical promise of advanced object detection models, such as YOLOv12, for malaria diagnosis workflows and support their future deployment in real-world microscopy labs in Rwanda and similar endemic regions.
APA
Issah, A.T. & Mukamakuz, C.. (2026). Bridging the Gap in Malaria Diagnostics: An Attention-Centric YOLO Framework with Species-Specific Augmentation for Tiny Parasite Detection in Low-Resource Settings. Proceedings of The Second AAAI Bridge Program on AI for Medicine and Healthcare, in Proceedings of Machine Learning Research 317:141-149 Available from https://proceedings.mlr.press/v317/issah26a.html.

Related Material