Abstract
Malaria diagnosis in resource-deficient regions increasingly relies on low-cost microscopes (LCMs), which offer an affordable alternative to high-cost microscopes (HCMs). Creating labeled LCM datasets is laborious because of poor image quality and limited field of view, and models trained on annotated HCM images often underperform on LCMs due to domain shift. Existing domain adaptation methods primarily focus on global feature alignment, despite malaria detection being a dense prediction task where local cellular characteristics are fundamental to microscopy-based examination. In this work, we propose MaLDAM, a cell-aware domain adaptation framework for LCM malaria detection that jointly performs global and masked local feature alignment between paired HCM and LCM images. By leveraging precomputed cell-containing patch masks, our method focuses adaptation on diagnostically relevant cellular regions while reducing the influence of background variations. We further apply microscopy-specific augmentations to improve robustness to low-cost imaging artefacts. On the M5-dataset, our method improves mean average precision (mAP) by 8.5 at 1000×, 7.3 at 400×, and 3 at 100× magnification without additional inference latency over the previous state-of-the-art. MaLDAM demonstrates more robust domain-invariant representation learning for practical malaria diagnosis using affordable microscopy devices. Our code is publicly available: https://github.com/TishanSathruwan/MaLDAM.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/HemaRAI_004.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: https://papers.miccai.org/miccai-2026-sat/supp/HemaRAI_004_supp.pdf
Link to Open Review
Open Review Page: https://openreview.net/forum?id=IpY0OWAC0B
BibTex
@InProceedings{FerMud_MaLDAM_MICCAISAT2026,
author = { Fernando, Muditha AND Rathnasekara, Tishan AND Nazar, Saeedha AND Perera, Avishka AND Kaluarachchi, Tharindu},
title = { { MaLDAM: Masked Localized Domain Adaptation for Malaria Detection in Low-Cost Microscopic Images } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17252},
month = {pending},
page = {pending}
}