CodaMal: Contrastive Domain Adaptation for Malaria Detection in Low-Cost Microscopes

TL;DR

CodaMal introduces an end-to-end framework for malaria parasite detection that trains on high-cost microscope images and tests on low-cost microscopy images. It adds a Domain Adaptive Contrastive Loss ($L_{DAC}$) to align HCM and LCM representations, enabling robust parasite detection without extra annotations. The method uses a single-stage detector with a non-linear projection head to minimize cross-domain discrepancy, achieving substantial mAP improvements (up to ~16-20% relative) and a notable speedup (about 21x) while reducing parameter count. Evaluated on the M5 dataset, CodaMal sets a new state-of-the-art for HCM-to-LCM malaria detection and demonstrates practical potential for scalable, low-cost diagnostics.

Abstract

Malaria is a major health issue worldwide, and its diagnosis requires scalable solutions that can work effectively with low-cost microscopes (LCM). Deep learning-based methods have shown success in computer-aided diagnosis from microscopic images. However, these methods need annotated images that show cells affected by malaria parasites and their life stages. Annotating images from LCM significantly increases the burden on medical experts compared to annotating images from high-cost microscopes (HCM). For this reason, a practical solution would be trained on HCM images which should generalize well on LCM images during testing. While earlier methods adopted a multi-stage learning process, they did not offer an end-to-end approach. In this work, we present an end-to-end learning framework, named CodaMal (COntrastive Domain Adpation for MALaria). In order to bridge the gap between HCM (training) and LCM (testing), we propose a domain adaptive contrastive loss. It reduces the domain shift by promoting similarity between the representations of HCM and its corresponding LCM image, without imposing an additional annotation burden. In addition, the training objective includes object detection objectives with carefully designed augmentations, ensuring the accurate detection of malaria parasites. On the publicly available large-scale M5-dataset, our proposed method shows a significant improvement of 16% over the state-of-the-art methods in terms of the mean average precision metric (mAP), provides 21x speed improvement during inference and requires only half of the learnable parameters used in prior methods. Our code is publicly available: https://daveishan.github.io/codamal-webpage/.

Figures (3)

• Figure 1: In a practical setup of computer-aided malaria detection: high-cost microscopes (HCM) facilitate ease of annotation for training, whereas low-cost microscopes (LCM) provide an affordable solution for testing in low-resource areas. To address this domain gap, we propose a domain adaptive contrastive loss-based framework, CodaMal.
• Figure 2: Schematic diagram of our end-to-end method for malaria detection. (a) During training, an HCM image ($X^i_h$) and its annotations ($Y^i_h$) are paired with the corresponding unannotated LCM image ($X^i_l$) that is captured from the same field-of-view, instance $i$. The HCM image ($X^i_h$) is utilized to train the backbone ($f$) and its detection head ($h$) using the standard object detection losses ($L_{OD}$). For domain adaptation, both $X^i_h$ and $X^i_l$ are processed through the same backbone ($f$) and non-linear projection head ($g$). The HCM representation, $Z^i_h$, is encouraged to maximize similarity with the corresponding LCM instance ($Z^i_l$) (indicated by the green arrow), while minimizing similarity with the representations of other LCM and HCM image instances ($j$,$k$) i.e. instances captured from different FOV or different blood smear (highlighted by the red arrow). (b) Once trained, the model receives LCM images and predicts the location and life-stage of the malaria parasite as its final output.
• Figure 3: Qualitative Results of our proposed method on LCM test images at 3 magnification scales of the microscope.