Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 Jul 19:12:1392269.
doi: 10.3389/fbioe.2024.1392269. eCollection 2024.

Embedded-deep-learning-based sample-to-answer device for on-site malaria diagnosis

Affiliations

Embedded-deep-learning-based sample-to-answer device for on-site malaria diagnosis

Chae Yun Bae et al. Front Bioeng Biotechnol. .

Abstract

Improvements in digital microscopy are critical for the development of a malaria diagnosis method that is accurate at the cellular level and exhibits satisfactory clinical performance. Digital microscopy can be enhanced by improving deep learning algorithms and achieving consistent staining results. In this study, a novel miLab™ device incorporating the solid hydrogel staining method was proposed for consistent blood film preparation, eliminating the use of complex equipment and liquid reagent maintenance. The miLab™ ensures consistent, high-quality, and reproducible blood films across various hematocrits by leveraging deformable staining patches. Embedded-deep-learning-enabled miLab™ was utilized to detect and classify malarial parasites from autofocused images of stained blood cells using an internal optical system. The results of this method were consistent with manual microscopy images. This method not only minimizes human error but also facilitates remote assistance and review by experts through digital image transmission. This method can set a new paradigm for on-site malaria diagnosis. The miLab™ algorithm for malaria detection achieved a total accuracy of 98.86% for infected red blood cell (RBC) classification. Clinical validation performed in Malawi demonstrated an overall percent agreement of 92.21%. Based on these results, miLab™ can become a reliable and efficient tool for decentralized malaria diagnosis.

Keywords: automated staining process; deep-learning algorithms; digital microscopy; malaria diagnosis; microscopy examination.

PubMed Disclaimer

Conflict of interest statement

Authors CYB, YMS, MK, YS, HJL, KWK, HWL, YJK, B-IK, SH, S-HC, BMW, CYL, K-HC were employed by Noul Co. Ltd. H-PB, DKL were a member of Scientific Advisory Board (Technical Consultant) in Noul Co., Ltd. The remaining author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Figures

FIGURE 1
FIGURE 1
Schematic of the embedded deep learning based on-site malaria diagnosis. (A) The miLab™ device not only automates the process (automated blood staining without liquid handling and autofocused digital images) of malaria diagnosis through microscopic analysis but also incorporates deep learning algorithm directly into the device for on-site review. (B) A web-based software allows experts to access the digital images for remotely reviewing the result through the internet. (C) Photograph of the result page in miLab™ for Plasmodium falciparum (P. falciparum) positive patient specimens. Users can review and confirm the results in the miLab™ for sample-to-answer, on-site malaria diagnosis. (D) Photograph of the screen shot of the result page from the same patient specimens on the web-based software, accessing remotely digital images and raw data from miLab™. Other experts can remotely review and confirm the same results from miLab™.
FIGURE 2
FIGURE 2
Characterization of the blood film in the miLab™. (A) Photograph of the prepared blood films from the miLab™ using a patient specimen with low hematocrit and high hematocrit from Malawi. Low hematocrit samples to be read in Zone B instead of Zone A, where high hematocrit samples were read. The miLab™ device automatically detects an appropriate area to observe RBCs in a monolayer. (B) Correlation of average RBC counts per FoV depending on the hematocrit of the clinical specimens (n = 37) was shown with open dots (Zone A) and close dots (Zone B). (C) Schematic of blood staining using three distinct staining patches in the cartridge and pictures of stained blood cells with Plasmodium-infected RBCs (black arrow) from each step of the staining procedure. The scale bars = 10 μm. (D) Comparison of microscopic cell image with the miLab™ blood film acquired from miLab™, 50x olympus microscopy with miLab™ blood film, and 100x microscopy with conventional Giemsa slides. The scale bars = 5 μm.
FIGURE 3
FIGURE 3
System verification of the miLab™ device. (A) Reproducibility of blood smear was represented with the box plot using RBC counts per FoV in seven clinical specimens (n = 20). Average RBC counts per FoV were demonstrated with low, middle, and high hematocrits. The RBC counts of the samples with the low (<30%) and the middle/high hematocrits (>30%) were selected from Zones B and A, respectively. (B) Reproducibility of blood staining was represented with the box plot using the red, green, and blue color value, which was obtained from the stained RBCs in FoVs of clinical specimens. The RGB color values of each RBC was conserved across FoVs (n = 4,000). (C) Cellular level classification performance for Plasmodium (ring, gametocytes) was represented with the ROC curve. The area under the curve (AUC) was 0.999 with 95% confidence interval in the range of 0.9986–0.9994. The confusion matrix was calculated at the optimal point where maximum accuracy was obtained. (D) Correlation of the detection rate of malaria positives (ring, gametocytes) between the deep learning algorithm (test group: miLab™) and naked eyes (control group: Microscopy) with the Pearson’s correlation coefficient (r) of 0.96 (p < 0.0001, n = 3,000).
FIGURE 4
FIGURE 4
Clinical validation of miLab™ in Malawi. (A) Design for a clinical study. A total of 555 clinical specimens were enrolled and subjected to microscopy and analysis by miLab™ for comparison with the reference tests (both local microscopy examination and RDT). Yellow cells indicate samples discordant with the reference test. (B) Agreement of analysis by miLab™ with the reference tests (microscopy and RDT). Based on the concordance of microscopy and RDT, overall percent agreement (OPA), positive percent agreement (PPA), and negative percent agreement (NPA) were 92.21%, 95.15%, and 91.43% respectively. (C) Correlation of the parasitemia level between microscopy and the miLab™ on a logarithmic scale. The Pearson’s correlation coefficient (r) is 0.8259 (95% CI: 0.7518–0.8794).
FIGURE 5
FIGURE 5
Performance of detecting stages and species in the miLab™. The confusion matrix represented stage and species classification using embedded deep learning algorithm with additional patient-level estimation. (p.f gameto; gametocyte, p. v trop.; trophozoite, p. v late; schizonts or gametocyte).

Similar articles

Cited by

References

    1. Bae C. Y., Esmaeili H., Zamin S. A., Seol M. J., Hwang E., Beak S. K., et al. (2023). Quantification of solution-free red blood cell staining by sorption kinetics of Romanowsky stains to agarose gels. Anal. Methods. 15, 5369–5379. 10.1039/D3AY01431B - DOI - PubMed
    1. Beck H.-P. (2022). Digital microscopy and artificial intelligence could profoundly contribute to malaria diagnosis in elimination settings. Front. Artif. Intell. 5, 510483. 10.3389/frai.2022.510483 - DOI - PMC - PubMed
    1. Chin L. K., Li H., Choi J.-H., Iwamoto Y., Oh J., Min J., et al. (2022). Hydrogel stamping for rapid, multiplexed, point-of-care immunostaining of cells and tissues. ACS Appl. Mater. Interfaces 14, 27613–27622. 10.1021/acsami.2c05071 - DOI - PubMed
    1. Choi J.-H., Chin L. K., Woo B. M., Song Y., Seol M. J., Hong Y., et al. (2021). Hydrogel-based stamping technology for solution-free blood cell staining. ACS Appl. Mater. Interfaces 13, 22124–22130. 10.1021/acsami.0c22521 - DOI - PubMed
    1. Das D., Vongpromek R., Assawariyathipat T., Srinamon K., Kennon K., Stepniewska K., et al. (2022). Field evaluation of the diagnostic performance of EasyScan GO: a digital malaria microscopy device based on machine-learning. Malar. J. 21, 122. 10.1186/s12936-022-04146-1 - DOI - PMC - PubMed

LinkOut - more resources