Abstract
Invasive venous blood draws remain the clinical standard for hematology, yet they are invasive, time-consuming, and costly. We introduce Video-to-Vessels, a computer-vision pipeline that converts high-magnification videos of bulbar conjunctiva capillaries into low-dimensional spatiotemporal vessel representations, reducing video dimensionality by ~200-fold while preserving hemodynamic information. These representations feed VesselNet, a multi-instance regression network that encodes each vessel with a modified ConvNeXt backbone, fuses vessel-specific thickness via cross-attention, and predicts blood biomarkers from concatenated embeddings. On a cohort of 224 participants with paired laboratory counts, VesselNet achieves a hemoglobin-based anemia ROC-AUC of 82.8% and a Spearman’s ρ of 0.47, while attaining a ρ of 0.46 for red-blood-cell (RBC) count regression. Removing local stabilization and segmentation-denoising lowers ρ by 38% for hemoglobin and 19% for RBC, underscoring their contributions. Our results mark a step toward a fully noninvasive complete blood count, coupling representation learning with ocular imaging.
Data availability
The anonymized bulbar conjunctiva vessel dataset, including RGB values extracted from high-magnification videos, and the associated blood test results, have been deposited in Figshare42. All personal identifying information has been removed to protect patient privacy. Additional data that support the findings of this study are available from the corresponding author upon reasonable request, subject to any necessary institutional or ethical approvals. No custom software was used beyond standard libraries.
Code availability
The custom code used for training the vessels dataset and testing the model has been deposited on GitHub and can be accessed via this link: https://github.com/tamirdennis/VesselNet.git. This code will be provided under the MIT License, an Open Source Initiative (OSI)-approved license, upon publication. Any subsequent updates to the code are tracked and versioned within GitHub to facilitate reproducibility and transparency.
References
Anemia. https://www.who.int/news-room/fact-sheets/detail/anaemia.
Safiri, S. et al. Burden of anemia and its underlying causes in 204 countries and territories, 1990–2019: results from the Global Burden of Disease Study 2019. J. Hematol. Oncol. 14, 185 (2021).
Kamel, M. M. et al. Evaluation of noninvasive hemoglobin monitoring in children with congenital heart diseases. Paediatr. Anaesth. 30, 571–576 (2020).
von Schweinitz, B. A., De Lorenzo, R. A., Cuenca, P. J., Anschutz, R. L. & Allen, P. B. Does a non-invasive hemoglobin monitor correlate with a venous blood sample in the acutely ill? Intern. Emerg. Med. 10, 55–61 (2015).
Nguyen, B.-V. et al. The accuracy of noninvasive hemoglobin measurement by multiwavelength pulse oximetry after cardiac surgery. Anesth. Analg. 113, 1052–1057 (2011).
Bell, S. et al. Comparison of four methods to measure haemoglobin concentrations in whole blood donors (<scp>COMPARE</scp>): a diagnostic accuracy study. Transfus. Med. 31, 94–103 (2021).
Kim, M. J., Park, Q., Kim, M. H., Shin, J. W. & Kim, H. O. Comparison of the accuracy of noninvasive hemoglobin sensor (NBM-200) and portable hemoglobinometer (HemoCue) with an automated hematology analyzer (LH500) in blood donor screening. Ann. Lab. Med. 33, 261–267 (2013).
Hadar, E., Raban, O., Bouganim, T., Tenenbaum-Gavish, K. & Hod, M. Precision and accuracy of noninvasive hemoglobin measurements during pregnancy. J. Matern. Neonatal. Med. 25, 2503–2506 (2012).
Amrutha, A. M. et al. Estimation of haemoglobin using non-invasive portable device with spectroscopic signal application. Sci. Rep. 14, 8697 (2024).
Das Mahapatra, P. P., Roy, C., Agarwal, K., Banerjee, J. & Sharma, S. Performance of the non-invasive point-of-care device, EzeCheck, for haemoglobin assessment in adults and children in community and institutional care settings. PLoS Digit. Health 3, e0000500 (2024).
Mitani, A. et al. Detection of anaemia from retinal fundus images via deep learning. Nat. Biomed. Eng. 4, 18–27 (2019).
Zhao, X. et al. Deep-learning-based hemoglobin concentration prediction and anemia screening using ultra-wide field fundus images. Front. Cell Dev. Biol. 10, 888268 (2022).
Das, S. et al. NiADA (Non-invasive Anemia Detection App), a smartphone-based application with artificial intelligence to measure blood hemoglobin in real-time: a clinical validation. Cureus https://doi.org/10.7759/cureus.65442 (2024).
Kato, S. et al. Machine/deep learning-assisted hemoglobin level prediction using palpebral conjunctival images. Br. J. Haematol. https://doi.org/10.1111/bjh.19621 (2024).
Zhao, L. et al. Prediction of anemia in real-time using a smartphone camera processing conjunctival images. PLoS ONE 19, e0302883 (2024).
Suner, S. et al. Prediction of anemia and estimation of hemoglobin concentration using a smartphone camera. PLoS ONE 16, e0253495 (2021).
Dimauro, G., Camporeale, M. G., Dipalma, A., Guarini, A. & Maglietta, R. Anaemia detection based on sclera and blood vessel colour estimation. Biomed. Signal Process. Control 81, 104489 (2023).
He, J. et al. Non-invasive quantitative blood cell counting using optical coherence tomography. Opt. Laser Technol. 179, 111313 (2024).
Ni, J., Li, G., Tang, W., Xiao, Q. & Lin, L. Noninvasive human red blood cell counting based on dynamic spectrum. Infrared Phys. Technol. 113, 103604 (2021).
Winer, M. M. et al. In vivo noninvasive microscopy of human leucocytes. Sci. Rep. 7, 13031 (2017).
Hu, D.-N. et al. Quantitative study of human scleral melanocytes and their topographical distribution. Curr. Eye Res. 45, 1563–1571 (2020).
Sun, J., Shen, Z., Wang, Y., Bao, H. & Zhou, X. LoFTR: Detector-free local feature matching with transformers. in Proc. IEEE Computer Society Conference on Computer Vision and Pattern Recognition. https://doi.org/10.1109/CVPR46437.2021.00881 (2021).
Liu, W. et al. Full-resolution network and dual-threshold iteration for retinal vessel and coronary angiograph segmentation. IEEE J. Biomed. Heal. Informatics 26, 4623–4634 (2022).
Ebrahimi, S., Bedggood, P., Ding, Y., Metha, A. & Bagchi, P. A high-fidelity computational model for predicting blood cell trafficking and 3D capillary hemodynamics in retinal microvascular networks. Investig. Ophthalmol. Vis. Sci. 65, 37 (2024).
Sun, Z. et al. Progress of bulbar conjunctival microcirculation alterations in the diagnosis of ocular diseases. Dis. Markers 2022, 4046809 (2022).
Xu, Z. et al. Measurement variability of the bulbar conjunctival microvasculature in healthy subjects using functional slit lamp biomicroscopy (FSLB). Microvasc. Res. 101, 15–19 (2015).
Brennan, P. F. et al. Assessment of the conjunctival microcirculation for patients presenting with acute myocardial infarction compared to healthy controls. Sci. Rep. 11, 7660 (2021).
Israel Central Bureau of Statistics. Statistical Abstract of Israel, Media Release. 371, https://www.cbs.gov.il/he/mediarelease/DocLib/2024/371/11_24_371b.pdf (2024).
Patel, K. V. et al. Haemoglobin concentration and the risk of death in older adults: differences by race/ethnicity in the NHANES III follow-up. Br. J. Haematol. 145, 514–523 (2009).
Dallman, P. R., Barr, G. D., Allen, C. M. & Shinefield, H. R. Hemoglobin concentration in white, black, and Oriental children: is there a need for separate criteria in screening for anemia? Am. J. Clin. Nutr. 31, 377–380 (1978).
Cresanta, J. L., Croft, J. B., Webber, L. S., Nicklas, T. A. & Berenson, G. S. Racial difference in hemoglobin concentration of young adults. Prev. Med. 16, 659–669 (1987).
Liu, Z. et al. A ConvNet for the 2020s. In Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 11966–11976. https://doi.org/10.1109/CVPR52688.2022.01167 (2022).
Gardner, W. M. et al. Prevalence, years lived with disability, and trends in anaemia burden by severity and cause, 1990–2021: findings from the Global Burden of Disease Study 2021. Lancet Haematol. 10, e713–e734 (2023).
Hornedo-González, K. D. et al. Non-invasive hemoglobin estimation for preoperative anemia screening. Transfusion 63, 315–322 (2023).
Cohen, J. F. et al. STARD 2015 guidelines for reporting diagnostic accuracy studies: explanation and elaboration. BMJ Open 6, e012799 (2016).
Karndumri, K. et al. Comparison of hemoglobin and hematocrit levels at 1, 4 and 24 h after red blood cell transfusion. Transfus. Apher. Sci. 59, 102586 (2020).
Chantepie, S. P. et al. Effect of single-unit transfusion in patients treated for haematological disease including acute leukemia: a multicenter randomized controlled clinical trial. Leuk. Res. 129, 107058 (2023).
Obama, K., Nakabeppu, S. & Inoue, H. Red blood cell deformation and progressive anemia following therapeutic intervention in patients with adult T-cell leukemia/lymphoma. Cureus 15, e34641 (2023).
Beutler, E. & Waalen, J. The definition of anemia: what is the lower limit of normal of the blood hemoglobin concentration? Blood 107, 1747–1750 (2006).
Badireddy M, B. K. Chronic Anemia. in StatPearls [Internet] (StatPearls Publishing, 2024).
NHANES. NHANES Questionnaires, Datasets, and Related Documentation. https://wwwn.cdc.gov/nchs/nhanes/.
Denis, T. Anonymized bulbar conjunctiva vessel data sets. figshare https://doi.org/10.6084/m9.figshare.28280930 (2026).
Acknowledgements
This study was supported by the Israeli Ministry of Innovation, Science and Technology (grant number: 3-17994, to HS and YR). IS is partially funded by The Nehemia Rubin Excellence in Biomedical Research, TELEM Program, Sheba Medical Center, Tel Hashomer, Israel.
Author information
Authors and Affiliations
Contributions
T.D., I.S., H.S., and Y.R. conceived and designed the research. Y.R., E.P., A.Z., and M.H. collected the data. T.D., L.W., and H.S. designed and developed the algorithms. T.D., L.W., H.S., I.S., O.B., A.A., K.A., R.D., and Y.R. participated in the data analysis. T.D., I.S., H.S., and Y.R. wrote the main manuscript text. T.D. and H.S. prepared the figures. All authors reviewed the manuscript.
Corresponding author
Ethics declarations
Competing interests
T.D., I.S., L.W., H.S., and Y.R. have patent applications related to the subject matter, vis Tel Aviv University. Patent applicant—institution—Ramot at Tel-Aviv University Ltd. (Tel-Aviv), Tel HaShomer Medical Research Infrastructure and Services Ltd. (Ramat-Gan). All other authors declare no competing interests.US Patent application 63/138,546 by Ramot at Tel-Aviv University Ltd. (Tel-Aviv), Tel HaShomer Medical Research Infrastructure and Services Ltd. (Ramat-Gan), Inventors: Ygal Rotenstreich, Ifat Sher Rosenthal, Haim Suchowski, Michael Mrejen, and Shahar Katz. Status: Application. Aspect of the manuscript covered in the patent application: the approach for imaging of bulbar blood vessels. U.S Provisional Application No. 63/829,813 by Ramot at Tel-Aviv University Ltd. (Tel-Aviv), Tel HaShomer Medical Research Infrastructure and Services Ltd. (Ramat-Gan), Inventors: Ygal Rotenstreich, Ifat Sher, Tamir Denis, Haim Suchowski and Lior Wolf. Status: provisional application. Aspect of manuscript covered in the patent application: the Video-to-Vessels pipeline and VesselNet.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Denis, T., Sher, I., Praisman, E. et al. Towards noninvasive blood count using a deep learning pipeline from bulbar conjunctiva videos. npj Digit. Med. (2026). https://doi.org/10.1038/s41746-026-02598-2
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41746-026-02598-2
