Abstract
Neonatal jaundice is a prevalent and potentially serious condition that can lead to severe complications if undiagnosed or untreated. While traditional diagnostic methods like blood sampling are invasive and time-consuming, and transcutaneous bilirubinometers remain costly, smartphone-based image analysis offers a promising low-cost, non-invasive alternative. However, most existing solutions rely on traditional machine learning techniques with limited accuracy and generalizability. In this study, we introduce a deep learning approach based on the Vision Transformer (T2T-ViT) and compare its performance with three other models, ResNet, Support Vector Machine (SVM), and K-Nearest Neighbors (k-NN), using a clinically annotated dataset of neonatal skin images captured via a smartphone camera. The models were evaluated using multiple performance metrics including accuracy, precision, recall, F1-score, Matthews Correlation Coefficient (MCC), and Area under the Curve (AUC). The T2T-ViT model achieved 99% across all metrics, significantly outperforming both convolutional and traditional machine learning models. These findings demonstrate the feasibility of applying transformer-based deep learning architectures for accessible, scalable, and accurate non-invasive neonatal jaundice screening, potentially enabling early intervention in resource-limited settings. This approach could serve as an accessible, scalable screening tool for neonatal jaundice detection, particularly in low-resource clinical settings.
Data availability
The dataset generated and/or analyzed during the current study are not publicly available due to patient privacy concerns and institutional ethical regulations. However, de-identified datasets and additional materials can be made available from the corresponding author upon reasonable request and subject to ethical approval.
References
Brown, A. K. Kernicterus: past, present, and future. Neoreviews 4 (2), e33–e40 (2003).
Brits, H. et al. The prevalence of neonatal jaundice and risk factors in healthy term neonates at National district hospital in Bloemfontein. Afr. J. Prim. Health Care Family Med. 10 (1), 1–6 (2018).
Bhutani, V. K. et al. Neonatal hyperbilirubinemia and rhesus disease of the newborn: incidence and impairment estimates for 2010 at regional and global levels. Pediatr. Res. 74 (1), 86–100 (2013).
Falke, M. The basics of neonatal hyperbilirubinemia. Neonatal Netw., 44(1). (2025).
Wickremasinghe, A. C. & Kuzniewicz, M. W. Neonatal hyperbilirubinemia. Pediatr. Clin. North. Am. 72 (4), 605–622 (2025).
Thomas, M. et al. Current and emerging technologies for the timely screening and diagnosis of neonatal jaundice. Crit. Rev. Clin. Lab. Sci. 59 (5), 332–352 (2022).
Ho, E. et al. BiliCheck transcutaneous bilirubinometer: a screening tool for neonatal jaundice in the Chinese population. Hong Kong Med. J. 12 (2), 99 (2006).
Boo, N. Y. & Ishak, S. Prediction of severe hyperbilirubinaemia using the Bilicheck transcutaneous bilirubinometer. J. Paediatr. Child Health. 43 (4), 297–302 (2007).
Hemmati, F. & Rad, N. A. K. The value of Bilicheck® as a screening tool for neonatal jaundice in the South of Iran. Iran. J. Med. Sci. 38 (2), 122 (2013).
Engle, W. D. et al. Evaluation of a transcutaneous jaundice meter following hospital discharge in term and near-term neonates. J. Perinatol. 25 (7), 486–490 (2005).
De Greef, L. et al. Bilicam: using mobile phones to monitor newborn jaundice. In: Proceedings of the 2014 ACM International Joint Conference on Pervasive and Ubiquitous Computing. (2014).
Taylor, J. A. et al. Use of a smartphone app to assess neonatal jaundice. Pediatrics, 140(3). (2017).
Swarna, S. et al. The smart phone study: assessing the reliability and accuracy of neonatal jaundice measurement using smart phone application. Int. J. Contemp. Pediatr., 5(285). (2018).
Mariakakis, A. et al. Biliscreen: smartphone-based scleral jaundice monitoring for liver and pancreatic disorders. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 1 (2), 1–26 (2017).
Rizvi, M. R. et al. A novel non-invasive technique of measuring bilirubin levels using bilicapture. Oman Med. J. 34 (1), 26 (2019).
Gupta, A., Kumar, A. & Khera, P. Jaundice prediction through non-invasive techniques: Issues and challenges. in Annual IEEE India Conference (INDICON) (IEEE, 2015).
Hashim, W. et al. Computer vision for jaundice detection in neonates using graphic user interface. In: IOP Conference Series: Materials Science and Engineering. (IOP Publishing, 2021).
Mansor, M. et al. New newborn jaundice monitoring scheme based on combination of pre-processing and color detection method. Neurocomputing 120, 258–261 (2013).
Munkholm, S. B. et al. The smartphone camera as a potential method for transcutaneous bilirubin measurement. PloS One. 13 (6), e0197938 (2018).
Saini, N., Kumar, A. & Khera, P. Non-invasive bilirubin detection technique for jaundice prediction using smartphones. Int. J. Comput. Sci. Inform. Secur. 14 (8), 1060 (2016).
Padidar, P. et al. Detection of neonatal jaundice by using an android OS-based smartphone application. Iran. J. Pediatr., 29(2). (2019).
Aydın, M. et al. Neonatal jaundice detection system. J. Med. Syst. 40, 1–11 (2016).
Hardalac, F. et al. Classification of neonatal jaundice in mobile application with noninvasive imageprocessing methods. Turkish J. Electr. Eng. Comput. Sci. 29 (4), 2116–2126 (2021).
Castro-Ramos, J. et al. Detecting jaundice by using digital image processing. in Three-Dimensional and Multidimensional Microscopy: Image Acquisition and Processing XXI. SPIE. (2014).
Leung, T. S. et al. Screening neonatal jaundice based on the sclera color of the eye using digital photography. Biomedical Opt. Express. 6 (11), 4529–4538 (2015).
Outlaw, F. et al. Screening for neonatal jaundice with a smartphone. In: Proceedings of the International Conference on Digital Health. (2017).
Laddi, A. et al. Non-invasive jaundice detection using machine vision. IETE J. Res. 59 (5), 591–596 (2013).
Outlaw, F. et al. Smartphone screening for neonatal jaundice via ambient-subtracted sclera chromaticity. PloS One. 15 (3), e0216970 (2020).
Parinyanut, P. et al. Digital camera image analysis of faeces in detection of cholestatic jaundice in infants. Afr. J. Pediatr. Surg. 13 (3), 131–135 (2016).
Gomez, M. et al. A graphical decision-theoretic model for neonatal jaundice. Med. Decis. Making. 27 (3), 250–265 (2007).
Hyperbilirubinemia, A. A. .o.P.S.o., Management of hyperbilirubinemia in the newborn infant 35 or more weeks of gestation. Pediatrics 114 (1), 297–316 (2004).
Moyer, V. A., Ahn, C. & Sneed, S. Accuracy of clinical judgment in neonatal jaundice. Arch. Pediatr. Adolesc. Med. 154 (4), 391–394 (2000).
De Luca, D. et al. The joint use of human and electronic eye: visual assessment of jaundice and transcutaneous bilirubinometry. Turk. J. Pediatr. 50 (5), 456–461 (2008).
Riskin, A., Abend-Weinger, M. & Bader, D. How accurate are neonatologists in identifying clinical jaundice in newborns? Clin. Pediatr. 42 (2), 153–158 (2003).
Nayagi, S. B. & Angel, T. S. Diagnosis of neonatal hyperbilirubinemia using CNN model along with color card techniques. Biomed. Signal Process. Control. 85, 104746 (2023).
Makhloughi, F. Artificial intelligence-based non-invasive bilirubin prediction for neonatal jaundice using 1D convolutional neural network. Sci. Rep. 15 (1), 11571 (2025).
Park, J. H. et al. Deep-learning-based smartphone application for self-diagnosis of scleral jaundice in patients with hepatobiliary and pancreatic diseases. J. Personalized Med. 11 (9), 928 (2021).
Althnian, A., Almanea, N. & Aloboud, N. Neonatal jaundice diagnosis using a smartphone camera based on eye, skin, and fused features with transfer learning. Sensors 21 (21), 7038 (2021).
Association, W. M. World medical association declaration of helsinki: ethical principles for medical research involving human subjects. Jama 310 (20), 2191–2194 (2013).
Yuan, L. et al. Tokens-to-token vit: Training vision transformers from scratch on imagenet. In: Proceedings of the IEEE/CVF international conference on computer vision. (2021).
Cortes, C. Support-Vector Networks. Machine Learning, (1995).
Harirchian, E. et al. A machine learning framework for assessing seismic hazard safety of reinforced concrete buildings. Appl. Sci. 10 (20), 7153 (2020).
Guo, G. et al. KNN model-based approach in classification. In: On The Move to Meaningful Internet Systems 2003: CoopIS, DOA, and ODBASE: OTM Confederated International Conferences, CoopIS, DOA, and ODBASE 2003, Catania, Sicily, Italy, November 3–7, 2003. Proceedings. (Springer, 2003).
Hasan, M. J. & Kim, J. M. A hybrid feature pool-based emotional stress state detection algorithm using EEG signals. Brain Sci. 9 (12), 376 (2019).
Madhusundar, N. & Surendran, R. Neonatal jaundice identification over the face and sclera using graph neural networks. In: 5th International Conference on Smart Systems and Inventive Technology (ICSSIT). (IEEE, 2023).
Hsu, W. Y. & Cheng, H. C. A fast and effective system for detection of neonatal jaundice with a dynamic threshold white balance algorithm. In: Healthcare. MDPI. (2021).
Abdulrhman, A. I., Jarjees, M. S. & Kasim, N. A. Neonatal jaundice detection system using convolutional neural network algorithm. Neuro Quantol., 20(8): 7536–7546. (2022).
Khanam, F. T. Z. et al. Non-invasive and non-contact automatic jaundice detection of infants based on random forest. Comput. Methods Biomech. Biomedical Engineering: Imaging Visualization. 11 (6), 2516–2529 (2023).
Abdulrazzak, A. Y. et al. Real-time jaundice detection in neonates based on machine learning models. BioMedInformatics 4 (1), 623–637 (2024).
Acknowledgements
The authors acknowledge the use of ChatGPT (OpenAI) solely for assistance with language polishing and minor grammatical corrections during manuscript preparation. All scientific content, study design, analysis, and conclusions were entirely conceived and written by the authors. The authors would also like to thank Dr. Alireza Vafaei Sadr for his valuable contributions during the early stages of this work.The authors would like to express their sincere gratitude to the Children’s Medical Center and the Tehran University of Medical Sciences for their continued support and cooperation throughout the course of this research. We are particularly thankful to the hospital’s neonatal care and laboratory staff for facilitating data acquisition and image collection. Special thanks are extended to the parents and legal guardians of the participating neonates for their kind consent and trust in this research project.
Funding
This study was conducted without the support of any external funding agency. The research was carried out independently as part of an academic initiative under the auspices of Amir Kabir and Tehran University of Medical Sciences.
Author information
Authors and Affiliations
Contributions
M.L. and M.R. conceptualized the study and contributed to manuscript preparation.M.L. drafted the initial version of the manuscript.MS.A. designed and implemented the deep learning model based on the Vision Transformer architecture, conducted the experiments, and managed the overall project.R.S. provided clinical supervision and contributed to data collection and validation.N.A. assisted with clinical data collection and neonatal skin photography.M.H.N. contributed to data annotation and literature review.All authors reviewed and approved the final version of the manuscript.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Ethics statement
This study was approved by the Institutional Review Board of the Children’s Medical Center, Tehran University of Medical Sciences (IR.TUMS.CHMC.REC.1399.001), and conducted in full compliance with the Declaration of Helsinki. Written informed consent was obtained from the legal guardians of all participating neonates prior to data collection. All data were anonymized to ensure confidentiality. In addition, consent was obtained for the publication of anonymized images in an open-access scientific journal.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, 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 changes were made. 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/4.0/.
About this article
Cite this article
Lotfi, M., Rabiee, M., Nazarpak, M.H. et al. Neonatal jaundice detection using a vision transformer-based deep learning model. Sci Rep (2026). https://doi.org/10.1038/s41598-026-40515-5
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-026-40515-5
