Copyright © 2025 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{186881,
        author = {Zeba Masroor and Syed Waheed Ali and Ahmed Rafi Farooqi and Hasan Ahmed Khan},
        title = {Smart Phone-Based Cross-Population Bilirubin Estimation using Multi-Region Color Fusion with Fairness and Uncertainty Awareness},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {12},
        number = {6},
        pages = {2366-2376},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=186881},
        abstract = {Early detection of jaundice is critical, especially for neonates and adults with liver dysfunction. Traditional laboratory tests for total serum bilirubin (TSB) are often invasive, expensive, and may not be available in low-resource settings. Smartphone-based imaging provides a promising non-invasive, affordable, and portable alternative, but existing solutions suffer from population bias, device calibration issues, and lack of fairness or uncertainty assessments. This study presents a novel approach to estimating bilirubin levels through smartphone-captured skin and scleral images, using a deep learning model that does not require device calibration. The proposed framework leverages colour-based spatial and temporal features using a cross-attention transformer. Additionally, fairness-aware adversarial learning is incorporated to ensure that the model performs equitably across different skin tones and devices. To further improve reliability, heteroscedastic uncertainty regression is employed to quantify model uncertainty. Pilot simulations on 120 synthetic samples resulted in an RMSE of 0.95 mg/dL and a fairness gap within ±4%, demonstrating the model’s robustness and fairness. This approach signifies a step toward affordable, accurate, and ethically sound jaundice detection using everyday smartphones, enhancing accessibility in diverse populations and resource-limited settings.},
        keywords = {Jaundice detection, smartphone imaging, bilirubin estimation, scleral colour, skin reflectance, deep learning, fairness, uncertainty.},
        month = {November},
        }