NeoJaundice-AI: Smartphone-Based Neonatal Jaundice Detection Using Dual-Input Deep Learning and Synthetic Augmentation

2026-06-14

AI Summary

This study addresses the critical challenges of limited laboratory access for neonatal jaundice screening in resource-constrained settings—particularly rural India—and the diagnostic difficulties posed by darker skin tones. To this end, the authors present the first end-to-end offline AI system tailored for Indian newborns, leveraging smartphone-captured images of both skin and sclera. The approach employs a dual-branch EfficientNet-B0 architecture that fuses handcrafted YCbCr features with deep features, enhanced by skin-tone-adaptive normalization and synthetic jaundice data augmentation to mitigate data scarcity. After INT8 quantization and ONNX deployment, the model occupies only 8.3 MB and achieves inference times under 3 seconds on standard Android devices, delivering a jaundice classification accuracy of 91.8% (sensitivity: 93.5%) and a mean absolute error of 1.4 mg/dL in serum bilirubin prediction.

Abstract

Neonatal jaundice (hyperbilirubinemia) is one of the most common conditions affecting newborns worldwide, with India alone recording roughly 15 million cases per year. Early detection is critical, yet standard diagnosis requires blood tests that are often impractical in rural clinics where laboratory facilities are limited. This paper presents NeoJaundice-AI, a smartphone-based screening system that uses photographs of a baby's skin and sclera (eye white) to estimate jaundice severity and predict serum bilirubin levels in under three seconds without requiring internet connectivity. The proposed system is built on a dual-branch EfficientNet-B0 architecture that independently processes skin and sclera images. Deep features are fused with handcrafted YCbCr color statistics to jointly perform four-class severity classification and continuous bilirubin regression. A key contribution is a synthetic jaundice generation method that simulates bilirubin-induced yellowing through controlled YCbCr channel modifications on normal neonatal skin images. This approach addresses data scarcity, particularly for severe jaundice cases and darker Indian skin tones (Fitzpatrick Types IV to VI). In addition, a skin-tone normalization module improves prediction consistency across diverse neonatal complexions. Experimental results demonstrate an overall classification accuracy of 91.8 percent, a clinical sensitivity of 93.5 percent, and a bilirubin mean absolute error of 1.4 mg/dL. After INT8 quantization and ONNX conversion, the model size is reduced to 8.3 MB while maintaining inference times below three seconds on standard Android devices. To the best of our knowledge, this is the first India-focused neonatal jaundice AI system that combines multimodal image fusion, skin-tone adaptation, synthetic data augmentation, and fully offline mobile deployment within a single framework.

Problem

• neonatal jaundice
• hyperbilirubinemia
• smartphone-based diagnosis
• data scarcity
• skin tone diversity

Innovation

• dual-input deep learning
• synthetic data augmentation
• skin-tone normalization
• offline mobile deployment
• neonatal jaundice detection