Abstract
We present the development of SpeCamX, a mobile application that transforms any unmodified smartphone into a powerful multispectral imager capable of capturing multispectral information. Our application includes an augmented bilirubinometer, enabling accurate prediction of blood bilirubin levels (BBL). In a clinical study involving 320 patients with liver diseases, we used SpeCamX to image the bulbar conjunctiva region, and we employed a hybrid machine learning prediction model to predict BBL. We observed a high correlation with blood test results, demonstrating the efficacy of our approach. Furthermore, we compared our method, which uses spectrally augmented learning (SAL), with traditional learning based on RGB photographs (RGBL), and our results clearly indicate that SpeCamX outperforms RGBL in terms of prediction accuracy, efficiency, and stability. This study highlights the potential of SpeCamX to improve the prediction of bio-chromophores, and its ability to transform an ordinary smartphone into a powerful medical tool without the need for additional investments or expertise.
Introduction
Monitoring liver health through blood bilirubin levels (BBL) is a critical diagnostic requirement, yet standard clinical blood tests remain invasive and infrastructure-dependent. While non-invasive bilirubinometers exist, they typically require expensive, specialized hardware. Previous attempts to leverage smartphone cameras for this purpose have been hindered by the physical limitations of the Bayer filter, where overlapping spectral sensitivities in standard RGB sensors result in low spectral resolution. This makes it difficult to isolate the specific absorption signatures of bilirubin, which peaks between 350–500 nm. SpeCamX addresses this by utilizing a software-only approach to reconstruct high-resolution multispectral data from standard RGB snapshots.
Methods
SpeCamX implements a sophisticated pipeline to transform an unmodified smartphone into a 27-channel multispectral imager. The process begins with the capture of a standard RGB image of the bulbar conjunctiva (the white part of the eye) using the phone's built-in flash in a controlled dark environment.
- Spectral Reconstruction: Using a Wiener estimation algorithm and pre-stored Transformation Matrices (TMs) calibrated for specific phone models, the app "unmixes" the RGB signal into a 27-channel multispectral data cube ranging from 420 to 680 nm.
- Spectrally Augmented Learning (SAL): Rather than training on raw RGB values, the system uses the reconstructed 27-channel spectra as input features. This allows the model to focus on physically relevant absorption bands.
- Hybrid Prediction Model: To ensure stability and reduce bias, the system employs a "committee" of four machine learning algorithms: Artificial Neural Networks (ANN), Support Vector Machines (SVM), k-Nearest Neighbors (KNN), and Random Forest (RF). Their outputs are linearly combined to produce the final BBL prediction.
Results and Discussion
The system was validated in a clinical trial involving 320 patients with various liver diseases. The ground truth was established using standard clinical diazo-method blood tests.
The hybrid SAL model achieved a Pearson correlation coefficient (R) greater than 0.90 and a mean bias of only 1.23 μmol/L. The Area Under the Receiver Operating Characteristic curve (AUROC) was 0.97, demonstrating exceptional diagnostic power for identifying elevated bilirubin.
A key finding was the efficiency of SAL over traditional RGB learning (RGBL). The SAL-based models reached high accuracy significantly faster; even with only 12.5% of the training data (n=40), the SAL model maintained an R of 0.9, whereas the RGBL model's performance degraded sharply. This suggests that the multispectral reconstruction process effectively "guides" the machine learning model toward the relevant physical features of bilirubin absorption, making the system more robust and data-efficient.
Conclusion
SpeCamX demonstrates that high-fidelity multispectral imaging can be achieved on unmodified consumer hardware through clever mathematical modeling. By transforming a smartphone into a clinical-grade bilirubinometer, this technology has the potential to democratize liver health monitoring, particularly in resource-limited settings. Future work will focus on expanding the platform to monitor other bio-chromophores such as hemoglobin and melanin, and validating the system across more diverse ethnic populations.
Summary updated