HueDx, Inc., Philadelphia, Pennsylvania, United States of America.
PLoS One. 2024 Oct 4;19(10):e0311343. doi: 10.1371/journal.pone.0311343. eCollection 2024.
Color correction is an important methodology where a digital image's colors undergo a transformation to more accurately represent their appearance using a predefined set of illumination conditions. Colorimetric measurements in diagnostics are sensitive to very small changes in colors and therefore require consistent, reproducible illumination conditions to produce accurate results, making color correction a necessity. This paper presents an image color correction pipeline developed by HueDx, Inc., using transfer algorithms that improve upon existing methodologies and demonstrates real-world applications of this pipeline in colorimetric clinical chemistry using a smartphone enabled, paper-based total protein diagnostic assay. Our pipeline is able to compensate for a variety of illumination conditions to provide consistent imaging for quantitative colorimetric measurements using white-balancing, multivariate gaussian distributions and histogram regression via dynamic, non-linear interpolating lookup tables. We empirically demonstrate that each point in the color correction pipeline provides a theoretical basis for achieving consistent and precise color correction. To show this, we measure color difference with deltaE (ΔE00), alongside quantifying performance of the HueDx color correction system, including the phone hardware, color sticker manufacturing quality and software correction capabilities. The results show that the HueDx color correction system is capable of restoring images to near-imperceptible levels of difference independent of their original illumination conditions including brightness and color temperature. Comparisons drawn from the paper-based total protein assay calibrated and quantified with and without using the HueDx color correction pipeline show that the coefficient of variation in precision testing is almost twice as high without color-correcting. Limits of blank, detection and quantitation were also higher without color-correction. Overall, we were able to demonstrate the HueDx platform improves reading and outcome of the total protein diagnostic assay and is useful for the development of smartphone-based quantitative colorimetric diagnostic assays for point-of-care testing.
颜色校正（Color correction）是一种重要的方法，它通过将数字图像的颜色进行转换，使其在使用预定义的照明条件下更准确地呈现外观。诊断中的比色测量对颜色的微小变化非常敏感，因此需要一致且可重复的照明条件才能得出准确的结果，这使得颜色校正是必要的。本文介绍了由 HueDx, Inc. 开发的图像颜色校正管道，该管道使用传输算法来改进现有的方法，并展示了该管道在使用智能手机实现的、基于纸张的总蛋白诊断测定中的比色临床化学中的实际应用。我们的管道能够补偿各种照明条件，通过白平衡、多元高斯分布和通过动态、非线性插值查找表的直方图回归，为定量比色测量提供一致的成像。我们通过经验证明，颜色校正管道中的每个点都为实现一致和精确的颜色校正提供了理论基础。为了证明这一点，我们使用 deltaE（ΔE00）测量颜色差异，同时量化 HueDx 颜色校正系统的性能，包括手机硬件、颜色贴纸制造质量和软件校正能力。结果表明，HueDx 颜色校正系统能够将图像恢复到几乎无法察觉的差异水平，而不受其原始照明条件（包括亮度和色温）的影响。从使用和不使用 HueDx 颜色校正管道校准和量化的基于纸张的总蛋白测定中得出的比较表明，在不进行颜色校正的情况下，精密度测试的变异系数几乎高出一倍。没有颜色校正时，空白极限、检测极限和定量极限也更高。总体而言，我们能够证明 HueDx 平台提高了总蛋白诊断测定的读数和结果，并且对于开发基于智能手机的定量比色诊断测定用于即时检测是有用的。
