White Blood Cell Count from Smartphone Captured Fingertip Video Using Deep Neural Network

This paper introduces a novel non-invasive method for estimating the number of white blood cells (WBC) using Photoplethysmography (PPG) signals from smartphone captured fingertip video. In this study, the fingertip videos from 90 participants have been captured before clinical Complete Blood Count (CBC) test. During the video recording process, individuals are directed to cover their index fingers on the smartphone camera and flashlight. After collecting fingertip video, the raw PPG signal is generated by evaluating the light intensity variations captured by the smartphone camera. The raw PPG signal has been preprocessed to remove noise and baseline drift. Therefore, 40 features have been retrieved from the clean PPG signal, including derivatives, Fourier analysis and single period characteristics. The deep neural network (DNN) model shows the strongest correlation with gold standard laboratory CBC tests with R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.934 and MAE of 0.609 ± 0.030. Smartphone-based WBC monitoring makes it easier to track immune health in real-time, helping people avoid frequent hospital visits and manage their health more conveniently.