Automatic Region of Interest Extraction from Finger Nail Images for Measuring Blood Hemoglobin Level
Shaharin Ahmed, Mohammad Ashfak Habib · 2021 International Conference on Electronics, Communications and Information Technology (ICECIT) · 2021
Point-of-care medical diagnosis has gained popularity for rapid diagnosis in medical field. Historically blood hemoglobin test was confined to medical laboratory and diagnostic centres. Currently, cell phone-based telehealth has empowered remote diagnosis. Researches on hemoglobin measurement by colorimetric analysis of user image taken from a smartphone are adding a new dimension to existing medical diagnosis. Images of body features like fingertip and fingernail are commonly used. Colorimetric analysis requires manually selecting perfect region of interest which has uniform color. The main objective of this paper is to extract region of interest automatically from fingernail images to detect blood hemoglobin level. A contour detection based process has been introduced to separate user hand image from the background, from which fingernails are detected. Our system can successfully subtract background if user hand is the closest object to the camera. By analyzing the color of fingernail data from smartphone images, our system calculates hemoglobin levels. Our application enables anyone with a smartphone to immediately detect blood hemoglobin level anywhere, anytime. A simple linear regression model was used to measure blood hemoglobin level. Our application is able to measure blood hemoglobin level with MAE value of 0.64, MSE value of 0.62, RMSE value of 0.78 and R2 of 0.69.