Improving hemoglobin estimation accuracy through standardizing of LED power

Nowadays, hemoglobin monitoring is essential during surgeries, blood donations, dialysis, etc. which are normally done using invasive methods. To monitor hemoglobin, a non-invasive hemoglobin meter was developed with five fixed LED wavelengths at 670 nm, 770 nm, 810 nm, 850 nm, 950 nm and controlled using an Arduino Uno Embedded Development Board. A photodetector with an on-chip trans-impedance amplifier was utilised to acquire the transmitted signal through the finger using the Photoplethysmography (PPG) principle. Before the standardization of LED power, we had tested the designed system on fifteen subjects for the five wavelengths and estimated the hemoglobin with an accuracy of 96.51 % and Root mean square error (RMSE) of 0.57 gm/dL.To further improve the accuracy, the LED Power was standardized and the PPG signal was reacquiredon the same subjects. With this, the accuracy improved to 98.29 % and also reduced the RMSE to 0.36 gm/dL. The designed system with LED power standardization showed a good agreement with pathology results with the coefficient of determination R 2 = 0.981. Also, Bland–Altman analysiswas used to evaluate the designed system and it showed good agreement between the two measurements.