Development of a portable spectrophotometer using artificial neural networks for non-invasive determination of glycated hemoglobin in blood by Raman spectroscopy

Abstract

BACKGROUND: Non-invasive diagnosis of diabetes is one of the major problems of contemporary medicine. The system being planned could be a new technology for measuring hemoglobin A1c (HbA1c) accurately and non-invasively. Therefore, a series of studies are to be conducted to assess the efficiency of the method under study and determine its potential for medical diagnosis and monitoring of HbA1c.
 AIMS:
 
 Investigation of the feasibility of Raman spectroscopy for non-invasive measurement of HbA1c.
 Development and design of a portable analyzer using this technology.
 Assessment of the efficiency and accuracy of the developed device.
 
 METHODS: Neural network creation requires collecting a training sample of measurements for subsequent application of TensorFlow library tools and performing laboratory measurements to calibrate the system for determining HbA1c. The device will use a 785-nm laser to take spectra according to the Raman spectroscopy. The obtained data will be fed to the input of the neural network based on the architecture of convolutional neural networks. Experiments will be conducted to train the model to determine the accuracy and efficiency of the device. A two-step data collection procedure is planned. First, a preliminary test will be done on 50 patients to see how the proposed method handles different age and gender groups and different HbA1c levels. Later, the data will continue to be collected on a larger scale, including patients with different types of diabetes and healthy individuals. Data will be collected using a portable spectrophotometer and monitored by high-performance liquid chromatography. Various metrics will be used to assess the efficiency and accuracy of the device such as accuracy, precision, recall, and F1-score.
 RESULTS: An analysis of the available literature was conducted and the following conclusions were drawn. In addition, a neural network model was developed using HbA1c measurements. Currently, our model is optimized to improve the accuracy and reliability of the results.
 CONCLUSIONS: The non-invasive Raman spectroscopy-based method has several advantages in measuring HbA1c levels. The procedure is faster and non-traumatic, and HbA1c levels can be monitored continuously. In particular, the non-invasive method eliminates errors associated with protein leakage outside the bloodstream.