Apparatus and method for estimating blood sugar based on heterogeneous spectrums
An apparatus and a method for determining blood sugar based on heterogeneous spectrums are provided. The blood sugar model generation apparatus includes a data obtainer configured to obtain blood sugar profile data of a user based on a first type spectrum-blood sugar profile relationship model, a second type spectrum measurer configured to measure training second type spectrum data for the skin of the user, and a processor configured to generate an individualized blood sugar model based on the obtained blood sugar profile data and the measured training second type spectrum data.
1. A blood sugar model generation apparatus comprising:
a storage that stores computer code; and
a processor that accesses and executes the computer code to:
generate a first type spectrum-blood sugar profile relationship model by training the first type spectrum-blood sugar profile relationship model using first type spectrum data and blood sugar profile data on a plurality of subjects whose first type spectrum data have been measured;
calculate blood sugar profile data of a user based on the first type spectrum-blood sugar profile relationship model and obtained first type spectrum data for skin of the user;
generate an individualized blood sugar model by training the individualized blood sugar model using second type spectrum data for the skin of the user and blood sugar values calculated from the blood sugar profile data of the user and corresponding to the second type spectrum data; and
estimate a blood sugar value of the user using the individualized blood sugar model and an obtained second type spectrum data for the skin of the user, wherein the first type spectrum is a different type of spectrum from the second type spectrum.
2. The blood sugar model generation apparatus of claim 1 , wherein the first type spectrum is a Raman spectrum, and the second type spectrum is a near infrared (NIR) spectrum.
3. The blood sugar model generation apparatus of claim 1 , wherein the blood sugar profile data is obtained by performing an oral glucose tolerance test (OGTT) on the subjects whose first type spectrum data has been measured.
4. The blood sugar model generation apparatus of claim 1 , wherein the first type spectrum-blood sugar profile relationship model is generated by training the first type spectrum-blood sugar profile relationship model using the first type spectrum data and an area under a curve (AUC) value calculated from blood sugar profile data.
5. The blood sugar model generation apparatus of claim 1 , wherein the training comprises a machine learning algorithm selected from among partial least squares regression, linear regression, neural network, decision tree, genetic algorithm, genetic programming, K-nearest neighbor, radial basis function network, random, forest, support vector machine, and deep-learning.
6. The blood sugar model generation apparatus of claim 1 , further comprising:
a light source configured to emit light to the skin of the user; and
a spectroscope configured to detect absorbed, scattered, or reflected light from the skin of the user,
wherein the second type spectrum data is measured based on the detected light.