**4. Development of non-invasive measurement of blood glucose of diabetic**

#### **4.1 Background**

Recently, increasing of diabetic has been brought to public attention. According to WHO, the number of diabetics is 170 million. To treat diabetes, diabetics should always check blood glucose monitoring. Therefore, they need to self-monitor blood glucose (SMBG).

But, increasing the number of factors too, applies Standard Error of Prediction (SEP) increases (Over Fitting). Therefore, we validate the optimal number of PLS factors by Leaveone-out method. The Prediction Residual Error Sum of Squares1 (PRESS1) values calculated

PRESSn=Σ(yobs-yref)2 (2)

After, the new objective variables are calculated by calibration curve adding a new PLS factor. PRESS2 value is calculated from the objective variables obtained again. If the residual is significant before and after (The difference between the PRESS1 and PRESS2), adding the PLS factor. If the residual is not significant before and after, select the model that was built before. After, we measured to the spectrum of sample of unknown amount, and calculated by using calibration curve and this spectrum. By the above process, it is possible to be

We used SIMCA for the qualitative analysis. A class is made by infrared spectra of known sample and builds a classification model. Each class is analysed by analysis of principal component and the distinction space is set. This space is called SIMCA box. It is classified into the class suited most by applying infrared spectra of unknown sample to SIMCA box. Moreover, the rest error is calculated by applying each spectrum that composes the class to other classes. And, Discrimination Power that can specify the factor in which it distinguishes between classes is obtained. We confirmed the validity of the classification model

KNN method is one of pattern analysis to determine the class by comparing the similarity between the patterns not based on specific statistical distributions. To determine the class of an unknown sample is made on "voting". First, calculate the Euclidean distance between samples for the known and unknown class samples. Next, select a known class samples for the number of "K" close to the distance from an unknown class sample. "K" is odd number. The class of an unknown sample is determined to a most numerous class in

For example, if the "K=5", analyse the closest class of 5 samples from an unknown sample class. Five classes (1,2,3,3,3), (1,3,3,2,1) are classified at Class "3", (1,1,2,1,2), (2,1,3,1,1) are classified at Class "1". For such analysis, impact on the accuracy of the analysis is a

**4. Development of non-invasive measurement of blood glucose of diabetic** 

Recently, increasing of diabetic has been brought to public attention. According to WHO, the number of diabetics is 170 million. To treat diabetes, diabetics should always check blood glucose monitoring. Therefore, they need to self-monitor blood glucose (SMBG).

combination of variables used to calculate the distance and the number of "K".

from the objective variables by the following equation 2

yobs: new objective variables, yref: reference values

measured quantitatively of unknown sample.

constructed by using Discrimination Power.

**3.4 K-Nearest Neighbor method (KNN)** 

the "K".

**4.1 Background** 

**3.3 Soft Independent Modeling of Class Analogy (SIMCA)** 

SMBG is measured by blood sampling method, but this method the patient suffering and stress, including issues such as the risk of infection. And, the economic burden on patients is very large, because medical needles and measurement kit are disposable. Medical expenses of diabetes and its complications are estimated at about 3,000 billion dollars worldwide, and are expected to continue to increase in the future. Those various studies have been conducted around the world, because the medical expenses have become large economic markets. But, the effective blood glucose measurement method to overcome these problems, have not yet been developed. Therefore, it is desired to develop a method to measure non-invasive blood glucose measurement. Over the past few years, several studies have been made on non-invasive blood glucose measurement based on ATR infrared spectroscopy. The purpose of this study is to examine the accuracy of blood glucose in clinical trial.

#### **4.2 Measurement system for non-invasive measurement of blood glucose**

This study used FT-IR (Travel-IR) and ATR method. The block diagram of measurement system is shown in Fig.3. In the ATR prism, used to the prism of diamond mounted on ZnSe (3 times reflection).

The measurement part is the tip of the left hand middle finger of subject. The middle finger was washed with ethanol. The 5μl squalene oil was applied on the prism by micropipette in each measurement. Squalene oil is used as an internal standard method described below. The measurement part of subject put on the prism, pressed from above with a constant pressure. We measured in this state, and got the absorption spectrum including the blood glucose value information.

Fig. 3. Non-invasive measurement system for blood glucose

The absorption spectrum was applied to the ATR correction. We used 1800 cm-1 wavenumber in standard wavenumber of ATR correction, because absorption peak and noise are not in this wavenumber. After, all measured spectra were applied data correction between 2700 cm-1 and 1750 cm-1 to remove the absorption noise of diamond prism, and, were applied normalization correction in the absorption peak of squalene oil. We used these corrected absorption spectra in analysis. In the measurement condition, measurement wavenumber range is 4000~700 cm-1, resolution is 4 cm-1, and accumulation is 30 times.

Introduction of Non-Invasive Measurement Method by Infrared Application 85

A

E

A

Zone Judge

C Little danger

Safety

Danger

glucose

Squalene oil

A B

D E

A

B

D

E

C

Fig. 4. Clarke grid

2

0

1

Absorbance

Estimated value (mg/dl)

Fig. 5. The absorption spectra of glucose powder and squalene oil

0 50 100 150 200 250 300 350 400

Reference value (mg/dl)

F

B

D D

C

B

C

E

**4.4 Non-invasive blood glucose measurement in subjects** 

**4.4.1 Development of calibration curve** 

absorption peak of glucose and oil were not overlap.

absorption peaks in 1377 cm-1 (Fig.5-D), 1462cm-1 (Fig.5-E), and 2922 cm-1 (Fig.5-F). The

4000 3500 3000 2500 2000 1500 1000 500

Wavenumber (cm-1)

The absorption spectra of the finger in before and after coating to squalene oil are shown in Fig.6. In absorption spectrum of finger, absorption peaks in 1377 cm-1 and 1462 cm-1 did not appear. But, absorption spectrum of finger coated with squalene oil has absorption peaks in 1377 cm-1 (Fig.6-D) and 1462 cm-1 (Fig.6-E). Therefore, it is possible to reduce individual differences in skin surface conditions due to the absorption spectra normalized at the absorption peak of squalene oil. We confirmed that the squalene oil is suitable as an internal standard method, and can measure the blood glucose value by absorption peaks of glucose.

The subjects were healthy eight men in their 20s. When developing a calibration curve, glucose tolerance test were performed on each subjects. In glucose tolerance test, first, the

### **4.2.1 Internal standard method by squalene oil**

In measuring the subject's blood glucose value, the good accuracy calibration curve is essential. To improve the accuracy of the calibration curve must be accurately extract glucose information on the infrared absorption spectrum.

In the ATR method, it is important to stick a sample to the prism. Many people of diabetes are elderly, person with dry skin on the fingertips are also often elderly. Squalene oil is used as internal standard method for the different dry skin of the subject's finger surface. To eliminate the effects of dry skin by applying the squalene oil, we can measure the subjects under the same conditions. And, there is no effect of squalene oil to apply normalization correction in the absorption peak of squalene oil. Furthermore, the S/N ratio of the spectrum is better, because to block the air from between the finger and prism by the oil.
