why a smartwatch can testing the blood lipid ?

Non-invasive blood lipid testing technical documentation

The four intraluminal imaging technologies currently used in clinical practice are: grayscale intravascular ultrasound (Grayscale-IVUS), virtual

Histology Intravascular Ultrasound (VH-IVUS), Optical Coherence Tomography (OCT) and Near Infrared Spectroscopy (NIRS). nearly red

Compared with other intracavity imaging techniques, external spectroscopy has stronger penetrating power and can analyze the chemistry of biological tissues.

It is currently the only intraluminal imaging technology that can directly detect lipid plaques.

Near Infrared Spectroscopy (NIRS)

Figure 1 Analysis of shortwave near-infrared spectroscopy results

Figure 2 shows the spectral curve obtained after multiple scans of the same sample. It can be seen that the spectra obtained by the scans overlap.

And use its average spectrum as the spectrum curve of the sample figure 3 Internal interactive verification and external prediction verification of the calibration modelConclusion: This study conducted short-wave near-infrared detection of human body cholesterol content, and a total of 236 samples were collected.

There are 177 calibration sets and 59 validation sets, which are used to establish the calibration model and analyze and evaluate its model performance.

In the short-wave near-infrared region of 600-1099 nm, the spectral information of each sample was extracted, and standardized and SG first-order derivatives were used.

preprocessing methods such as number, difference derivation, SG smoothing, standard normal variable exchange, detrending correction and baseline correction.

Calibrate the spectral background and optimize the model. Use the partial least squares method to build a model and compare the effects to obtain the best

The preprocessing method is standardization and baseline correction and eliminating outliers twice. The obtained model parameters are Rc=0.801 1,

sEC=6.699 8, Rp=0.803 4, sEP=7.529 6, MF is 4, sEP/sEC=1.12, which proves that the model is more robust.

The prediction accuracy is high.