Brevet US5606164 - Method and apparatus for biological fluid analyte concentration measurement ...

1 .Collect calibration data set.

2. Perform data pretreatment including mean centering, variance scaling, n^ derivative, smoothing, k ratiometric method.

3. Decompose data into score and loading vectors using methods such as non-linear iterative partial least squares.

4. Determine appropriate number ot factors to include in a model which adequately represents the data.

9. Determine generalized distance by methods including Mahalanobis distance (MD), and Robust distance (RD).

11. Determine probability of class membership by evaluation of chi-squared distribution function or Hotelling's T-statistic.

12. Identify outliers as having large generalized distance which results in a low probability of class membership.

13. Remove outliers from the calibration set.

14. Do outliers have very large distances? |—*- Yes

No A

!±:

15. Determine calibration model by methods including principal component regression,

partial least squares, multiple linear regression and artificial neural networks.

16. Reduce data set down to significant factors by computing principal component analysis scores

or partial least squares scores.

17. Use significant factors and reference data

to calculate regression coefficients and artificial neural networks weights.

Fig. 5

18. Collect data from samples where concentration of analyte of interest is unknown.

19. Apply pretreatment techniques used during the calibration phase.

20. Project data into principal component space defined by the calibration model.

Brevets