Calibration transfer for analytical instruments

• US 5,459,677 A
• Filed: 06/24/1991
• Issued: 10/17/1995
• Est. Priority Date: 10/09/1990
• Status: Expired due to Term

• Current Assignee: Board of Regents of The University of Washington (University of Washington)
• Original Assignee: Board of Regents of The University of Washington (University of Washington)
• Inventors: Wang, Yong D., Veltkamp, David J., Kowalski, Bruce R.
• Primary Examiner(s): Voeltz, Emanual T.
• Assistant Examiner(s): Pipala, Edward
• Application Number:
  US07/720,256
• Time in Patent Office: 1,576 Days
• Field of Search: 364/571.04, 364/571.02, 364/497, 364/498, 250/252.1, 73/1 R
• US Class Current: 703/2
• CPC Class Codes: G01N 2030/025 Gas chromatography
• CPC Class Codes: G01N 2030/626 calibration, baseline
• CPC Class Codes: G01N 21/274 Calibration, base line adjustment, drift correction
• CPC Class Codes: G01N 30/8672 not depending on an individual instrument, e.g. retention time indexes or calibration transfer

Abstract

A technique for transferring a multivariant calibration model from a reference analytical instrument to a target analytical instrument that may be a different instrument, or the same instrument at a later time. In a "direct" approach, a plurality of transfer samples are selected, and a plurality of measurements are made for each transfer sample using the reference instrument, producing a reference instrument response for each sample. These measurements are repeated for the target instrument, to produce a target instrument response for each transfer sample. One then generates transfer coefficients capable of performing a multivariate estimation of the reference instrument responses for the transfer samples from the target instrument responses for those samples. The transfer coefficients may then be used to convert a target instrument response for an unknown sample into the equivalent response for the reference instrument. In an important "piecewise" variation, the transfer coefficients comprise a plurality of estimation coefficients for estimating each reference data value from more-than-one but less-than-all target data values. "Classical" and "inverse" transfer techniques are also described, wherein target instrument responses for the transfer samples are combined with reference instrument responses for the full set of calibration samples, to derive a multivariate prediction model for the target instrument.

Claims

1. A method of producing a multivariate prediction model for predicting a property of a sample using a target spectrometer, based in part upon data measured by a reference spectrometer, the method comprising:

producing property data by measuring said property for each of a plurality of calibration samples;

for each calibration sample, producing a reference spectrometer response by making a plurality of measurements using the reference spectrometer;

for each transfer sample in a subset of the calibration samples, producing a target spectrometer response by making a plurality of measurements using the target spectrometer;

using the reference spectrometer responses for the calibration samples, the target spectrometer responses for the transfer samples, and the property data, producing estimates of the responses of the target spectrometer to the calibration samples; and

producing a multivariate prediction model for the target spectrometer by combining the estimates of the responses of the target spectrometer to the calibration samples with the property data.

2. The method of claim 1, comprising the further steps of producing a target spectrometer response for an unknown sample using the target spectrometer, and predicting said property for the unknown sample by combining the multivariate prediction model for the target spectrometer with the target spectrometer response for the unknown sample.

3. The method of claim 1, wherein the step of producing estimates comprises determining

4. The method of claim 3, comprising the further steps of producing a target spectrometer response for an unknown sample using the target spectrometer, and predicting said property for the unknown sample by combining the multivariate prediction model for the target spectrometer with the target spectrometer response for the unknown sample.

5. A method of producing a multivariate prediction model for predicting a property of a sample using a target spectrometer, based in part upon data measured by a reference spectrometer, the method comprising:

producing property data by measuring said property for each of a plurality of calibration samples;

for each calibration sample, producing a reference spectrometer response making a plurality of measurements using the reference spectrometer;

for each transfer sample in a subset of the calibration samples, producing a target spectrometer response by making a plurality of measurements using the target spectrometer;

producing a multivariate prediction model for the reference spectrometer by using the reference spectrometer responses for the calibration samples the property data; and

using the reference spectrometer responses for the transfer samples, the target spectrometer responses for the transfer samples, the multivariate prediction model for the reference spectrometer, and the property data, producing a multivariate prediction model for the target spectrometer.

6. To the method of claim 5, comprising the further steps of producing a target spectrometer response for an unknown sample by using the target spectrometer, and predicting said property for the unknown sample by combining the multivariate prediction model for the target spectrometer with the target spectrometer response for the unknown sample.

7. The method of claim 5, wherein the multivariate prediction model for the target spectrometer is determined as follows

8. The method of claim 7, comprising the further steps of producing a target spectrometer response for an unknown sample using the target spectrometer, and predicting said property for the unknown sample by combining the multivariate prediction model for the target spectrometer with the target spectrometer response for the unknown sample.

9. A method of using a target spectrometer to perform a plurality of measurements on an unknown sample so as to produce an estimate of a reference spectrometer response for the unknown sample, the reference spectrometer response comprising values that would have been obtained by performing said plurality of measurements on the unknown sample using a reference spectrometer, the method comprising:

providing a plurality of transfer samples;

producing a reference spectrometer response for each transfer sample, by making a plurality of measurements using the reference spectrometer;

producing a target spectrometer response for each transfer sample, by making a plurality of measurements using the target spectrometer;

producing transfer coefficients for performing a multivariate estimation of the reference spectrometer responses for the transfer samples from the target spectrometer responses for the transfer samples;

producing a target spectrometer response for the unknown sample, by making a plurality of measurements using the target spectrometer; and

converting the target spectrometer response for the unknown sample into an estimate of the reference spectrometer response for the unknown sample, using the transfer coefficients.

10. The method of claim 9, wherein each reference spectrometer response comprises a reference data value for each of M measurements made using the reference spectrometer, wherein each target spectrometer response comprises a target data value for each of M measurements made using the target spectrometer, and wherein the transfer coefficients comprise a plurality of estimation coefficients for estimating each reference data value from a window comprising a plurality of substantially less than M target data values.

11. The method of claim 10, wherein each reference spectrometer response includes one or more nonlinear functions of the measurements made using the reference spectrometer, and wherein each target spectrometer response includes the same nonlinear functions of the measurements made using the target spectrometer.

12. The method of claim 9, wherein the transfer coefficients F are produced by determining:

space="preserve" listing-type="equation">F=R.sub.2 +R.sub.1

where R₂⁺ is the inverse of a matrix formed from the target spectrometer responses for the transfer samples of R₁ is a matrix formed from the reference spectrometer response for the transfer samples.

13. The method of claim 9, wherein each reference spectrometer response and each target spectrometer response comprises a spectrum comprising a plurality of absorbance values at a corresponding plurality of wavelengths.

14. The method of claim 9, wherein the target spectrometer and the reference spectrometer are different spectrometers.

15. The method of claim 14, wherein each reference spectrometer response comprises a reference data value for each of M measurements made using the reference spectrometer, wherein each target spectrometer response comprises a target data value for each of M measurements made using the target spectrometer, and wherein the transfer coefficients comprise a plurality of estimation coefficients for estimating each reference data value from a window comprising a plurality of substantially less than M target data values.

16. The method of claim 14, wherein the transfer coefficients F are produced by determining

17. The method of claim 9, comprising the further steps of:

providing a plurality of calibration samples, the calibration samples including the transfer samples plus additional samples;

producing a reference spectrometer response for each calibration sample, by making a plurality of measurements using the reference spectrometer;

producing property data by measuring a property for each calibration sample;

producing a multivariate prediction model for the reference spectrometer, using the reference spectrometer responses for the calibration samples and the property data; and

predicting said property for the unknown sample by combining the estimate of the reference spectrometer response for the unknown sample with the multivariate prediction model for the reference spectrometer.

18. The method of claim 17, wherein said property is a concentration of an analyte included in the calibration samples.