Efficient spectral matching, particularly for multicomponent spectra

• US 7,698,098 B2
• Filed: 02/18/2008
• Issued: 04/13/2010
• Est. Priority Date: 02/18/2008
• Status: Active Grant

• Current Assignee: Thermo Electron Scientific Instruments LLC (Thermo Fisher Scientific Incorporated)
• Original Assignee: Thermo Electron Scientific Instruments LLC (Thermo Fisher Scientific Incorporated)
• Inventors: McCarthy, William J., Grenov, Alexander I., Larson, Kathleen A., Ritter, Gary L., Izzia, Federico
• Primary Examiner(s): TSAI, CAROL S W
• Application Number:
  US12/032,794
• Publication Number:
  US 20090210194A1
• Time in Patent Office: 785 Days
• Field of Search: 702 27- 32, 702/179, 702/189, 702/194, 702/134, 702/137, 702/181, 702/40, 250/306, 250/307, 250/310, 250/311, 250/282, 250/347, 250/461.2, 250/339.05, 250/339.07, 250/339.12, 250/349, 250/353, 356/328, 356/303, 356/307, 356/326, 356/301, 356/456, 356/402
• US Class Current: 702/179
• CPC Class Codes: G01N 21/35 using infrared light G01N21/39 takes precedence

Abstract

An unknown spectrum obtained from infrared or other spectroscopy can be compared to spectra in a reference library to find the best matches. The best match spectra can then each in turn be combined with the reference spectra, with the combinations also being screened for best matches versus the unknown spectrum. These resulting best matches can then also undergo the foregoing combination and comparison steps. The process can repeat in this manner until an appropriate stopping point is reached, for example, when a desired number of best matches are identified, when some predetermined number of iterations has been performed, etc. This methodology is able to return best-match spectra (and combinations of spectra) with far fewer computational steps and greater speed than if all possible combinations of reference spectra are considered.

Claims

1. A method of identifying candidate spectra within an unknown spectrum, the method including the steps of:

a. obtaining a set of comparison spectra;

b. comparing the unknown spectrum to the comparison spectra;

c. identifying candidate spectra, each candidate spectrum being one of the comparison spectra having a desired degree of correspondence to the unknown spectrum;

d. performing the following substeps one or more times;

(1) generating new comparison spectra, each new comparison spectrum including;

(a) one of the candidate spectra, and

(b) one of the comparison spectra from the set;

(2) comparing the unknown spectrum to the new comparison spectra;

(3) identifying further candidate spectra, each candidate spectrum being one of the new comparison spectra having a desired degree of correspondence to the unknown spectrum,

wherein, when performing substeps (1)-(3) again, the candidate spectra of substep (3) are used as the candidate spectra when generating new comparison spectra in substep (1).

2. The method of claim 1 further including the step of presenting a list of at least some of the candidate spectra.

3. The method of claim 2 wherein the candidate spectra within the list are presented in ranked order, with the candidate spectra having greater correspondence to the unknown spectrum being listed first.

4. The method of claim 1 wherein, prior to comparing the unknown spectrum to one of the comparison spectra, one or both of the unknown spectrum and the comparison spectrum are transformed.

5. The method of claim 1 wherein, when comparing the unknown spectrum to one of the comparison spectra, regression is performed on the comparison spectrum versus the unknown spectrum.

6. The method of claim 5 wherein:

a. when performing regression on the comparison spectrum, a regression weight is calculated; and

b. the comparison spectrum is weighted by its regression weight.

7. The method of claim 6:

a. further including the step of presenting a list of at least some of the candidate spectra, and

b. wherein the regression weights of the comparison spectra defining the candidate spectra are presented in the list.

8. The method of claim 1 wherein, when generating new comparison spectra, each comparison spectrum includes:

a. one of the candidate spectra, and

b. one of the comparison spectra, wherein this comparison spectrum is not already included within the candidate spectrum.

9. The method of claim 1 wherein if any candidate spectrum contains the same comparison spectra as a previously-identified candidate spectrum, one of the candidate spectrum and the previously-identified candidate spectrum is discarded.

10. The method of claim 1 wherein, if a candidate spectrum corresponds to the unknown spectrum by at degree greater than or equal to a qualifying correspondence value, the comparison spectra therein are excluded from any later generation of new comparison spectra.

11. The method of claim 1 wherein step d. is halted once the new comparison spectra include a desired number of comparison spectra.

12. A method of identifying candidate spectra within an unknown spectrum, the method including the steps of:

a. providing one or more candidate spectra, each candidate spectrum being assumed to represent at least a portion of the unknown spectrum;

b. performing the following substeps one or more times;

(1) generating comparison spectra, each comparison spectrum being a combination of;

(a) one of the candidate spectra, and

(b) a reference spectrum corresponding to an identified substance,

wherein each candidate spectrum is combined with several different reference spectra to generate several comparison spectra;

(2) comparing the unknown spectrum to the comparison spectra;

(3) identifying further candidate spectra, each candidate spectrum being one of the comparison spectra which at least partially matches the unknown spectrum,

wherein, when performing the foregoing substeps (1)-(3) again, the candidate spectra of substep (3) are used as the candidate spectra when generating comparison spectra in substep (1).

13. The method of claim 12 wherein the step of providing one or more candidate spectra includes:

a. obtaining a set of several different reference spectra, each corresponding to an identified substance;

b. comparing the unknown spectrum to the reference spectra;

c. identifying the candidate spectra, each candidate spectrum being one of the reference spectra which at least partially matches the unknown spectrum.

14. The method of claim 12 further including the step of presenting a list of at least some of the candidate spectra, wherein the candidate spectra within the list are presented in ranked order with the candidate spectra having higher matching to the unknown spectrum being listed first.

15. The method of claim 12 further including the step of performing a transform on at least one of:

a. the unknown spectrum;

b. at least one of the comparison spectra;

c. at least one of the reference spectra; and

d. at least one of the candidate spectra.

16. The method of claim 12 further including the step of performing regression on at least one of the comparison spectra versus the unknown spectrum.

17. A method of identifying candidate spectra within an unknown spectrum, the method including the steps of:

a. comparing the unknown spectrum to at least some comparison spectra, each comparison spectrum being obtained from a spectral library;

b. saving at least some of the comparison spectra that best match the unknown spectrum, each such comparison spectrum being a candidate spectrum;

c. creating new comparison spectra, each new comparison spectrum being a combination of;

(1) one of the candidate spectra, and

(2) one of the comparison spectra within the spectral library,

d. comparing the unknown spectrum to at least some of the new comparison spectra,

e. saving at least some of the new comparison spectra that best match the unknown spectrum, each such comparison spectrum being a candidate spectrum;

f. repeating the foregoing step c. onward at least once;

g. presenting at least some of the candidate spectra.

18. The method of claim 17 wherein steps c.-f. are halted once the new comparison spectra include a predefined number of comparison spectra.

19. The method of claim 17 wherein the candidate spectra are presented in ranked order, with the candidate spectra having greater correspondence to the unknown spectrum being presented first.

20. The method of claim 17 wherein at least one of the following substeps is performed when comparing the unknown spectrum to one of the comparison spectra:

a. transforming one or both of the unknown spectrum and the comparison spectrum; and

b. regressing the comparison spectrum versus the unknown spectrum.

21. A system for identifying candidate spectra within an unknown spectrum, the system including a computer which:

a. receives an unknown spectrum;

b. accesses one or more records, each record including one or more comparison spectra;

c. compares the unknown spectrum to at least some of the comparison spectra from the records;

d. identifies candidate spectra, each candidate spectrum being one of the comparison spectra which matches the unknown spectrum with a desired degree of correspondence;

e. compares the unknown spectrum to new comparison spectra, each new comparison spectrum being a combination of;

(1) one of the candidate spectra, and

(2) one of the comparison spectra from the records,

f. identifies new candidate spectra, each new candidate spectrum being one of the new comparison spectra which matches the unknown spectrum with a desired degree of correspondence;

g. presents at least some of the candidate spectra to a user.

22. The system of claim 21 wherein the system, after identifying new candidate spectra:

a. compares the unknown spectrum to new comparison spectra, each new comparison spectrum being a combination of;

(1) one of the new candidate spectra, and

(2) one of the comparison spectra from the records,

b. identifies further new candidate spectra, wherein each new candidate spectrum is one of the new comparison spectra which matches the unknown spectrum with a desired degree of correspondence.

23. The system of claim 21 wherein the candidate spectra are presented to the user in ranked order, with the candidate spectra having the highest degree of correspondence to the unknown spectrum being presented first.

24. The system of claim 21 wherein the computer transforms at least one of:

a. the unknown spectrum, and

b. the comparison spectrum,

prior to comparing them.

25. The system of claim 21 wherein the computer performs regression on the comparison spectrum versus the unknown spectrum when comparing them.