Systems and methods for interpreting high energy interactions
View Patent ↗Systems and methods for interpreting high-energy interactions on a sample are described in this application. In particular, this application describes an analysis method that comprises impinging radiation from a source on an analyte, detecting the energy interactions resulting from the impinging radiation using a radiation detector, adjusting the signal from the radiation detector using a machine learning module to emphasize specific parts of the detector signal, training the machine learning module in a supervised or unsupervised manner, producing quantitative and qualitative models using the machine leaning module, and then applying the machine learning module to additional energy interactions. The signal received by the detector can be preprocessed to emphasize specific parts of the detector signal, which is then mapped to a machine learning module for training in a supervised or unsupervised manner. The quantitative and qualitative models derived from this training can be applied to new detector inputs from the same or similar instruments. Other embodiments are described.
1. An analysis method, comprising:
training a machine learning module for interpreting high energy interactions, wherein training the machine learning module comprises:
impinging radiation from a source on an analyte;
detecting energy interactions resulting from the impinging radiation using a radiation detector, wherein the radiation detector produces a signal indicative of the detected energy interactions;
adjusting the signal from the radiation detector to emphasize specific parts of the signal of the radiation detector that are associated with quantitative or qualitative values;
producing quantitative and qualitative models derived from the machine leaning module; and
applying the machine learning module to additional energy interactions.
2. The method of claim 1 , wherein the source is an X-ray source, and wherein the radiation detector is an X-ray detector.
3. The method of claim 2 , wherein the signal of the X-ray detector is indicative of the fluorescent X-rays emitted from the analyte, and wherein the machine learning module determines a chemistry of the analyte.
4. The method of claim 1 , wherein the signal is indicative of fluorescence peaks, background radiation scatter, coherent (Bragg) scattering, inelastic (Compton) scattering, Bremsstrahlung scattering, or elastic (Rayleigh) scattering.
5. The method of claim 4 , wherein the machine learning module provides an output showing energy spectra of elements that are present in the analyte.
6. The method of claim 5 , wherein the machine learning module provides the output to a remote computing device.
7. The method of claim 1 , wherein the radiation detector that produces the signal is a first radiation detector, and wherein the additional energy interactions are detected by a second radiation detector that produces a second signal indicative of the additional energy interactions.
8. The method of claim 1 , wherein applying the machine learning module to additional energy interactions comprises:
impinging radiation from an additional source on a second analyte; and
detecting energy interactions resulting from the impinging radiation using a second radiation detector.
9. The method of claim 1 , wherein the analyte comprises solid material.
10. The method of claim 1 , wherein training the machine learning module comprises: receiving detector input from samples with known or estimated composition or qualities related to final output values obtained using the machine learning module.
11. The method of claim 10 , wherein training the machine learning module comprises: analyzing, based on a family of standards, data of preprocessed analytic output from a given sensor.
12. The method of claim 1 , wherein the machine learning module produces a diagram of detector input influence that identifies regions of the radiation detector inputs that have predictive power over quantitative or qualitative outputs.
13. The method of claim 1 , wherein the machine learning module is configured to identify both direct and indirect signals of a given value associated with an element within the analyte.
14. The method of claim 1 , wherein the machine learning module identifies an element of the analyte based at least in part on a fluorescence line produced by the analyte.
15. The method of claim 1 , wherein the machine learning module uses an equation having a structure of:
C i =re i ×re i+1 × . . . re i+n
wherein C i represents a concentration of a given element within the analyte, wherein e i represents a given energy calculated from outputs of the radiation detector, and wherein r represents a given weight or series of weights and/or other modifications for a particular energy (e).
16. The method of claim 15 , wherein the machine learning module uses a regression framework to evaluate variable (e i ) importance.
17. The method of claim 15 , wherein the machine learning module uses a decision tree, support vector machine (SVM), or k-nearest neighbor classifier related to output of the radiation detector and information about the given element.
18. The method of claim 1 , wherein the detector comprises a silicon-drift detector, a silicon pin diode, a silicon lithium detector, a germanium detector, a cadmium-telluride detector, a monochromater, a polychromater, a photomultiplier, a charged-doubled device, an image intensifier, a camera, and/or a digital flat panel detector.
19. The method of claim 1 , wherein the camera is a CMOS camera.
20. The method of claim 1 , wherein training the machine learning module is iterative.
21. The method of claim 1 , wherein adjusting the signal from the radiation detector to emphasize specific parts of the signal of the radiation detector that are associated with quantitative or qualitative values comprises normalization.
22. The method of claim 1 , wherein the analyte comprises liquid material.
23. The method of claim 1 , wherein applying the machine learning module to additional energy interactions comprises applying quantitative and qualitative models of the machine learning module to the additional energy interactions to interpret the additional energy interactions.
24. A method for training a machine learning module for interpreting high energy interactions, wherein the method comprises:
using a radiation detector to detect energy interactions resulting from radiation impinging on an analyte, wherein the radiation detector produces a signal indicative of the detected energy interactions; and
adjusting the signal from the radiation detector to emphasize specific parts of the signal of the radiation detector that are associated with quantitative or qualitative values; and
producing quantitative and qualitative models using the machine leaning module.
25. An analysis system, comprising:
a radiation source configured to impinge radiation on an analyte;
a radiation detector configured to:
detect energy interactions resulting from the impinging radiation; and
produce a signal indicative of the detected energy interactions; and
a machine learning module configured to;
produce quantitative and qualitative models,
wherein the machine learning module is trained by:
impinging radiation from the radiation source on the analyte;
detecting energy interactions resulting from the impinging radiation using the radiation detector;
adjusting the signal from the radiation detector to emphasize specific parts of the signal of the radiation detector that are associated with the quantitative or qualitative values.