Patent Application
20190026439
It is becoming increasingly important and urgent to rapidly and accurately identify toxic materials or pathogens with a high degree of reliability, particularly when the toxins/pathogens may be purposefully or inadvertently mixed with other materials. In uncontrolled environments, such as the atmosphere, a wide variety of airborne organic particles from humans, plants and animals occur naturally. Many of these naturally occurring organic particles appear similar to some toxins and pathogens even at a genetic level. It is important to be able to distinguish between these organic particles and the toxins/pathogens.
In cases where toxins and/or pathogens are purposely used to inflict harm or damage, they are typically mixed with so-called “masking agents” to conceal their identity. These masking agents are used to trick various detection methods and apparatus to overlook or be unable to distinguish the toxins/pathogens mixed therewith. This is a recurring concern for homeland security where the malicious use of toxins and/or infectious pathogens may disrupt the nation's air, water and/or food supplies. Additionally, certain businesses and industries could also benefit from the rapid and accurate identification of the components of mixtures and materials. One such industry that comes to mind is the drug manufacturing industry, where the identification of mixture composition could aid in preventing the alteration of prescription and non-prescription drugs.
One known method for identifying materials and organic substances contained within a mixture, or in elemental form, is to measure the absorbance, transmission, reflectance or emission of each material as a function of the wavelength or frequency of the illuminating or scattered light transmitted through the material. In the case of a mixture this requires that the mixture be separable into its component parts. Such measurements as a function of wavelength or frequency produce a plot that is generally referred to as a spectrum. The spectra of the material or object, i.e., sample spectra, can be identified by comparing the sample spectra to a set of reference spectra that have been individually collected for a set of known elements or materials. The set of reference spectra are typically referred to as a spectral library, and the process of comparing the sample spectra to the spectral library is generally termed a spectral library search.
Spectral library searches have been described in the literature for many years, and are widely used today. Spectral library searches using infrared (approximately 750 nm to 100 μm wavelength), Raman, fluorescence or near infrared (approximately 750 nm to 2500 nm wavelength) transmissions are well suited to identify many materials due to the rich set of detailed features these spectroscopy techniques generally produce. The above-identified spectroscopy techniques produce a rich fingerprint of the various pure entities and can be used to identify the component whether alone or in a mixture.
Conventional library searches generally and other such applications are time consuming and memory intensive. The process is also memory intensive because spectral libraries can be substantial in size. The instant application overcomes this and other problems by searching a sub-set of the library.
In one embodiment, the disclosure relates to a method for determining an identity of an unknown material by (a) obtaining a spectrum of the material wherein the spectrum represents the unknown material; (b) representing the spectrum as a target vector; (c) providing a vector space containing a plurality of known vectors representing the spectra of known materials; (d) mapping the target vector into the vector space; (e) determining a correlation between the target and the known vectors; (f) identifying the unknown material as the known material based upon the determined correlation.
In another embodiment, the disclosure relates to a method for conducting a spectral library search to identify an unknown compound comprising acquiring one or more spectra of the compound; representing each spectrum as a target vector; providing an n-dimensional space having a plurality of partitioned spaces, at least one of the partitioned spaces containing at least one known vector representing a known material; mapping each target vector in one of the plurality of the partitioned spaces to form a mapped partitioned space; identifying one or more known vectors within the mapped partitioned space which approximate the target vector; and identifying the unknown compound by comparing the target vector to the known vectors within the mapped partitioned space which closely approximate the target vector.
In still another embodiment, the disclosure relates to a system for identifying the composition of an unknown material comprising: acquiring one or more spectra of the unknown material; a processor programmed with a first instruction for representing each spectrum as a target vector; a database for providing a plurality of partitioned spaces, at least one of the partitioned spaces containing at least one known vector representing a known material; the processor programmed with second instruction for: (i) mapping one of the target vectors in one of the partitioned spaces to form a mapped partitioned space; (ii) identifying one or more known vectors within the mapped partitioned space which approximate the target vector; and (iii) determining the identification of the unknown material by selecting a candidate which provides the closes approximation to the target vector of the unknown material.
In still another embodiment, the disclosure relates to a method for identifying of an unknown material comprising acquiring one or more spectrum of the unknown material; representing each spectrum as a target vector; providing a plurality of partitioned spaces, wherein at least one of said partitioned spaces contains at least one known vector representing a known material; mapping one of the target vectors into one of the partitioned spaces to form a mapped partitioned space; identifying at least one known vector within the mapped partitioned space which approximates the target vector; identifying adjacent mapped partitioned spaces having at least one vector approximating the target vector; and calculating a correlation between the target spectrum and the known vectors in the partitioned space and the adjacent known partitioned spaces to identify the best candidate.
In yet another embodiment, the disclosure relates to a system for identifying of an unknown material comprising acquiring one or more spectra of the unknown material; a processor programmed with a first set of instructions for representing each spectrum as a target vector; a database for providing a plurality of partitioned spaces, wherein at least one of said partitioned spaces contains at least one known vector representing a known material; the processor programmed with a second set of instructions for: (i) mapping one of the target vectors into one of the partitioned spaces to form a mapped partitioned space; (ii) identifying at least one known vector within the mapped partitioned space which approximates the target vector; (iii) identifying adjacent mapped partitioned spaces having at least one vector approximating the target vector; and (iv) calculating a correlation between the target spectrum and the known vectors in the partitioned space and the adjacent known partitioned spaces to identify the best candidate.
A chemical image is compiled from several frames having a plurality of spectra. A pixel of the image can be deconstructed into a plurality of frames where each frame of the pixel denotes a relationship between intensity and wavelength (or wave-number).
The spectrum can be collected using various spectroscopical techniques, including infrared, Raman, Fluorescence and near infrared techniques. The spectra of many materials are often collected in a library of known materials. The library spectra should be corrected to remove all signals and information that are not due to the chemical compositions of the samples and known elements/material. Such anomalies include various instrumental effects, such as the transmission of optical elements, the detector's responsiveness, and any other non-desired sample effect due to the instrument utilized for collecting the spectra. It is noted that the uncorrected spectra may also be used without departing from the principles disclosed herein. Thus, an optional step according to an embodiment of the disclosure may include removing instrument-dependent error from the spectra and/or the library. This step can be implemented by using the transfer function of the instrument.
PCA is a data dimensionality reduction technique based on a multivariate least-squares calculation. It is similar to an Eigenvector analysis calculation (typically associated with the mathematical field of linear algebra). PCA results in the representation of a set of data by a reduced set of factors where some percentage of the variance of the data set is explained (typically 95, 99, or 99.5%). Thus, the relative position of the points in the reduced n-dimensional space with respect to each other is unchanged relative to the position of the points in the original-dimensional space to the extent that the variance of the data set is explained. In practical terms, this means that a set of spectra can be represented by a greatly reduced set of abstract factors (alternatively termed principal components or eigenvectors). Typically a set of 1024 spectral point spectra can be represented as a set of 10-15 point abstract factors with virtually no loss in accuracy. This reduction allows the partitioning of a dataspace with much smaller numbers of dimensions than the original dataspace. Without this reduction, the n-dimensional partitioning would not be practical for most spectral data sets.
In one embodiment of the disclosure an n-dimensional space (or a partitioned space) is used to relate the unknown composition with the known materials. In contrast with the conventional library searches, this method provides an expedited operation. An n-dimensional space can have any general form, for example, a sphere with multiple axial vectors intercepting at one point in space. For the sake of simplicity, the inventive concepts will be discussed in relation to a three-dimensional space; however, the disclosure is not limited thereto. Extending the inventive concepts from a three-dimensional space to an n-dimensional space is well within the skill of one of ordinary skill in the art.
According to one embodiment of the disclosure, the identity of an unknown material can be detected by obtaining a spectrum of the material. This step can be implemented using any conventional spectroscopic device. Next, the spectrum can be reduced to one or more target vectors. Conventional algorithms, such as those identified above including PCA, can be used for reducing the spectrum to a target vector. To determine the identity of the unknown material a vector space containing a plurality of known vectors can be then provided. The vector space can be an n-dimensional space containing the spectra of known components in the vector form. In other words, the n-dimensional space can be constructed around one or more vectors representing spectra of known material. Such vectors can be defined based on the spectra of known material. For example, an n-dimensional space can be constructed based on vectors representing such known material as sugar, flour, salt, anthrax, etc. Each known vector defines a point of origin and an end point, which in turn define direction and magnitude of the vector. A plurality of such vectors then form an n-dimensional space within which the spectrum of the known material (i.e., the unknown spectrum) can be identified.
Once the n-dimensional space is constructed, the unknown spectrum (interchangeably, the target vector) can be mapped into the vector space. Typically, the target vector is mapped such that the vector's origin is consistent with the origin of the other known vectors. The target vector then extends in a particular direction to its end point in the vector space. Once mapped into the vector space, a correlation between the target vector and one or more the known vectors can be determined. Such correlation may be, for example, mathematical or geometrical. Once a correlation between the target vector and at least one known vector is determined, the identity of the unknown material can be readily ascertained.
Accordingly to one embodiment of the disclosure, the process of correlating the target and the known vector can be implemented by partitioning the vector space into sub-spaces (interchangeably, partitioned spaces) containing one or more of the known vectors. The sub-spaces can have any form. In one embodiment, the sub-spaces define cubes of consistent size. Moreover, each sub-space can have one or more known vectors thereon. The partitioning of the n-dimensional space into a set of subspaces is performed by the following steps. 1) A minimum and maximum value is determined for each coordinate axis in the n-dimensional space (based on the projection of the known vectors onto that axis). 2) Each coordinate axis that has a length greater than some minimum length (based on the range of the coordinate axis with the largest range) is then divided by an integer M. 3) The resultant divided coordinate axes form a set of subspaces in the n-dimensional space. 4) M is generally set to 2 initially and is incremented by increments of 1 until the desired degree of partitioning is reached. 5) The degree of partitioning is determined by the density of the number of known vectors that map into any given subspace.
After the sub-spaces are defined, the identity of the target vector can be determined by correlating the target vector to an appropriate sub-space. For example, for one of the sub-spaces into which the target vector is mapped, a correlation between the target vector and each of the known vectors in said sub-space can be determined. The identity of the unknown vector can be then determined based on such a correlation.
In one exemplary embodiment, the step of providing a plurality of partitioned spaces includes providing an n-dimensional space with a plurality of known vectors in n-dimensions. Each of the coordinate axes of the n-dimensional space can then be divided by an integer M to get a set of n-dimensional subspaces where each n-dimensional subspace is populated by a plurality of the known vectors. Next, a ratio of known vectors occupying each n-dimensional subspace as a percentage of a total number of known vectors can be determined. Once such ratio is defined, the n-dimensional space can be divided into further sub-spaces if the ratio of the known vectors in any of the n-dimensional subspace exceeds a threshold. This process is aided by finding a minimum point and a maximum point for each of the coordinate axes. In an alternative embodiment, the threshold can be selected algorithmically. The threshold can have any range. For example, the threshold can be in the range of 20-80&, 5-90% or 1-99%.
The embodiments described above can be implemented with a processor in communication with a database and other electronic peripherals. For example, an image forming spectra of an unknown material can be first obtained. Each spectra is a function of at least one of intensity, wavelength, wave number or frequency. Next, a processor can be programmed with a set of instructions to represent each spectrum as a target vector. The processor can communicate with a database for providing a plurality of partitioned spaces, each partitioned space containing at least one known vector representing a known component. The same or a different processor can also be programmed with a second set of instructions for, among others, (1) mapping one of the target vectors in one of the partitioned spaces to form a mapped partitioned space; (2) identifying one or more known vectors within the mapped partitioned space which approximate the target vector; and (3) determining the identity of the unknown material by selecting a candidate which provides the closest approximation to the target vector of the unknown material. According to this embodiment, the step of-mapping each target vector to one of the plurality of partitioned spaces may further include representing each target data point as a vector; and identifying each n-dimensional space where the target vector resides.
The embodiments disclosed herein are exemplary in nature and are intended to illustrate, not limit, applicant's inventive principles.
The instant application relates to and claims the filing-date benefit of PCT Application No. PCT/US05/13036 filed Apr. 15, 2005, which claims the filing-date benefit of U.S. patent application Ser. No. 10/812,233 filed Mar. 29, 2004; the specification of each of these applications is incorporated herein in its entirety for background information.
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/US2006/006213 | 2/23/2006 | WO | 00 | 1/6/2009 |
| Number | Date | Country | |
|---|---|---|---|
| 20090171593 A1 | Jul 2009 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | PCT/US2005/013036 | Apr 2005 | US |
| Child | 11918486 | US |
