Device and method for material identification from a set of monochromatic images using spectral computed tomography

The device and method use a predefined set of known materials and AI to accurately identify material composition and density from monochromatic images, addressing integration and noise challenges in sCT systems for diverse applications.

WO2026240663A2PCT designated stage Publication Date: 2026-11-19VINAVER MEDICAL D O O
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/RS2026/000006
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-15
Filing Date
2026-05-13
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

Existing methods for material identification using spectral computed tomography (sCT) face challenges in achieving precise chemical composition and density analysis from monochromatic images, particularly due to noise amplification and the need for detailed knowledge of CT system geometry, limiting integration with existing hardware and accuracy in material differentiation.

Method used

A device and method utilizing a predefined set of known materials, combined with machine learning and artificial intelligence, to identify materials from a set of monochromatic images, integrating with existing sCT systems and providing full material composition and density analysis.

Benefits of technology

Enables accurate material identification with improved image quality and integration capabilities, overcoming noise issues and enabling applications in medical diagnostics, security screening, and cultural heritage analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure RS2026000006_19112026_PF_FP_ABST
    Figure RS2026000006_19112026_PF_FP_ABST
Patent Text Reader

Abstract

A device and method for identification of materials from a set of monochromatic images by spectral computed tomography belongs to the field of devices that enable determination of the chemical composition and density of materials obtained directly by a monochromatic X-ray beam or indirectly from a set of virtual monochromatic images by means of spectral computed tomography (sCT), for the purpose of enabling precise medical diagnostics, planning of therapies and treatments, application of safety in scanning, precise calculation and deposition of doses, correction of X-ray absorption effects, non-destructive resting and detection of explosives, quality control of food, professional analysis of materials, and study of pigments, coatings and structural materials in historical artifacts without damage. This is achieved in that the device (1) comprises : an X-ray source (2) emitting an X-ray beam (3) toward an object (4) being scanned, a detector (5) receiving (6) attenuated by passing through the object (4), a control device (7) operatively connected with a computer (8) and a cloud (9), configured so as to enable the use of a predefined set of known materials for identification, instead of attempting to calculate quantitative properties of the material exclusively on the basis of the acquired data.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] DEVICE AND METHOD FOR MATERIAL IDENTIFICATION FROM A SET OF MONOCHROMATIC IMAGES USING SPECTRAL COMPUTED TOMOGRAPHY

[0002] Technical Field

[0003] The field of technology to which the invention relates, generally speaking, belongs to the field of medical equipment, more specifically to the field of devices for examination of materials used for medical diagnostics and detection of contrast agents and detection of harmful substances, and in particular relates to a device and a method for identification of materials from a set of monochromatic images by spectral computed tomography.

[0004] According to the International Patent Classification (IPC), the subject matter of the invention is classified and designated by the basic classification symbol A61B 6 / 00, which relates to apparatus or devices for diagnostics and radiation, as well as by the secondary classification symbols A61B 6 / 03, which includes computed tomography, GOIN 21 / 17, which relates to testing or analysis of materials by systems in which incident light is modified depending on the properties of the material under examination, GOIN 23 / 046, which includes testing or analysis of materials using tomography, for example computed tomography, and GOIN 23 / 00, which includes testing or analysis of materials using wave radiation, for example X- rays.

[0005] Technical Problem

[0006] The technical problem solved by the present invention consists in the following: how to constructively design a device and define methods which enable identification of the chemical composition and density of materials from a set of two or more monochromatic images obtained directly by a monochromatic X-ray beam or indirectly from a set of virtual monochromatic images by means of spectral computed tomography (sCT), for the purpose of enabling precise medical diagnostics, which is widely applied in therapy planning and treatment, application of safety in scanning, precise calculation and deposition of doses, correction of X-ray absorption effects, non-destructive testing and detection of explosives, quality control of food, professional analysis of materials, study of pigments, coatings andstructural materials in historical artifacts without damage, such that the obtained data are of sufficient quality for the stated application and that such a device and applied method enable integration with existing hardware systems, are compatible with sCT devices of different technological constructions, particularly intended for identification of specific materials in detection of contrast agents or dangerous and harmful substances.

[0007] Background Art

[0008] The development of technology has been accelerated by the application of computers, which has significantly facilitated the solution of numerous problems, especially in medicine, and has contributed to significant progress in the efficiency of diagnostics and treatment, so that from the first images obtained by X-rays using invasive methods, both for doctors and for patients, there has been a transition to the application of new-generation devices in which harmful effects during operation are significantly reduced and significantly safer conditions are created during medical examinations and implementation of treatment. This, in addition to application in medicine, also relates to improvement of efficiency and working conditions in industry, military and police, as well as in the analysis of cultural heritage when defining differentiation and quantification of multiple materials.

[0009] In recent times, intensive work has been carried out on the construction of devices based on various technical solutions for processing data obtained by the method of spectral computed tomography with the application of artificial intelligence. Computed tomography (CT) is a diagnostic method which reconstructs a three-dimensional (3D) structure of an object from a set of X-ray projections recorded from different angles. Spectral computed tomography enables acquisition of data at different energy levels of X-rays using spectral capabilities of detectors or X-ray sources, or a combination of both. Some of the detectors with spectral capability are sandwich detectors (dual-layer detectors) and detectors with photon-counting technology. Monochromatic images within CT diagnostics represent reconstruction of an object from data obtained by monochromatic X-rays. Virtual monochromatic images are digitally reconstructed images from spectral data, which represent a prediction of an image that would be formed by monochromatic X-rays.CT systems intended for application in industry generate X-rays by means of a cathode tube, through interaction of electrons with a solid anode or an anode in the form of a jet stream. X-ray beams obtained by this method are polychromatic. Another method of generating X- rays is by means of synchrotron radiation, which arises from accelerated motion of charged particles. This method produces a large quantity of X-rays with a characteristic spectrum, which can then be filtered by various techniques in order to select a desired monochromatic energy. This approach can also be used for CT imaging, but involves high costs and technical challenges, and is still not applicable in clinical medicine. Monochromatic rays can also be obtained by decay of radioactive sources. sCT systems have been developed with the aim of improving differentiation and quantification of materials by X-ray imaging. Monochromatic images (virtual or native) provide better results in differentiation of materials, since differences in attenuation of X-rays between different materials are more pronounced at a single energy level than when attenuation is observed as an average across a certain energy range.

[0010] Examples of diagnostic devices include X-ray systems, magnetic resonance systems, ultrasound systems, computed tomography (CT) systems, positron emission tomography (PET) systems, ultrasound, nuclear medicine systems and other types of imaging systems. Common to all CT systems is an X-ray source that emits a beam toward the object being scanned, such as a patient, luggage, a security-risk object and the like. The beam, after being attenuated by the scanned object, reaches an array of radiation detectors. The intensity of the attenuated radiation of the beam received at the detector array depends on attenuation of the X-ray beam by the material under examination. Each detector element of the detector array produces a separate electrical signal indicating characteristics of the attenuated beam. The electrical signals are transmitted to a data processing system for analysis, which ultimately produces an image. Generally, the X-ray source and the detector array rotate around the opening of the gantry within the imaging plane and around the subject. X-ray sources are usually X-ray tubes that emit an X-ray beam from a focal point. X-ray detectors usually contain a collimator for collimating beams received at the detector.

[0011] For understanding the subject invention, within consideration of the state of the art, it is necessary to take into account the following:The basic method for interpretation of sCT data - material decomposition - was introduced by Alvarez and Macovski in 1976. Material decomposition may be applied to sCT data in the projection domain (before 3D reconstruction) or in the image domain (after 3D reconstruction). For two (or more) acquired series at different X-ray energies, attenuation of each point may be expressed as a linear combination of attenuations of two basis materials. A basis material represents a selected pair of physical or virtual materials of known composition. A common choice for decomposition of biological tissues is the basis pair polymethyl methacrylate and aluminum. By using estimated coefficients of basis materials obtained by decomposition and their known attenuation properties, it is possible to reconstruct virtual monochromatic images at any energy within the diagnostic range. Virtual monochromatic images have demonstrated clinical value in improving contrast and reducing artifacts in specific diagnostic tasks, and now represent a standard output in first clinical photon-counting CT scanners.

[0012] Material decomposition may be extended to other tasks such as calcium detection, separation of water and fat, separation of bone and soft tissues, K-edge differentiation and the like. The specific task depends on the selection of basis materials. For example, in separation of bone and soft tissues, one basis material represents bone tissue and the other soft tissue, thereby obtaining maps of bone and soft tissues. The most common example of K-edge imaging in medicine relates to quantification of iodine contrast agents. In that case, one of the basis materials is iodine, which exhibits abrupt changes in X-ray attenuation due to the presence of the K-shell absorption edge (K-edge) within the diagnostic energy range. The energy of the absorption edge corresponds to the binding energy of electrons in the K- shell of the atom. The weight associated with the iodine basis material in the decomposition process is proportional to the concentration of iodine in the patient. Very often there is a need for differentiation and quantification of multiple tissues and / or contrast agents. In analysis of cultural heritage, as well as in industrial and security applications, there is a need for differentiation and quantification of multiple materials. In such cases, the basis of decomposition may include more than two materials - this approach is known as multimaterial decomposition.A step closer to material identification is determination of material density and effective atomic number. A unique definition of effective atomic number does not exist in the literature, but it is generally accepted that the effective atomic number represents a type of averaged atomic numbers of the elements present in a material or mixture. The inventors have provided a comparative overview of multiple methods for estimation of the effective atomic number.

[0013] Density and effective atomic number may be obtained using the fundamental physical interactions of X-rays with matter, such as Compton scattering and the photoelectric effect, which depend on density and atomic number. The contribution of each physical mechanism may be separated by a procedure similar to material decomposition, wherein the basis materials are replaced by theoretical or empirical functions describing the energy dependence of the interactions. Alternatively, density and effective atomic number may also be obtained by a calibration approach. United States patent US 7697657 describes a system and method for obtaining density and effective atomic number from sCT data.

[0014] Extraction of full material composition by decomposition into physical or virtual materials, or by using theoretical basis functions describing X-ray interaction, requires well-separated spectral channels. In practical CT systems, limited energy resolution makes it difficult to acquire fully independent spectral data, which leads to instability in the decomposition process and amplification of noise. In order to overcome problems of an underdetermined or poorly conditioned system of equations, multi-material decomposition requires an equal number of spectral channels and basis materials. For a fixed radiation dose, increasing the number of spectral channels reduces the number of detected X-rays per channel. Since X-rays follow Poisson statistics, at a low number of photons, noise begins to dominate the signal, further degrading the signal-to-noise ratio. In addition, error propagation in decompositions with three or more materials follows a multistep dependency, which means that even small fluctuations in input values lead to exponential amplification of noise in the output. Finally, existing approaches often depend on detailed knowledge of CT system geometry and proprietary technical specifications, which limits flexibility of software solutions for material decomposition and prevents broader application without direct access to data controlled by the manufacturer.In addition to approaches based on material decomposition, material properties may also be obtained by calibration methods. An example is scanning of a calibration phantom composed of multiple known materials and creation of a calibration function that maps unknown materials to known ones. Another example is the use of public databases containing recorded attenuation values, where the measured value is mapped to the closest one in the database. By this approach, the need for physical scanning of a large number of materials is avoided, but it requires a very precise understanding of detector response and the shape of the polychromatic X-ray spectrum. Many databases contain full material composition and density data along with attenuation values, which enables access to identified materials. Identification of materials from a set of predefined materials serving as a reference table is known as dictionary-based material identification.

[0015] Monochromatic images (virtual or native), by definition, provide attenuation values independent of the scanner and method, and therefore knowledge of detector response or X-ray spectrum is not required. Wohlfahrt et al. provided an overview of current approaches to material identification in the literature, particularly for application in radiotherapy. For sCT, the most represented methods rely on calibration that maps virtual monochromatic images to material density or effective atomic number. Multiple virtual monochromatic images may be useful because certain energies provide better contrast for precise definition of critical structures, while others minimize artifacts resulting from preferential attenuation of the beam (beam hardening) in the case of a polychromatic spectrum. The selection of energy level is therefore specific for each type of imaging. In addition, many materials have identical or very similar attenuation values at certain energies, which further complicates correct material characterization. United States patent US 8787519 describes a method for optimization of display of dual-energy CT images, which essentially includes optimal selection of virtual monochromatic images for specific diagnostic tasks.

[0016] It should be emphasized that, despite careful and detailed searching of patent and scientific literature in the relevant field, no devices have been found that use a set of two or more (virtual or native) monochromatic CT images for estimation of full composition and density of materials of an object. Compared to previous solutions, the technical solution according to the subject application provides for the use of all available information across the energyrange of the imaging system. If a monochromatic source is not available, the method applies traditional material decomposition for estimation of virtual monochromatic images. Since monochromatic datasets are independent of the imaging system, their use as input data enables simple integration with existing sCT systems. However, characterization of materials from two or more monochromatic datasets may still be challenging due to the previously stated reasons, particularly in the case of virtual images where quantitative accuracy depends on the quality of decomposition. Full characterization requires multiple energy channels, usually equal to the number of different elements composing the compounds in the object of interest. The present application includes a hybrid and industry-specific solution reflected in the selection of a reference table (lookup table). This means that different sets of materials may be used for diagnostic imaging, radiotherapy, security screening or cultural heritage analysis, which will be described in detail later in the invention.

[0017] It should be noted that all previous solutions enabled only partial quantitative evaluation of spatial spectral data, wherein the obtained data were not of sufficient quality for many applications, with difficult or impossible integration with existing imaging systems. In contrast, the present invention enables full material identification with significantly improved image quality and the possibility of integration with existing hardware systems. By review of available patent documentation and other technical documentation from the relevant field, as well as available technical and promotional literature, the following has been found:

[0018] United States patent US10898727 B2, granted on 26 January 2021, relates to a method for obtaining various data suitable for use in planning radiation of patients. This method is based on measurements performed using computed tomography CT devices with a quantum X-ray detector in order to generate results, taking into account the specific characteristics of such obtained data. The results obtained are forwarded to an interface through which they are used for planning patient irradiation. From review and comparison of this patent and the subject invention, the basic difference is observed in that this US patent uses a quantum detector.In United States patent US7697657 B2, granted on 13 April 2010, a system and method for imaging density and effective atomic number is presented, comprising a high-frequency electromagnetic energy source emitting a beam of high-frequency electromagnetic energy toward an object being scanned, a detector receiving attenuated electromagnetic high- frequency energy weakened by the object, a data acquisition system (DAS) operatively connected with the detector and a computer operatively connected with the DAS and programmed to: acquire projection data from the detector during acquisition of projection data, generate a density image for an unknown material (based on two or more known basis materials), generate an effective atomic number (Z) based on two or more known basis materials and on an arctg function of the ratio of atomic numbers of two or more known materials, and store density and effective material images on a computer-readable storage medium. The computer according to this invention is programmed to generate a density image for an unknown material based on calibration of two or more known basis materials and to generate an effective atomic number (Z) for an unknown material based on calibration of two or more known basis materials and on an arctg function of the ratio of atomic numbers of multiple known materials and to forward images for storage on a computer-readable storage medium. By comparing schematic representations, it is observed that the disclosed devices and the subject device have different constructions.

[0019] In United States patent US12201851 B2, granted on 21 January 2025, a method for obtaining a treatment plan determining irradiation of a patient is presented, wherein the method comprises: a dose inference phase, including use of a model for inference of spatial dose based on patient data; a dose mimicking phase, including execution of a complex optimization process for forming a treatment plan in accordance with the inferred spatial dose, wherein the complex optimization considers multiple scenarios relating to uncertainty of patient data.

[0020] In European patent EP38O6747 Bl, published on 3 July 2024, an invention entitled a method for X-ray imaging using dual energy is presented, comprising obtaining a first and a second image of material, wherein the first image includes one density obtained by one energy beam, while the second image includes another density obtained by another energy beam, combining the first and second image to generate a combined image such that a techniquefactor is obtained around which tissue thickness is determined and volumetric data are recorded. Dual X-ray images may also be generated by combining images and multiple energies of materials of the subject with weighting factor values determined based on thickness of different tissue types corresponding to the recorded material. Different weighting factor values may be applied to different pixels of higher or lower energy images. Weighting factor values may, for example, be retrieved from memory, calculated, interpolated or extrapolated. Images of higher and lower energy and weighting factor values may be linearly combined to generate a dual-energy X-ray image. In some cases, weighting factor values may increase material contrast. From the above, it is evident that the methods according to this invention differ from the methods presented in the subject invention.

[0021] In United States patent US8787519 B2, granted on 22 July 2014, a system and method for optimizing display of sCT images is presented, comprising an X-ray source forming an X-ray beam toward a material being scanned, a detector receiving attenuated rays, a data acquisition system (DAS) operatively connected with the detector and a computer. The computer is programmed to receive data from two or more spectra of incident energy, decompose the obtained CT scanner data from two or more spectra of incident energy, decompose the obtained scanned data into projection CT data of a first basis material and a second basis material, generating images which form a final monochromatic image with reduced noise. The obtained image is then transmitted to a computer-readable medium for further storage.

[0022] In European patent EP2052279 Bl, published on 11 September 2019, entitled a device and method of computed tomography, a device is presented comprising a detector element sensitive to radiation and providing output data indicating detected radiation at at least a first and second energy level or energy range. Photon-counting detectors with energy resolution further classify detector signals according to their corresponding energies. Correctors correct classified signals and a combiner compares signals according to a combined function to generate outputs (EL, EH) indicating radiation detected at at least a first or second energy level or range. By careful review of drawings of this and the subject invention, it is easily observed that they represent different conceptual solutions.In United States patent US9354331 B2, granted on 31 May 2016, entitled an X-ray computed tomography device and radiation detector, a device is presented comprising an X-ray tube, a detector and two DAS devices for data acquisition. The radiation detector contains multiple detection elements. Each detection element repeatedly detects X-rays generated by the X- ray tube and transmitted through a subject and repeatedly generates an electrical signal corresponding to detected X-rays. A first DAS collects electrical signals detecting a part of the imaging region of each detection element in an integral mode, while a second DAS collects electrical signals detecting another part of the imaging region of each detection element in a photon-counting mode. From this it is evident that the methods according to this US patent and the subject device are different.

[0023] Disclosure Of Invention

[0024] The essence of the invention is reflected in that, in accordance with the inventive idea, a device is constructively designed which comprises: an X-ray source emitting an X-ray beam toward the object being scanned, a detector receiving X-rays attenuated by passing through the object, a control device operatively connected with the detector and a computer and a cloud, which are configured such that, according to a corresponding method, they enable the use of a predefined set of known materials for identification, instead of attempting to calculate quantitative properties of the material exclusively on the basis of the acquired data.

[0025] By the device and method according to the invention, it is achieved that the use of a defined set of materials has a dual purpose:

[0026] 1. providing a certain form of regulation and adaptation of the material identification problem to a specific industrial field;

[0027] 2. enabling prediction of the full composition of the material and its density, with the possibility of calculating a wide range of material properties useful in various industrial domains.

[0028] The essence of the invention also lies in that the sCT device, implemented according to the invention, acquires data from two or more energy spectra or monochromatic energies,wherein it uses two or more detectors or one detector with internal spectral capabilities (for example a photon-counting detector), wherein the sCT device may also represent any combination of the stated detectors and sources and provides for the use of all available information across the energy range of the imaging system.

[0029] The essence of the invention further lies in that the computer is programmed such that control of the data acquisition process is performed via the control device, wherein identification and generation of results about the material are performed from a set of real or virtual monochromatic images.

[0030] The novelty of the invention is also reflected in that material identification is preferably performed using a set of expected materials, by application of a physical model or machine learning (ML) and artificial intelligence (Al) according to the principle of deep learning, informed by a physical solution.

[0031] The essence of the invention also lies in that information on the level of confidence of material identification is provided and adjusted via input controls, wherein the computer is operatively connected with a cloud platform or, for improved data processing, the cloud is directly connected with the control device.

[0032] The essence of the invention further lies in that the obtained results are suitable for application in planning of radiotherapy of patients, multi-contrast diagnostic visualization, identification of materials in medical diagnostics, archaeometry, quality control and security inspections.

[0033] The essence of the invention also lies in that it enables full characterization of materials which was not previously achievable by material decomposition procedures due to the requirement for a large number of energy channels and radical amplification of noise during the identification process. Owing to the introduction of spectral information, which may comprise only two independent energy channels, the method presented in the description of this invention surpasses the accuracy of material identification in comparison with methods intended for operation on non-spectral data sets.

[0034] The essence of the invention further lies in that the subject device can be technically implemented as a fully integrated system in which the control device, computer and cloudsystem represent an independent "add-on" module that is upgraded onto existing sCT systems.

[0035] Brief Description Of Drawings

[0036] For the purpose of easier understanding of the invention, the inventors, by way of example only, refer to the accompanying drawings of the application, wherein:

[0037] - Figure 1 shows a schematic representation of the device for identification of materials from a set of monochromatic images by spectral computed tomography with a control device and a computer connected to a cloud;

[0038] - Figure 2 shows a schematic representation of the steps in the operation process of the subject device, from acquisition of spectral data to identification of materials; - Figure 3 shows a diagram of information which, according to the subject method, is used for identification of materials.

[0039] Detailed Description of the Invention

[0040] By the present invention, the above-defined technical problem is fully solved.

[0041] As can be easily observed by viewing Figure 1, the device 1, according to this invention, comprises: an X-ray source 2 emitting an X-ray beam 3 toward an object 4 being scanned, a detector 5 receiving X-rays 6 attenuated by passing through the object 4, a computer 8 operatively connected with a control device 7 and a cloud 9, which are configured so as to enable the use of a predefined set of known materials for identification, instead of calculating quantitative properties of the material exclusively on the basis of the acquired data. The described invention can be implemented such that, for the purpose of improving data processing, the control device 7, computer 8 and cloud 9 represent an independent "add-on" module.

[0042] The invention includes a method of using an sCT system for spatial identification of materials on the basis of a set of known materials. It should be emphasized that the invention is applicable to sCT systems of different technological constructions and purposes, provided that they enable acquisition of spectrally and spatially resolved data. A unique characteristicof the material identification method, given in the description of this invention, is its integrability with multiple different sCT systems. This integrability arises from the fact that the monochromatic data set must contain quantitatively accurate, energy-dependent values for each type of material.

[0043] The method according to the subject invention is applicable in the sinogram space, prior to image reconstruction, or after reconstruction in 3D space. Units in sinogram space represent line integrals of attenuation coefficients, whereas in 3D reconstruction the data are expressed in units of linear attenuation coefficients or in Hounsfield units.

[0044] Depending on the technical characteristics of the device 1, the monochromatic data set is obtained by using a monochromatic X-ray source. The subject method is also broadly applicable to devices that do not acquire direct monochromatic data, but perform spectral separation within a certain energy range. For such, more commonly used devices, an additional data preparation step is required in which virtual monochromatic images are generated by a material decomposition procedure or other relevant methods known to experts in the field.

[0045] In addition to improved integration, this method differs from prior solutions in that it uses a predefined set of known materials for identification, instead of attempting to calculate quantitative properties of materials exclusively from acquired data. The use of a defined set of materials has a dual purpose:

[0046] 1. it provides a certain form of regularization and adaptation of the material identification problem in a specific industrial field,

[0047] 2. it enables prediction of the full composition of the material and its density.

[0048] This second characteristic enables calculation of a wide range of material properties useful in various industrial domains.

[0049] The operation method of the subject device is most easily understood by observing the schematic representation of the steps in the operation process from acquisition of spectral data to identification of the examined material shown in Figure 2 of the drawings of the invention. From this figure it is observed that the operation process of the device 1comprises: acquisition 10 of spectral CT data, reconstruction 11 of two or more monochromatic images, identification 12 of materials from a list of known materials for each spatially resolved data point, on the basis of which data are obtained on: electron density 13, mass density 14, material concentration 15, stopping power 16 and material elemental composition 17.

[0050] The method of operation of the subject device is shown in Figure 3, through a set of predefined, expected materials with dashed lines representing their known dependence of attenuation on energy. From this figure it is easily observed that the points indicate that, instead of continuous values, only certain well-defined energies may be used. It is also evident that some materials may have very similar or identical attenuation values at a certain energy level, but the differences increase at other energies. This indicates the need for spectral measurement for accurate material identification. The points on the solid line represent measured monochromatic values for each measured spatial point. Essentially, the method for material identification operates on the principle of finding the minimum mathematical distance between the measured monochromatic values of an unknown material and the expected values of known materials from the defined set.

[0051] The method comprises combinations of two or more attenuation curves from the predefined set of materials, in order to achieve a higher level of matching with the measured monochromatic values.

[0052] The method also comprises the use of information of technical solutions according to which specific weights may be applied to the set of monochromatic data, for the purpose of optimizing input data for material detection, with the use of any analytical or artificial intelligence-based algorithm for estimation of the importance of each energy level from the set.

[0053] The method comprises the application of various analytical solutions aimed at minimizing the mathematical distance between the data of the unknown material and the set of known materials. In addition, it is possible to combine multiple materials from the defined set in order to further reduce the distance, thereby creating a virtual material with a composition and density proportional to the mixing ratio. A method of promoting sparsity of the solutionis particularly emphasized, which essentially minimizes the number of selected materials from the predefined set of materials used to describe the measured monochromatic values. An advantage of this method is that, for the purpose of improving the identification process, it is not limited to processing only one spatial point at a given moment, but also uses information about neighboring points (for example neighboring voxels). This enables the application of machine learning and artificial intelligence algorithms as support in material identification.

[0054] In addition to the use of a predefined set of materials for limiting the problem to a specific industrial application, the method according to the invention is applicable to an industryspecific data-driven approach for supporting material identification by artificial intelligence. The subject device is constructed to assist in identification of materials from a predefined set by training on multiple industry-specific data (for example patient scans, luggage at airports, etc.). A particular characteristic of the device described in this invention is that it is not exclusively data-driven, but also includes physical information during the training process, wherein it is limited to identification of materials from a predefined set of data.

[0055] Physical information is introduced into the training process in the following manner: the device uses as input data from the sCT system (input spectral data) and attempts to identify one or more materials from the predefined set of materials. When the device performs an estimation, an output set of spectral data is mathematically generated based on the composition and density of the assumed material. The output spectral data are calculated either exactly based on known properties or by using experimentally measured values. This information is then compared with the input spectral data through various known loss functions, such as Ll-loss, L2-loss, cross-entropy, dice loss and others.

[0056] The number of output spectral data is equal to or less than the number of input spectral channels. Individual channels are specifically weighted in order to contribute to the overall loss function. The overall loss function also includes other penalty factors that are not based exclusively on the difference between input and output spectral data.

[0057] The advantage of using an artificial intelligence model for performing or supporting material identification lies in the strength that the model acquires by training on large datasetsspecific to a particular industry. For example, in medical diagnostics, numerous models are designed for organ segmentation. This information may be incorporated into the present invention as a weighting factor in the material identification algorithm, particularly in cases where it is necessary to distinguish tissues with similar attenuation properties at different energy levels.

[0058] Regardless of the above-described method of material identification, the present invention also provides statistical information on the level of confidence of the performed identification. This is particularly useful when searching for specific types of materials within the examined object - for example detection of contrast agents in patients or detection of dangerous and harmful substances. Information on the level of confidence of material identification is, according to the invention, provided and adjusted via input controls.

[0059] The statistical approach may be in the form of a linear or nonlinear model, but also all other statistical methods of machine learning and artificial intelligence. The problem may be defined through a design matrix containing the main regressors, i.e., materials of interest defined in a predefined material file, as well as nuisance regressors which may also be used in the identification process. Nuisance regressors may represent other materials from the defined set, motion artifacts, slight signal shifts or any other factors that may affect the identification result.

[0060] Confidence intervals for ML or Al methods described earlier may be indirectly applied using techniques such as permutation testing, for the purpose of assigning significance levels to individual regions or patterns.

[0061] By this invention, a significant problem of noise present in many previous solutions is solved. Namely, full characterization of materials is enabled, which was not previously achievable by material decomposition procedures due to the requirement for a large number of energy channels and radical amplification of noise during the process. On the other hand, owing to the introduction of spectral information, which may comprise only two independent energy channels, the method presented in the description of this invention surpasses the accuracy of material identification compared to methods intended for operation on non-spectral data sets.Such a device provides complete information about the examined material, which enables broad application including, but not limited to:

[0062] - calcium detection (calcium scoring): quantification of calcium deposits in blood vessels, which represents a key indicator for assessment of cardiovascular risk;

[0063] - fat-water separation: critical in musculoskeletal and metabolic imaging for distinguishing fat infiltration in organs or tissues;

[0064] - separation of bone and soft tissues: assists in orthopedic diagnostics and surgical planning by improving contrast between different tissue types;

[0065] - contrast differentiation (K-edge differentiation): enables differentiation of materials with similar attenuation properties using their K-edge absorption characteristics, such as iodine, calcium or gadolinium, which is particularly important in contrast-enhanced medical diagnostics;

[0066] - effective atomic number: differentiates materials based on their atomic composition, improving tissue classification and application in security scanning;

[0067] - electron or mass density: useful in radiotherapy for precise dose calculation and in structural analysis of materials;

[0068] - absolute and relative stopping power: essential for planning proton and heavy ion therapy, ensuring precise dose deposition in particle therapy;

[0069] - Monte Carlo simulations: useful for estimating dose distribution in radiotherapy during treatment planning;

[0070] - attenuation correction: improves quantitative accuracy in positron emission tomography (PET) and single-photon emission computed tomography (SPECT) by correcting X-ray absorption effects;

[0071] - industrial and security applications: enable non-destructive testing, detection of explosives and food quality control by identifying specific material compositions;

[0072] - cultural heritage analysis: facilitates study of pigments, coatings and structural materials in historical artifacts without damage.

[0073] The subject invention is not limited only to imaging devices based on computed tomography, but may be applied to any imaging system that emits radiation, as well as to all other devicesthat enable acquisition of information from at least two different imaging channels for obtaining a representation. Such channels may be spectral channels of X-ray attenuation, phase shift, scattering and other interactions of X-rays with the object. Energy sources may also include ultrasound devices, magnetic field devices (MRI), or devices based on nuclear emitters.

[0074] The application of the subject invention is not limited exclusively to systems that acquire two or more energy levels, but also includes all systems that can measure two or more independent pieces of information for a single spatial point. In the case of X-ray imaging, these may be information on attenuation, phase shift, X-ray scattering from the object, or any combination of these parameters.

[0075] Industrial Or Other Application Of The Invention

[0076] Industrial production of the subject invention is fully possible in factories for manufacturing computers and sophisticated medical equipment. The device shown in the invention may be manufactured according to documentation which experts in the field can easily produce using the description and drawings from this application.

[0077] The invention is suitable for mass production and its application, in addition to diagnostics, therapy planning and treatment in medicine, is particularly recommended in the field of security scanning, precise calculation and deposition of doses, correction of X-ray absorption effects, non-destructive testing and detection of explosives, food quality control, professional material analysis, and study of pigments, coatings and structural materials in historical artifacts without damage.

[0078] Testing on a prototype has shown that the subject device and method enable integration with existing hardware systems, that they are compatible with sCT devices of different technological constructions, particularly intended for identification of specific materials in detection of contrast agents or dangerous and harmful substances.

[0079] For the applicant

[0080] Gordana Primovic, attorney at law

Claims

PATENT CLAIMS1. A device for identification of materials from a set of monochromatic images by spectral computed tomography, comprising: an X-ray source (2) emitting an X-ray beam (3) toward an object (4) being scanned, a detector (5) receiving X-rays (6) attenuated by passing through the object (4), a control device (7) and a computer (8) operatively connected with the device (1) and a cloud (9), characterized in that the computer (8) of the device (1) is programmed to use a predefined set of known materials for identification instead of calculating quantitative properties of the examined material.

2. The device for identification of materials from a set of monochromatic images by spectral computed tomography according to claim 1, characterized in that the computer (8) is configured such that control of the data acquisition process is performed via the control device (7).

3. The device for identification of materials from a set of monochromatic images by spectral computed tomography according to claim 1, characterized in that the device (1) uses radiation sources for generating an Al-based image.

4. The device for identification of materials from a set of monochromatic images by spectral computed tomography according to claim 1, characterized in that the computer (8) is programmed to perform material identification using a set of expected materials, by applying a physical model or machine learning and artificial intelligence according to the principle of deep learning.

5. The device for identification of materials from a set of monochromatic images by spectral computed tomography according to claim 1, characterized in that the computer (8) is programmed to provide and adjust information on the level of confidence of material identification via input controls.

6. The device for identification of materials from a set of monochromatic images by spectral computed tomography according to claim 1, characterized in that the device (1) is configured to acquire data from two or more energy spectra or monochromatic energies,and / or uses two or more detectors or one detector with internal spectral capabilities, and / or any combination of the stated detectors and sources.

7. The device for identification of materials from a set of monochromatic images by spectral computed tomography according to claim 1, characterized in that the computer (8) is programmed to perform identification of materials from a set of real or virtual monochromatic images and to generate results.

8. The device for identification of materials from a set of monochromatic images by spectral computed tomography according to claim 1, characterized in that the device (1) is configured to use nuclear emitters for generating an Al-based image.

9. The device for identification of materials from a set of monochromatic images by spectral computed tomography according to claim 1, characterized in that the device (1) is configured to use magnetic emitters for generating an Al-based image.

10. The device for identification of materials from a set of monochromatic images by spectral computed tomography according to claim 1, characterized in that the device (1) is configured to use ultrasound for generating an Al-based image.

11. A method for identification of materials from a set of monochromatic images by spectral computed tomography on a device (1) according to claim 1, characterized in that it comprises acquisition (10) of spectral CT data, reconstruction (11) of two or more monochromatic images, identification (12) of materials from a list of known materials for each spatially resolved data point, on the basis of which data are generated on: electron density (13), mass density (14), material concentration (15), stopping power (16) and material elemental composition (17), wherein, for the purpose of radical noise reduction in identification of the examined material, spectral information from at least two independent energy channels is used.

12. The method for identification of materials from a set of monochromatic images by spectral computed tomography according to claim 11, characterized in that the process of material identification also comprises information about surrounding points, for example neighboring voxels.

13. The method for identification of materials from a set of monochromatic images by spectral computed tomography according to claim 11, characterized in that it enables finding the minimum mathematical distance between measured monochromatic values of an unknown material and expected values of known materials from a defined set.

14. The method for identification of materials from a set of monochromatic images by spectral computed tomography according to claim 11, characterized in that it uses at least two different energy levels for generating an Al-based image.

15. Use of the device and method for identification of materials from a set of monochromatic images by spectral computed tomography according to claims 1 and 11, characterized in that information from the device (1) is used for planning radiotherapy in cancer.

16. Use of the device and method for identification of materials from a set of monochromatic images by spectral computed tomography according to claims 1 and 11, characterized in that the device (1) uses the image for industrial testing of materials.

17. Use of the device and method for identification of materials from a set of monochromatic images by spectral computed tomography according to claims 1 and 11, characterized in that the device (1) uses the image for military testing of materials.

18. Use of the device and method for identification of materials from a set of monochromatic images by spectral computed tomography according to claims 1 and 11, characterized in that radiotherapy is particle-based, for example using protons, carbon nuclei, helium and others.For the applicantGordana Primovic, attorney at law