Generating an estimate of patient radiation dose resulting from medical imaging scans

0001Embodiments of the invention generally cover approaches for predicting patient radiation exposure during computed tomography (CT) scans.

0002As is well known, CT scan systems use ionizing radiation (X-rays) to generate images of tissues, organs and other structures in the body. The X-ray data resulting from a CT scan can be converted into an image on a computer display screen. For example, a CT scan provides a collection of data used to create a three-dimensional (3D) volume that corresponds to a scanned portion of a patient's body. The 3D volume is then sliced at small intervals along the axis of the patient's body to create an image of body tissue. Such slices may include both lateral and transverse slices (as well as other slices), depending on the tissue or structure imaged.

0003The use of CT scans and ionizing radiation for medical imaging has increased dramatically over the last decade. In addition, modern technologies such as CT scans provide much more detailed and valuable diagnostic information than traditional radiography. At the same time, however, the patient is exposed to a significant amount of radiation. For example, a typical chest CT will have a dose 100-250 times that of a conventional chest x-ray, depending on the voltage and current of the CT scanning system, the protocol to follow to perform the procedure, and the size and shape of the patient being scanned. Expose the patient.

0004Despite the increased use of CT scans (and the resulting exposure to radiation), the radiation dose that the patient is exposed to during the procedure, and more importantly, the cumulative dose over numerous procedures, is one person. It is not a parameter that is regularly tracked for the patient, nor is these parameter an easily accessible part of the patient's medical record. This is partly because the radiation dose absorbed by organs and tissues in the body cannot be measured directly in living patients as part of a CT scan, and the results obtained from the corpse are more accurate, This is because it does not correspond very well to the dose absorption of living tissues.

0005Similarly, the currently used approaches to dose prediction also provide inaccurate results. For example, one approach is to rely on a limited number of anthropometric phantoms to represent a given patient. However, the available imaging phantoms do not adequately represent wide variations in the size and weight of people in the population undergoing CT scans. As a result, single-point surface measurements are currently being performed in most of the cases where even the slightest dose is predicted. However, depending on where the single point dose is measured, this can lead to poor quality and highly variable results. More generally, surface measurements of radiation exposure do not provide an accurate measurement of the actual absorption of tissues, organs and structures in the body.<u style="single">Prior art document information related to the invention of this application includes the following (including documents cited at the international stage after the international filing date and documents cited when domestically transferred to another country).</u><u style="single">(Prior art document)</u><u style="single">(Patent document)</u><u style="single"> (Patent Document 1) U.S. Patent Application Publication No. 2008/0298540</u><u style="single"> (Patent Document 2) US Pat. No. 5,844,241</u><u style="single"> (Patent Document 3) U.S. Pat. No. 6,345,112</u><u style="single"> (Patent Document 4) U.S. Patent Application Publication No. 2001/0027262</u><u style="single"> (Patent Document 5) U.S. Patent Application Publication No. 2010/02888916</u><u style="single"> (Patent Document 6) U.S. Patent Application Publication No. 2008/0292055</u><u style="single"> (Patent Document 7) U.S. Pat. No. 5,341,292.</u><u style="single"> (Patent Document 8) U.S. Patent Application Publication No. 2008/0089569</u>

<p num="0006"> The embodiment provides a technique for predicting radiation exposure of a patient during a computed tomography (CT) scan. One embodiment includes a computer implementation method for creating a personalized imaging model. This method generally selects an initial imaging phantom for an individual undergoing an image scan (the imaging phantom has one or more associated localized images) and receives one or more scout images of the individual. May include doing. The method may further include determining the transformation between one or more localized images associated with the imaging phantom, and deforming the initial imaging phantom based on the transformation.</p><p num="0007"> In certain embodiments, the imaging scan is a computed tomography (CT) scan, and in other cases, the imaging scan is a fluoroscopic scan, a PET scan, an angiographic scan, and the like. This method receives a set of parameters that describe the CT scanning device used to perform imaging scans and CT scans, a modified imaging phantom and an imaging scan using the received set of parameters. And, based on the simulation, predicting the radiation dose absorbed by an individual as a result of performing an imaging scan. In certain embodiments, the simulation is a Monte Carlo simulation.</p><p num="0008"> Another embodiment includes a method for creating an imaging model for an individual. This method selects an initial imaging phantom for an individual undergoing computed tomography, and divides the reference CT scan associated with that individual into multiple dissections of that individual present in the reference CT scan. Identifying a three-dimensional (3D) volume of a scientific landmark can generally be included. This method matches one or more identified anatomical landmarks on a split reference CT scan with the corresponding anatomical landmarks on the initial imaging phantom, and the matched anatomical landmarks. Deformation of the initial imaging phantom based on the mark may also be included.</p><p num="0009"> An additional embodiment implements a computer-readable storage medium storing an application that performs the above-mentioned method when executed on a processor, and a system having the processor, and the above-mentioned method when executed on the processor. Includes memory for storing corporate information asset management application programs.</p>

0010The method by which the above aspects are achieved can be understood in detail and a more specific description of the embodiments of the invention briefly summarized above can be made by reference to the accompanying drawings. However, it should be noted that the accompanying drawings exemplify only typical embodiment inventions and therefore do not limit their scope as the present invention may recognize other similarly effective embodiments.<figref num="1">FIG. 1 illustrates an example of a CT scan system and a related computing system configured to provide prediction of patient radiation dose according to one embodiment of the present invention.</figref><figref num="2">FIG. 2 illustrates an example of an imaging system used to acquire CT scan data according to one embodiment.</figref><figref num="3">FIG. 3 illustrates an example of a dose prediction system used for predicting and tracking cumulative patient radiation doses according to one embodiment.</figref><figref num="4">FIG. 4 illustrates a method for creating a suitable model for predicting a patient's radiation dose resulting from a CT scan, according to one embodiment.</figref><figref num="5A">FIG. 5A illustrates an exemplary image representing a deformable phantom according to one embodiment.</figref><figref num="5B">FIG. 5B illustrates an example of a two-dimensional (2D) reference image of a portion of the human body corresponding to the phantom shown in FIG. 5A, according to one embodiment.</figref><figref num="6">FIG. 6 illustrates another method for creating a suitable model for predicting the patient's radiation dose resulting from a CT scan, according to one embodiment.</figref><figref num="7">FIG. 7 illustrates an example of a phantom slice superimposed on a patient's corresponding CT slice according to one embodiment.</figref><figref num="8">FIG. 8 illustrates an example of a transverse slice of an imaging phantom superimposed on a patient's corresponding transverse CT slice according to one embodiment.</figref><figref num="9">FIG. 9 illustrates an example of the displacement of the organ volume of the CT image division and imaging phantom according to one embodiment.</figref><figref num="10">FIG. 10 illustrates a method of dose prediction service for providing patient dose prediction to multiple CT scan providers according to one embodiment.</figref><figref num="11">FIG. 11 illustrates an example of the computing infrastructure of a patient dose prediction service system configured to support multiple CT scan providers in one embodiment.</figref>

0011Embodiments of the present invention generally cover approaches for predicting radiation exposure of a patient during a computed tomography (CT) scan. More specifically, embodiments of the present invention predict patient dose by interpolating the results of multiple simulations, an efficient approach for creating appropriate patient models used to make such predictions. a to approach, and a plurality of CT scans provider provides an approach for the service provider that hosts the available dose prediction service. As detailed below, dose management systems provide a single system for tracking radiation doses across modality and presenting information to physicians in a meaningful and easily understandable format. Routine consideration of cumulative doses when directing diagnostic imaging tests leads to a more informed decision process, ultimately benefiting patient safety and care.

0012In one embodiment, a virtual imaging phantom is created to model a given patient undergoing a CT scan. Virtual imaging phantoms are created by modifying existing mathematical phantoms to better match the size, shape, and / or organ location of patients exposed to CT scan radiation. First, the mathematical phantom can be selected, for example, based on the patient's age and gender. A patient-specific geometry can be achieved by transforming the selected mathematical phantom using the transformations obtained by analyzing the patient's scout image localizer. In this context, as will be appreciated by those skilled in the art, "localizer" generally refers to a patient's 2D imaging (typically anterior / posterior x-ray and / or lateral x-ray). In such an approach, the selected math phantom may have its own set of localized images. A reference image for a given virtual phantom is selected from images obtained from multiple individuals, selected to match the geometry, size and position of the phantom (eg, with arms raised or on the side of the body). Can be selected.

0013Image registration techniques are then used to map points in the patient's localized image to points in the reference image associated with the virtual phantom. This results in a series of transformations that can be used to deform the virtual phantom to better match the patient's geometry. A similar approach uses a reference set of 3D data (selected CT scans) for the phantom, and uses 3D image registration techniques to relate the points of the CT scan of a given patient to a given phantom. Accompanied by mapping to points in the reference CT scan.

0014Similarly, image segmentation can be used to identify the 3D volume during a CT scan that corresponds to the organ, tissue, or structure of interest in the patient's CT scan. The 3D volume can be a more accurate 3D volume that is thought to represent a bounding box, or organ, etc. Once identified, the displacement is determined between the location of the organ in the phantom and the corresponding location of the patient's CT scan. Instead of targeting individual image points (as in 2D / 3D image registration techniques), the image splitting approach is from CT images as data points to determine transformations from virtual phantoms and given patients. Works by using larger 3D volumes.

0015In each of these examples, the resulting hybrid phantom provides a much more accurate mathematical representation for a particular patient for use in dose simulation than the unmodified phantom alone. Once the transformation is determined, a hybrid virtual phantom can be used to simulate a given CT procedure for the patient. For example, well-known Monte Carlo simulation techniques have been developed to predict organ absorbed doses in virtual phantoms. Such simulation techniques, in addition to the virtual phantom (after conversion for a given patient), have many settings related to the CT scanner model and the procedures performed to calculate accurate predictions of organ absorbed dose. To use. For example, CT scanners include kVp (ie peak tube voltage), X-ray generator target angle, fan angle, geometry, slice thickness, focus-to-axis distance, flat filter (material and thickness), and beam shaping. Can be modeled using filters (materials and geometry). Of course, these (and other parameters) can be selected when available or as needed to meet the needs of a particular example.

0016However, predicting organ-absorbing organ doses using Monte Carlo simulations can require a significant amount of computational time, much longer than the time required to perform an actual CT scan. Given the high utilization of CT scans in many imaging facilities, this delay is not manageable if the total cumulative dose prediction does not exceed a given maximum. Even if the predictions are not used before performing the procedure, the simulation is currently being performed unless the patient dose prediction can be determined in a relatively equal amount of time to perform the procedure. Maintaining a record of dose predictions for a given scanning system becomes cumbersome as it becomes increasingly late from the scan. This problem is dramatically greater for SaaS providers who host dose prediction services in the cloud for multiple imaging facilities.

0017Thus, in one embodiment, the patient dose prediction determined for a given procedure can be generated by interpolating between two (or more) previously completed simulations. If a "close" simulation is not available, then hybrid virtual phantoms, CT scanners and procedural data can be added to the queue for the full Monte Carlo simulation to be performed. Over time, a large library of simulations makes it possible to provide dose predictions in real time as procedures are scheduled and implemented. By doing so, it is possible to capture the cumulative dose of a given patient and observe the cumulative dose limit.

0018Further, in one embodiment, software as a service (SaaS) or cloud provider model may be used to perform dose prediction, maintain a library of calculated simulations, and perform Monte Carlo simulations. In such cases, the CT scan provider may provide the SaaS provider with the parameters of the prescribed CT procedure. For example, to provide SaaS providers with the virtual phantoms of their choice, as well as the transformations used to create hybrid phantoms tailored to the equipment and protocols used to perform specific individuals and CT procedures. Client software (or even a secure web-based portal) can be used. Once received, the service provider can select and interpolate the appropriate simulation from the library and return the prediction of the patient's absorbed dose to the imaging center.

0019Importantly, SaaS providers do not need to receive actual identification information about a given individual or patient undergoing a CT scan. Instead, the SaaS provider receives only information related to virtual phantoms and CT systems / procedures. As a result, the service provider's business may not require compliance with various laws and / or regulations relating to the privacy of personal health information. In addition, by providing dose predictions to multiple imaging centers, the resulting simulation library is more diversified and interpolated than a simulation library created solely from scanning procedures performed at a single imaging center. It is much more likely that you will find a candidate for it. Furthermore, centralization of simulation libraries and Monte Carlo simulations will enable improvements in phantoms, Monte Carlo simulation engines, and interpolation techniques shared by all imaging centers using cloud-based services. Finally, this approach leaves the imaging center to maintain the information that links the cumulative dose to a particular patient. Therefore, the actual patient data can be limited to each individual provider. At the same time, of course, SaaS providers include, for example, digital Imaging and Communications in Medicine (DICOM), Picture Archiving and Communication Systems (PACS), Health Level Seven International (HL7) standards, ICD-9, ICD-10 diagnostics and procedure codes, etc. Various standard protocols for image and data exchange, including, can be used to communicate with the imaging center.

0020Further, the following description refers to embodiments of the present invention. However, it should be understood that the present invention is not limited by any particular descriptive embodiment. Instead, any combination of the following functions and elements, whether related to different embodiments or not, is contemplated for the practice and practice of the present invention. Moreover, embodiments of the invention may achieve advantages over other possible solutions and / or prior art, but whether certain advantages are achieved by a given embodiment does not limit the invention. .. Therefore, the following aspects, functions, embodiments and advantages are for illustration purposes only and are not considered elements or limitations of the appended claims unless expressly stated in the claims. Similarly, reference to the "invention" shall not be construed as a generalization of any subject matter disclosed herein, and unless expressly stated in the claims, ancillary claims. It shall not be considered an element or limitation of.

0021As will be appreciated by those skilled in the art, aspects of the invention may be embodied as systems, methods or computer program products. Accordingly, aspects of the invention may take the form of a complete hardware embodiment, a complete software embodiment (including firmware, resident software, microcode, etc.) or a combination of software and hardware embodiments. , All of which may be commonly referred to herein as "circuits", "modules" or "systems". Further, aspects of the present invention may take the form of a computer program product embodied in one or more computer readable media having computer readable program code on it.

0022Any combination of one or more computer-readable media can be utilized. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, electronic, magnetic, optical, electromagnetic, infrared, or a semiconductor system, device or device, or any suitable combination described above, but is not limited thereto. More specific examples of computer-readable storage media are electrical connections with one or more wires, portable computer diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable PROM (EPROM or flash). Memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage, magnetic storage, or any suitable combination described above. In the context of this document, a computer-readable storage medium can be any tangible medium that can contain or store programs for use by or in connection with an instruction execution system, device or device.

0023The flowcharts and block diagrams of the diagrams illustrate the architecture, functionality, and behavior of possible implementations of systems, methods, and computer program products according to various embodiments of the invention. In this regard, each block in the flowchart or block diagram may represent a module, segment or portion of code, which contains one or more executable instructions for performing the specified logical function. In some alternative implementations, the functions described in the blocks can occur in a different order than described in the figure. For example, two blocks shown in succession can actually be executed at substantially the same time, depending on the functionality involved, or the blocks can sometimes be executed in reverse order. Each block of the block diagram and / or flowchart illustration, and a combination of blocks of the block diagram and / or flowchart illustration, with a special purpose hardware-based system or special purpose hardware that performs a specified function or operation. It can be executed by a combination of computer instructions.

0024Embodiments of the present invention may be provided to end users through a cloud computing infrastructure. Cloud computing generally refers to the provision of scalable computing resources as a service on a network. More formally, cloud computing can be defined as the computing power that provides an abstraction between computing resources and their underlying technology architecture (eg, servers, storage, networks), with minimal management effort. Or enable convenient, on-demand network access to a shared pool of configurable computing resources that can be quickly delivered and released with the interaction of service providers. Therefore, cloud computing allows users to virtualize computing resources (or the location of these systems) in the "cloud" without considering the underlying physical systems (or locations of these systems) used to provide the computing resources. Allows access to storage, data, applications and even fully virtualized computing systems).

0025Typically, cloud computing resources are provided to the user on a pay-as-you-go rate, and the user actually uses the computing resources (eg, the amount of storage space consumed by the user or the virtualization system created by the user). Is only charged for the number of). Users can access any resource in the cloud at any time, anywhere on the Internet. In the context of the present invention, the service provider may provide the imaging center with patient predictions from both predictive and reporting perspectives. For example, a dose prediction interface can be used to submit virtual phantom and CT data to cloud-based providers.

0026The flow charts and block diagrams of the figures illustrate the structures, functionality and behavior of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, segment or portion of code, which contains one or more executable instructions for performing the specified logical function. It should also be noted that in some alternative implementations, the functions described in the blocks can occur in a different order than described in the figure. For example, two blocks shown in succession can actually be executed at substantially the same time, depending on the functionality involved, or the blocks can sometimes be executed in reverse order. Each block of the block diagram and / or flowchart diagram, and a combination of blocks of the block diagram and / or flowchart diagram, with a special purpose hardware-based system or special purpose hardware that performs a specified function or operation. Also note that it can be executed by a combination of computer instructions.

0027In addition, certain embodiments of the invention below rely on specific examples of computed tomography CT scans that use a client-server architecture to provide a set of imaging dose predictions. However, of course, the techniques described herein are limited to radiation limited as part of imaging techniques (eg, PET scans, conventional radiography, and fluorescence fluoroscopy and angiography). May be adapted for use with other medical imaging techniques that rely on exposure.

0028FIG. 1 illustrates an example of a CT scan environment 100 and a related computing system configured to provide prediction of patient radiation dose according to one embodiment of the present invention. As shown, the CT scan environment 100 includes a CT scan system 105, an imaging system 125, and a dose prediction system 130. In addition, the dose prediction system 130 includes a database of imaging phantoms 132 and a simulation library 134.

0029As is well known, the CT scanner 105 provides a device used to irradiate subject 120 with X-rays from an astrophysical source 110. X-rays emitted from the X-ray source 110 pass through the tissues, organs, and structures of subject 120 at different rates, depending on the density and type of substance that the X-rays pass through (some of which are such tissues, Absorbed by organs and structures). A sensor placed with the ring 115 detects the radiation dose passing through the subject 120. The resulting sensor information is passed to the imaging system 125. The imaging system 125 provides a computing device configured to receive, store, and create images from sensor data obtained from a CT scanner.

0030The imaging system 125 allows the operator to receive data obtained from performing a CT scan in addition to performing a predetermined CT procedure. For example, the imaging system 125 may be configured to provide "windows" in various body structures based on its ability to block X-rays emitted from the source 110. CT scans (often referred to as "slices") are typically made perpendicular to the long axis of the body with respect to the axis or cross section. However, the CT scanner 105 may allow the imaging data to be reformatted into various planes or as a volume (3D) representation of the structure. Once the CT scan has been performed, the imaging data created by the CT scanner 105 can be saved and the resulting scanned image can be reviewed or otherwise evaluated. In one embodiment, the imaging data is formatted using well-known DICOM standards and stored in a PACS repository.

0031In one embodiment, the dose prediction system 130 provides a computing system and a software application configured to predict a patient absorbed dose to a given patient undergoing a given CT scan. Such predictions are made in a predictive sense (ie, before the scan is performed), but can also be made after the fact.

0032In the prediction example, the dose prediction system 130 may provide a prediction of patient dose before performing a CT scan. In yet one embodiment, the dose prediction system 130 may be configured to automatically generate alerts based on configurable thresholds. Criteria for alert generation can use a rules engine that can take into account age, gender, ICD9 / ICD10 coding, and other information (eg, specified cumulative dose limits) for a given patient or procedure. More generally, dose thresholds can be flexible enough to reflect any legal, institutional, or therapeutic requirements for dose monitoring. In one embodiment, the resulting dose prediction may be stored as part of the patient's medical record / history maintained by the imaging center, hospital, or other provider.

0033In addition, dose thresholds can optionally be used to produce incident reports sent to the appropriate physician. Incident reports may include procedural instructions and dose predictions that exceed rules or thresholds, as well as additional information necessary to provide background for physician intervention or decision. In one embodiment, such a report can be printed / emailed using a customizable XML template.

0034The imaging phantom 132 can provide a generally accepted mathematical model of parts such as human tissues, organs, structures and the like. For example, the imaging phantom 132 may provide a set of non-uniform rational splines (NURBS) used to create a three-dimensional (3D) model of the human body (or part thereof). Alternatively, the imaging phantom can be represented using Constructive Solid Geometry (CSG) or other mathematical representation. Different imaging phantoms 132 may be provided for general personalization based on series and gender. However, as mentioned above, the virtual geometry and body shape of the imaging phantom selected based solely on age and / or gender may correspond to the size, shape and organ position of the actual person undergoing the CT procedure. (Or may not correspond). Thus, in one embodiment, the dose prediction system 130 may be configured to deform the virtual phantom to better fit a particular patient. An example embodiment for deforming the virtual imaging phantom 122 is described in more detail below.

0035Once the imaging phantom has been modified to suit a particular individual, the dose prediction system 130 can perform simulations to predict a first past dose dose resulting from a given CT scan procedure. For example, in one embodiment, CT Monte Carlo simulations can be performed using CT scan parameters, CT procedure parameters, and modified phantoms to reach dose predictions. However, other simulations can be used. The results of a given dose prediction simulation can be stored in the simulation library 134.

0036For example, CT scanners include X-ray tube current and voltage, CT scanner mode, kVp, X-ray generator target angle, fan angle, collimation, slice thickness, focus-to-axis distance, flat filter (material and thickness). ), And can be parameterized for simulation based on the beam shaping filter (material and geometry). Various approaches can be used in the simulation process, but in one embodiment, "Computation of bremsstrahlung X-ray spectra over an energy range 15 KeV to 300 KeV ??" Calculation of X-ray spectrum) WJ Iles, Regne Unit. National Radiological Protection Board ??, NRPB, 1987, kVp, target angle and filtering are used to model the X-ray spectrum. Will be done.

0037In addition, the focal-to-axis distance determines the distance from the X-ray source to the axis of rotation, and the fan angle determines how much the beam spreads on the slice plane. Of course, these (and other parameters) can be selected when available or as needed to meet the needs of a particular example. But typically, for each anatomical region defined by the phantom, the energization is conserved slice by slice. Normalization simulation of CTDIvol phantom can be performed for each CT model. This slice-by-slice energization information combined with the mass of each anatomical region is sufficient to calculate the absorbed dose to each region of a given scan region.

0038However, typically, Monte Carlo simulations require much longer processing time to complete than the CT scan itself. Thus, in one embodiment, the dose prediction system 130 predicts the dose by interpolating between two (or more) simulations in the simulation library 134. For example, existing multivariate scattering interpolation can be used to calculate the patient dose in the first pass of the simulation data. Patient dose information will improve as more appropriate simulations are added. Similarly, new scanner models may be added to the simulation library 134 when calibration measurements and specifications for these scanners are obtained.

0039Simulation library 134 provides a database of Monte Carlo simulation results. In one embodiment, the simulation library 134 provides individual information about dose / energy delivery to a set of phantoms for a group of supported medical imaging scanners (eg, CT, RF, XA imaging modality, etc.). Store as both on-feed and deformed for patients with. In one embodiment, the simulation library 134 is used to provide a real-time search and / or calculation of the dose distribution given the given acquisition parameters, patient description, and scan area.

0040As mentioned, the simulation library 134 can be auto-expanded over time as additional Monte Carlo simulations are completed. For example, as a CT scan examination is performed, simulations for execution may be added to the queue. Priority can be given to the simulation of areas where data points are sparsely distributed. This improves the probability of identifying a simulation for interpolation, i.e., the simulation "space" covered by the simulation library 134. Similarly, the more simulations available in the simulation library 134, the tighter the threshold for selecting the simulation to interpolate in a given example, leading to better dose prediction accuracy.

0041Although shown in FIG. 1 as part of the CT scan environment 100, the dose prediction system 130 (and the phantom 132 and library 134) may be provided by or as a host service accessed by the CT scan environment 100. It should be noted that. For example, the imaging center may use a client interface on the imaging system 125 (eg, a secure web portal or a dedicated client application) to interact with the host dose prediction provider. Examples of such embodiments are described in more detail below with respect to FIGS. 11 and 12.

0042FIG. 2 illustrates an example of an imaging system 125 used to acquire CT scan data and manage patient dose predictions according to one embodiment. As shown, the imaging system 125 includes, but is not limited to, a central processing unit (CPU) 205, a CT system interface 214, a network interface 215, an interconnect 217, a memory 225 and a storage 230. The imaging system 125 may also include an input / output device interface 210 that connects an input / output device 212 (eg, a keyboard, display, and mouse device) to the imaging system 125.

0043The CPU 205 reads and executes the program instructions stored in the memory 225. Similarly, the CPU 205 stores and reads the application data in the memory 225. The interconnect 217 facilitates the transmission of program instructions and applications between the CPU 205, I / O interface 210, storage 230, network interface 215 and memory 225. The CPU 205 is included to represent a single CPU, a plurality of CPUs, a single CPU having a plurality of processing cores, and the like. The memory 225 is generally included to represent a random access memory. The storage 230 can be a disk storage device. Although shown as a single unit, Storage 230 is a fixed and / or removable storage device such as a disk drive, solid state storage (SSD), network attached storage (NAS), or storage area network (SAN). Can be a combination of. In addition, Storage 230 (or a connection to a storage repository) can meet a variety of criteria for data storage related to healthcare environments (eg PACS repositories).

0044As shown, the memory 220 includes an imaging control component 222, an image storage component 224, and a dose prediction interface 226. In addition, it includes storage 235, imaging protocol 232 and alarm threshold 234. The imaging control component 222 corresponds to a software application used to perform a predetermined CT scanning procedure as specified by imaging protocol 232. Imaging protocol 232 generally specifies the location, time, and duration for performing a particular CT procedure using a particular scan modality. The image storage component 224 is configured to store images and CT data derived during a given CT procedure, or interacts with the appropriate storage repository to store such images and data. Provide software to do. For example, CT scan data can be sent from / to the PACS repository (via a network interface) over a TCP / IP connection.

0045The dose prediction interface 226 provides a software component configured to interact with the dose prediction system 130 to obtain patient dose predictions that may result from a particular CT procedure. As described, in one embodiment, the dose prediction interface 226 can interact with the local system of the CT imaging environment. However, in an alternative embodiment, the dose prediction interface 226 can interact with the host service provider. In such cases, interface 226 may send a request for patient dose prediction to the host service provider. In addition, such claims may indicate an imaging phantom, a conversion to that phantom, and a protocol to follow for a CT scanning device and a given imaging scan. In either case, when used in a predictive sense (ie, before performing the procedure), patient dose prediction is to determine whether an alarm should be issued before the prescribed procedure is performed. It can be compared to alarm thresholds and rules (eg, alarms indicating that a given procedure exceeds (or is likely to) a cumulative dose limit for a given patient, organ or body part).

0046FIG. 3 illustrates an example of a dose prediction system 130 used for predicting and tracking cumulative patient radiation doses according to one embodiment. As shown, the dose prediction system 130 includes, but is not limited to, a central processing unit (CPU) 305, a network interface 315, an interconnect 320, a memory 325 and a storage 330. The dose prediction system 130 may also include an input / output device interface 310 that connects the input / output devices 312 (eg, keyboard, display and mouse devices) to the dose prediction system 130.

0047Like the CPU 205, the CPU 305 is included to represent a single CPU, multiple CPUs, a single CPU with multiple processing cores, etc., and the memory 325 is generally included to represent a random access memory. The interconnect 317 is used to transmit program instructions and applications between the CPU 305, I / O interface 310, storage 330, network interface 315 and memory 325. The network interface 315 is configured to transmit data over a communication network, for example, to receive a request from the imaging system for dose prediction. Storage 330, such as a hard disk drive or solid state (SSD) storage drive, can store non-volatile data.

0048As illustrated, memory 320 includes a dose prediction tool 321 which provides a set of software components. Illustratively, the dose prediction tool 321 includes a Monte Carlo simulation component 322, a simulation selection component 324, an image registration / splitting component 326, and a dose interpolation component 328. And storage 330 includes imaging phantom data 332, CT imaging protocol 334 and simulation library 336.

0049The Monte Carlo simulation component 322 is configured to predict patient radiation doses based on simulations using imaging phantom data 322 and a specific set of CT imaging devices and a designated imaging protocol 334. As mentioned, in one embodiment, the imaging phantom data 332 can be transformed or otherwise transformed to better suit the physical characteristics of a given patient.

0050The image registration / division component 326 may be configured to determine a set of transformations for transforming the imaging phantom data 332 before performing a Monte Carlo simulation using the phantom. For example, the image registration / splitting component 326 can use image registration techniques to evaluate a patient's scout localization image, along with a reference or localization image associated with the phantom. Image registration is the process of aligning two images in a common coordinate system. The image registration algorithm determines a set of transformations for setting the correspondence between two images. Once the transformation between the patient's scout image and the phantom's reference image is determined, the same transformation can be used to transform the phantom. Such deformations can scale, translate and rotate the geometry of the virtual phantom to accommodate the patient.

0051In another embodiment, image segmentation is used to determine the size and relative location of the patient's organs, tissues and anatomical structures. In such cases, the patient's available CT scan data can be divided to identify geometric deposits that may correspond to the organ (or other structure of interest). For example, in one embodiment, image segmentation may be used to identify a bounding box that is likely to contain a particular organ or structure. Other divisional approaches may be used to provide a more deterministic 3D volume area corresponding to the organ or structure. Once identified, this information is used to displace the geometry of the corresponding organ (or structure of interest) of the virtual phantom.

0052Although shown as part of the dose prediction server 130, in one embodiment the image registration / splitting component 326 is part of the imaging system 125 or otherwise in the computing infrastructure of the imaging facility. It should be noted that it is part. By doing so, the provider hosting the dose prediction service can transform a given virtual phantom without receiving information that could be used to identify patients undergoing CT scans at the imaging facility. It becomes possible to receive the conversion of. This approach can simplify (or exclude) certain legal or regulatory requirements associated with an entity that processes protected health information or medical records.

0053After the Monte Carlo simulation is complete, the resulting patient dose predictions, as well as the parameters provided in simulation component 322, are stored in simulation library 335. The dose interpolation component 328 is then used to determine the patient dose prediction from the simulation in the simulation library 335 without performing a complete Monte Carlo simulation. In this way, the simulation selection component 324 can compare the parameters of the CT scan, the equipment used to perform the CT scan, and the imaging phantom modified to represent a particular individual. This information is used to identify a set of two (or more) simulations to interpolate. Various approaches can be used, but in one embodiment the selection component 324 uses distance measurements to compare the modified phantom, CT procedure, and CT instrument with those in the simulation library 335. Can be done. In one embodiment, the top two (or N) choices are selected for interpolation. Alternatively, a simulation with overall similar measurements within a specified threshold is selected for interpolation. In such cases, the simulation is used for interpolation by adjusting the threshold up or down.

0054Given a scanner for testing and a set of parameters that describe the patient (kVp, target angle, gantry tilt, height, weight, etc.), the system can set customizable tolerances for each variable (eg,). , The actual kVp is within 10kV of the simulation). When searching for simulations, only simulations that are acceptable for all given parameters are included in the calculation. In one embodiment, the simulation results can be interpolated using known shepherd methods. The standard deviation across the simulation result set is used as a measure of uncertainty (for example, for the 5 simulation sets used, the SD of absorbed dose to the chest is 0.2 mSv and the absorbed dose to the liver is 0.2 mSv. SD is 0.15 mSv).

0055FIG. 4 illustrates method 400 for creating a suitable model for predicting a patient's radiation dose resulting from a CT scan, according to one embodiment. More specifically, Method 400 illustrates an example of an embodiment in which the image registration technique is used to transform a virtual phantom. As illustrated, Method 400 begins at step 405, where the dose prediction tool selects a virtual phantom with a pre-mapped localizer image. As mentioned, the virtual phantom can be selected based on the age and gender of the individual undergoing the CT scan procedure in question. In step 410, the dose prediction tool receives a scout image of the individual performing the dosimetry. Scout images provide a 2D image projection of an individual, such as anterior / posterior and / or side scout images acquired by a CT scanning system before performing a full CT procedure. Alternatively, the scout image can be a 3D volume of the individual acquired as part of the pre-procedure for a CT scan. At step 415, a pre-mapped locaiser image is obtained that corresponds to the use for transforming the selected virtual phantom. Pre-mapped images can be selected based on the relevant area of the patient being scanned. For example, in a patient undergoing (or receiving) a chest CT scan, the selected reference image may depict this area of an individual with a body geometry that closely matches the virtual phantom.

0056FIG. 5A illustrates an exemplary image representing a deformable phantom according to one embodiment. As illustrated, image 500 provides front / rear view 501 and side view 502 of the virtual image phantom. As illustrated on displays 501 and 502, the geometry of this phantom includes bone structures representing ribs 505, spine 515 and legs 522. In addition, indications 501 and 502 include geometry representing organs including stomach 510 and kidney 515. Virtual phantoms (as illustrated in indications 501 and 502) provide rough approximations of the size, shape and location of human organs, tissues and structures.

0057Although apparently a rough approximation of the actual human anatomy, virtual phantoms are generally accepted as providing a fairly accurate prediction of dose absorption. FIG. 5B illustrates an example of a 2D reference image of a portion of the human body corresponding to the phantom shown in FIG. 5A, according to one embodiment. As shown, the relative positions, sizes, and shapes of the bones, tissues, and organs in the reference image match well with the corresponding positions of the virtual phantom.

0058Revisiting Method 400, at step 420, the dose prediction tool performs an image registration process to determine the transformation between the patient's scout image and the reference image used to represent the virtual phantom. The result of image registration is the mapping from the point of the 2D Scout Localizer to the point of the reference image (or vice versa). Similarly, in the example of a 3D scout image of a patient (ie, a current or previous CT scan), the 3D image registration technique involves the point of the patient's 3D scout image and the point of the reference image corresponding to the phantom in 3D coordinate space. Can be mapped between.

0059In step 425, this same transformation is used to transform the geometry that represents the virtual phantom. By transforming the virtual phantom using the transformations obtained from the image registration process, the size, shape and organ position represented by the geometry of the virtual phantom will be more accurately the geometry of the actual patient. Matches. For example, performing an image registration process using the reference image shown in FIG. 5B and the patient scout localizer provides a transformation that can be used to transform the virtual phantom shown in FIG. 5A. The deformed virtual phantom can be used to predict the organ absorbed dose resulting from a given CT procedure. That is, the dose predictions obtained from the Monte Carlo simulation are patient-tailored and are more accurate and more consistent when used to predict patient doses across multiple scans.

0060FIG. 6 illustrates another method for creating a suitable model for predicting the patient's radiation dose resulting from a CT scan, according to one embodiment. More specifically, Method 600 illustrates an example of an embodiment in which the image splitting technique is used to transform a virtual phantom. Similar to Method 400, Method 600 begins by selecting an imaging phantom that the dose prediction tool deforms, for example, based on the patient's age and gender (step 605). However, instead of reading the patient's 2D image localizer, the dose prediction tool receives a 3D scan volume of some part of the patient (eg, a CT scan from a previous chest and abdominal CT) (step 610). Once acquired, image splitting is used to identify tissue, organs, structures or other landmarks in the image volume (step 615). A variety of available division approaches are available, but in one embodiment image division provides a minimal bounding box that surrounds each identified organ or structure.

0061At step 620, the dose prediction tool matches the organs and other anatomical landmarks identified in the CT scan segment (eg, bone location) with the corresponding landmarks of the virtual phantom. For example, FIG. 7 illustrates an exemplary slice of a CT scan overlaid on a corresponding CT slice of a virtual phantom, according to one embodiment. In this example, the virtual phantom slice 700 includes the heart 701, lung 703, spine 704, and humerus 705, in addition to the line 702, which represents the volume surrounded by the phantom. However, the locations and locations of the heart and lung organs of the virtual phantom do not correspond well to the locations of these organs as shown on CT. For example, the spatial area of the lung (706) does not match the size or location of the phantom lung 702 organ. Similarly, the phantom border 702 does not respond well to the patient. Therefore, when this phantom is used to predict dose, the phantom does not explain the large amount of adipose tissue in this patient, resulting in much greater dose absorption than would actually occur.

0062On the other hand, other landmarks of the phantom align well with the patient. For example, the spine and arms are nearly aligned on both the phantom (spine 704, humerus 705) and CT. Therefore, in step 625, the dose prediction system determines the 3D displacement map based on the matched anatomical or structural landmarks.

0063For example, in FIG. 7, the phantom slice 700 shows an unmodified or undeformed phantom, and the phantom slice 710 is displaced using the method of FIG. 6 (or using the image registration technique of the method of FIG. 4). Shows the same phantom slice (after being transformed).

0064After being deformed using the identified organ volume and specific patient displacement as illustrated on the phantom slice 710, the borderline 702'is closer to the contour of the patient CT scan, the phantom lung 703' and the heart 701. 'Is displaced to better reflect the location of these organs on the scan. In addition, other anatomical landmarks such as the spine and humerus remain in the same general position. The imaging phantom shown on slice 700 is shown superimposed on the patient's corresponding CT scan slice on slice 720. Similarly, the deformed phantom shown on slice 710 is shown superimposed on the patient's corresponding CT scan slice on slice 730.

0065Revisiting Figure 6, at step 630, the dose prediction tool creates a rasterized 3D representation of the displaced organs, tissues and structures of the virtual phantom. As mentioned above, virtual phantoms are described as a series of non-uniform rational splines (NURBS), while CT data is typically a series of 3D coordinate singles called "voxels" (abbreviations for "volume elements"). Expressed as point values, voxels extend the concept of pixels in three dimensions and various known approaches are available to "voxelize" a set of NURB or CSG data. This transforms the geometric or mathematical representation of NURB or CSG data into a 3D array of voxel values. In one embodiment, step 630 (voxelization step) is performed to avoid discontinuities that are often problematic in Monte Carlo simulations with mathematical phantoms (NURB or CSG based). In addition, to achieve speed improvements, voxel-based models are well suited for GPU-based computing methods.

0066Once a rasterized phantom is created, it can be used to predict the organ absorbed dose resulting from a given CT procedure (either before or after performing such a procedure). Similar to the image segmentation approach, dose predictions performed using a phantom deformed using the segmentation approach are patient-tailored and more accurate and more consistent for both individual and multiple scans. Brings some dose measurement.

0067FIG. 8 illustrates an example of a transverse slice of an imaging phantom superimposed on a patient's corresponding transverse CT slice according to one embodiment. In this example, the cross-sectional display 800 corresponds to the display 710 in FIG. 7, and the cross-sectional display 850 corresponds to the display 730 in FIG. The cross-sectional display is created by compositing the straight portions of the individual slices to create a vertical image. As illustrated, the transverse representations 800 and 805 provide a full length view, including components not found in the patient's superposed CT images (eg, brain 801 and kidney 802). As illustrated in Display 800, the virtual phantom boundary 810 does not correspond well to the patient's contour (ie, the size of the body surrounded by the patient's skin). However, on display 850, the phantom boundary 815 is displaced to better match the patient's reference CT scan data. Similarly, internal organs, structures and other tissues can be displaced.

0068Importantly, this example shows that displacement can occur with respect to elements of the virtual phantom that are not part of the patient's CT scan data. For example, as illustrated by the displacement position of kidney 802'on display 850, kidney 802 can be displaced by the movement of other organs for which CT scan data is available. Furthermore, this example shows that a virtual phantom is needed to predict patient dose, even when CT scan data is available. This occurs because the CT scans in this example were limited to the chest and abdomen, but X-ray scattering results in partial absorption by the patient's brain, kidneys and other organs and tissues. In other words, a virtual phantom is needed to predict organ dose absorption for organs that are not imaged as part of a given CT scan or procedure.

0069FIG. 9 illustrates another example of CT image segmentation and organ volume replacement of an imaging phantom according to one embodiment. In this example, the speculative CT volume 900 contains a set of bounding boxes that represent the split image positions of various organs (eg, liver 905, gallbladder 910, and right adrenal gland 915). In addition, volume 900 indicates an arrow representing the displacement of these organs based on the image division of CT scan data. In this particular example, the liver 905 was displaced downward to the right, the gallbladder 910 was displaced upward and in front of the liver 905, and the right adrenal gland 915 was occupied by the upper left, formerly liver 905. It has been moved to space. Further, in this example, the organ is represented by a boundary box and is displaced based on the geometric center of gravity. However, in an alternative embodiment, image division (for either the phantom or the patient's CT image data) provides a more accurate geometric volume that represents the structural elements of the organ, tissue or body. In such cases, the displacement is based on the mass center of gravity of the organ (eg, the center of gravity of the liver is unilaterally localized based on mass), or to other approaches primarily driven by the topology of a given organ volume. Can be based.

0070Displacement of one organ of the phantom (eg, liver 905) based on its corresponding position on a CT reference scan, as described in this example, results in other organs (eg, gallbladder 910 and right kidney 915). ) May be required. This happens because when the phantom is used to perform dose predictive analytics, the two organs clearly do not occupy the same physical volume. Thus, in one embodiment, the dose prediction tool can displace an organ, tissue or structure until a "steady state" is reached.

0071It should be noted that the examples of embodiments shown in FIGS. 4 and 6 can be used separately or in combination with each other to transform the virtual phantom. The specific approach or combination of selected approaches is based on the available imaging phantoms, mapped 2D and / or 3D images, and the availability of localized scout images and / or previous CT scan data for a given patient. And based on the type, it can be adapted to suit the needs of a particular example.

0072In one example, the cloud provider creates a host system used to perform dose prediction and maintains a library of calculated simulations, as well as augments the simulation library with a new example. To perform a Monte Carlo simulation. For example, FIG. 10 illustrates Method 1000 of a dose prediction service for providing patient dose predictions to multiple CT scan providers.

0073As illustrated, Method 1000 begins at step 1005, where the dose prediction service receives a 2D or 3D image registration transformation or 3D volume displacement field and phantom voxelization in addition to the image phantom (or reference to the image phantom). In an alternative embodiment, the dose prediction service may receive data describing the deformed phantom, such as a transformed NURBS resulting from a 2D or 3D image registration process or the CT field displacement technique described above.

0074In step 1010, the dose prediction service receives the parameters of the CT scan system and the imaging plan of the CT scan performed (or planned to be performed) on the patient. Once the patient, scanning equipment and CT scan provider parameters are received, the dose prediction service can identify two (or more) simulations that match the transformed phantom, CT scanning system parameters and imaging regimen. (Step 1015). The provider can set a customizable configuration tolerance for each variable (for example, the actual kVp is within 10kV of the simulation). In addition, the simulation is evaluated and only simulations that are acceptable for all (or a particular set) of given parameters are included in the calculation. In one embodiment, the simulation results can be interpolated using known shepherd methods. The standard deviation across the simulation result set is used as a measure of uncertainty (for example, for the 5 simulation sets used, the SD of absorbed dose to the chest is 0.2 mSv and the absorbed dose to the liver is 0.2 mSv. SD is 0.15 mSv).

0075At step 1020, the dose prediction service determines whether the matching simulation identified in step 1015 is within the permissible range of the parameters (or meets other thresholds or criteria). If not within range, the image phantom (and transformation / transformation) and received parameters are added to the queue for the simulated patient / scanner / image planning scenario (step 1025). As mentioned, the simulation is to determine adapted organ absorbed dose predictions for both individual patients based on the deformed phantom, and for a particular imaging facility based on CT scanners and calibration / configuration data. , Monte Carlo simulation can be used.

0076However, as SaaS providers' simulation libraries grow, most bills should be able to identify a set of simulations to interpolate. At step 1030, the dose prediction service performs multivariable scattering interpolation using the matching simulation identified in step 1015 to predict the absorbed dose to the organ for a particular patient and associated CT scan procedure. Such an analysis can be performed much faster than a complete Monte Carlo simulation, and dose prediction does not lag behind the sequence of procedures performed at a given imaging facility, and further (eg, cumulative dose limits). It should be noted that it will be possible to be provided at the same time as the prescribed procedure (to ensure that it does not exceed). In one embodiment, the currently used multivariable scattering interpolation method is called the "shepherd method". An example of this method is Shepard, Donald (1968). "A two-dimensional interpolation function for irregularly-spaced data." Proceedings of the 1968 ACM National Conference. Pp. 517-524.

0077At step 1035, once the interpolation process is complete, the dose prediction is returned to the billing system (eg, the dose prediction client program running on the computing system at the imaging facility). At the client, the dose management system tracks patient organ equivalent dose, effective dose, CTDI, DLP, DAP down to the test level. This information is also summarized to provide cumulative tracking of organ equivalent dose, effective dose, CTDI, DLP, DAP for a given patient history. In addition, this aggregate of information is used to provide an institution-wide presentation of per capita organ equivalent dose, patient effective dose, CTDI, DLP, and DAP. Therefore, dose prediction services can provide a wide range of [incomplete sentences] to imaging facilities. This same information is also available to imaging facilities that carry out regional cases of dose prediction systems.

0078FIG. 11 illustrates an exemplary computing infrastructure 1100 of a patient dose prediction service system configured to support multiple CT scan providers in one embodiment. As shown, the cloud-based provider 1125, which hosts the dose prediction service 1130, receives a dose prediction request from the imaging facility 11051-2 on the network 1120. At each imaging facility 1105, CT system 1110 is used to provide imaging services to patients. Imaging / Dose Client 1115 communicates with Dose Prediction Service 1130 to request and receive patient dose predictions, where dose predictions are adapted based on procedures and patients. As mentioned, claims can include CT procedures, parameters for scanning equipment and modality, and phantoms (or transformations used to deform the phantom) that have been deformed based on the particular patient's body morphology.

0079In the dose prediction service 1130, the simulation library 1135 selects a simulation to interpolate patient doses using the billing data and the module of CT scanner and procedure (phantom / CT system data 1140 shown in Figure 11). Used for. If good simulation candidates are not available for interpolation, service 1130 may bill the queue of simulations to perform. A Monte Carlo simulation is then performed on demand to provide both patient dose prediction and imaging procedures for a given patient, as well as new simulation data points to be added to library 1125.

0080Advantageously, embodiments of the present invention provide a variety of techniques for predicting radiation doses resulting from CT (and other) X-ray techniques. As described, image registration and / or image segmentation techniques can be used to create hybrid imaging phantoms that more closely match the size and shape of an individual's body. Doing so improves the accuracy of the dose predictions determined from the simulation. That is, the resulting hybrid phantom provides a much more accurate mathematical representation of a particular patient for use in dose simulation than the unmodified phantom alone.

0081Once the transformation is determined, a hybrid virtual phantom can be used to simulate a given CT procedure for the patient. For example, Monte Carlo simulation techniques can be used to predict the organ absorbed dose of a virtual phantom. Such simulation techniques, in addition to the virtual phantom (after conversion for a given patient), have many parameters related to the CT scanner model and the procedures performed to calculate accurate predictions of the absorbed dose to the organs. To use. However, predicting organ-absorbing organ doses using Monte Carlo simulations can require a significant amount of computational time, much longer than the time required to perform an actual CT scan. Thus, in one embodiment, patient dose predictions determined for a given procedure can be generated by interpolating between two (or more) previously completed simulations. If a "close" simulation is not available, then hybrid virtual phantoms, CT scanners and procedural data can be added to the queue for the full Monte Carlo simulation to be performed. Over time, a large library of simulations makes it possible to provide dose predictions in real time as procedures are scheduled and implemented. By doing so, it is possible to capture the cumulative dose of a given patient and observe the cumulative dose limit. Further, in one embodiment, the SaaS provider hosts a dose prediction service provided to multiple imaging facilities. In such cases, the service provider may have a robust library of simulations for use in interpolating dose predictions for imaging providers.

0082Although the description is directed to embodiments of the invention, other and further embodiments of the invention can be devised without departing from its basic scope, the scope of which is determined by the following claims.