Systems, methods, and computer readable media for bone volume assessment

Abstract

The invention relates to a bone volume evaluation system (100) based on magnetic resonance (MRI) patient scan data and / or at least one combined computed tomography (CT) data, comprising at least one processor module (110) and at least one memory module (120) for storing patient data. The processor module (110) is configured to process and / or analyze MRI patient scan data, at least one combination of CT data and / or simplified patient data, preferably to process the data combinations to determine at least one bone volume parameter of the first patient, including bone mineral density, fat percentage, calcium concentration, total bone trabecular interstitial density, etc. In addition, the invention also relates to a corresponding method 200 and a computer readable medium.

Description

Technical Field

[0001] The present invention relates to a system, method, and computer-readable medium for bone volume assessment based on patient magnetic resonance imaging (MRI) scan data and / or at least one set of synthetic computed tomography (CT) imaging data. Background Technology

[0002] In the existing technology, both magnetic resonance imaging (MRI) and computed tomography (CT) have inherent limitations. In recent years, attempts have been made to overcome these shortcomings by using these two technologies in parallel.

[0003] For example, bone marrow can be affected by a variety of conditions, such as post-traumatic bone contusion, diffuse tumor infiltration, or osteoporosis. To date, the primary imaging modalities for these conditions are MRI and CT, used alone or in combination, for example in radiotherapy planning, positron emission tomography (PET) attenuation correction, and bone quality assessment. Therefore, several studies have also analyzed methods for assessing bone mineralization using CT imaging, such as those supported by enhancement image processing techniques:

[0004] "Quantitative analysis of bone minerals by dual-energy computed tomography", Genant HK, Boyd D, Radiological Research, Nov-December 1977, 12(6): 545-51, PMID: 591258.

[0005] "Precise Measurement of Vertebral Mineral Content by Computed Tomography", Cann CE, Genant HK, Journal of Computed Tomography, August 1980, 4(4): 493-500, PMID: 7391292.

[0006] Quantitative computed tomography for spinal mineralization assessment: current status, Genant HK, Cann CE, Ettinger B, Gordon GS, Kolb FO, Reiser U, Arnaud CD, Journal of Computational Tomography, May-June 1985, 9(3): 602-604, PMID: 11536558.

[0007] However, with the development of dual-energy computed tomography (DECT) technology, CT imaging has become more valuable as an adjunct to traditional MRI diagnosis. Specifically, based on material breakdown technology, DECT can break down bone marrow into components such as bone marrow matrix, calcium, water, adipose tissue, hydroxyapatite, and collagen matrix, thereby achieving bone marrow segmentation and visualization (color coding). Therefore, several studies have explored DECT-based methods for assessing bone mineral density.

[0008] "Bone mineral assessment: a new dual-energy CT technique", Nickoloff EL, Feldman F, Atherton JV, Radiology, July 1988, 168(1): 223-228, PMID: 3380964.

[0009] "Assessment of Vertebral Trabeculae Based on Dual-Energy CT", Wesarg S, Kirschner M, Becker M, Kirschner M, Becker M, Erdt M, Kafchitsas K, Khan MF, Medical Information Methods, 2012, 51(5): 398-405, doi: 10.3414 / ME11-02-0034, published online on September 28, 2012, PMID: 23038636.

[0010] "Three-dimensional bone mineral density assessment of the lumbar spine in vivo without a phantom based on dual-energy CT", Wichmann JL, Booz C, Wesarg S, Kafchitsas K, Bauer RW, Kerl JM, Lehnert T, Vogl TJ, Khan MF, Radiology, June 2014, 271(3): 778-784, doi: 10.1148 / radiol.13131952, published online on January 16, 2014, PMID: 24475863.

[0011] Furthermore, DECT bone mineral assessment methods based on phantom-free technology have been explored in depth:

[0012] "Comparison between lumbar vertebral bone mineral density assessment and dual-energy X-ray absorptiometry based on a novel phantom-free dual-energy CT post-processing algorithm", Booz C, Hofmann PC, Sedlmair M, Flohr TG, Schmidt B, D'Angelo T, Martin SS, Lenga L, Leithner D, Vogl TJ, Wichmann JL, European Journal of Experimental Radiology, 2017, 1(1): 11, doi: 10.1186 / s41747-017-0017-2, published online on September 20, 2017, PMID: 29708178.

[0013] European Journal of Radiology, June 2015; 25(6): 1714-20, doi: 10.1007 / s00330-014-3529-7, published online on December 7, 2014. "Assessment of pedicle cancellous bone mineral density in a phantom-free system based on quantitative dual-energy CT: Correlation with pedicle screw pull-out strength", Wichmann JL, Booz C, Wesarg S, Bauer RW, Kerl JM, Fischer S, Lehnert T, Vogl TJ, Khan MF, Kafchitsas K.

[0014] American College of Radiology (2013) Quantitative Computed Tomography (QCT) Bone Mineral Density Testing Implementation Guidelines (Resolution 32), available at: http: / / www.acr.org / ~ / media / ACR / Documents / PGTS / guidelines / QCT.pdf, July 2017.

[0015] Furthermore, in recent decades, the use of MRI in radiotherapy planning, PET attenuation correction, and bone quality assessment has increased significantly. This trend is partly attributed to the superior soft tissue contrast of MRI compared to CT imaging.

[0016] However, while the combined use of CT and MRI technologies can improve medical imaging results, using two independent technologies in parallel increases the workload and cost of system configuration and operation.

[0017] Against this backdrop, synthetic CT technology has been developed in recent years, for example, to support radiotherapy procedures based solely on magnetic resonance (MR).

[0018] Synthetic CT generates X-ray attenuation and electron density information from MR scans to simulate CT images. Different substances in the image correspond to different Huntsfield units (HUs), and these HU values ​​can be converted into electron density in a treatment planning system (TPS) using calibration curves. The purpose of synthetic CT images is to provide X-ray attenuation and electron density information to achieve dose calculations with the same accuracy as CT images, such as enabling MR-based radiotherapy planning without the need for additional CT images.

[0019] However, the latest synthetic CT cannot provide the same level of tissue information as DECT. Therefore, DECT tissue analysis is superior to traditional synthetic CT analysis, especially in assessing tissue composition such as bone structure. Summary of the Invention

[0020] The present invention aims to provide a bone volume assessment system capable of providing bone volume assessment for individual patients based on magnetic resonance imaging (MRI) data and / or synthetic computed tomography (CT) data. By performing specific transformations and processing on the data, additional information is determined from the MRI and / or synthetic CT data, resulting in a more optimized and efficient workflow in terms of cost and workload, enabling the relevant assessment for individual patients to be completed with only a single medical imaging session. Furthermore, the present invention also aims to provide corresponding methods and computer-readable media.

[0021] The above objectives are achieved by the system of claim 1, the method of claim 13, and the computer-readable medium of claim 15, with preferred embodiments provided in the dependent claims.

[0022] According to the present invention, a bone volume assessment system based on MRI patient scan data and / or at least a set of synthetic CT data is provided, comprising:

[0023] At least one processor module for processing and analyzing MRI patient scan data;

[0024] At least one storage module is used to store patient data of at least the first patient, particularly MRI patient scan data and / or synthetic CT data and / or associated data;

[0025] The processor module has at least one data connection with at least one storage module, particularly a bidirectional data connection, so that data can be transferred for processing and / or storage.

[0026] The processor module is also configured as follows:

[0027] Receive at least one set of synthetic CT data from the first patient; and / or

[0028] The system receives a request for MRI patient scan data of the first patient from the storage module or MRI scanning device and / or applies a first conversion element to the MRI patient scan data of the first patient to generate at least one set of synthetic CT data of the first patient.

[0029] A second conversion element is applied to at least one set of synthetic CT data and / or MRI patient scan data of the first patient to generate simplified patient data providing information on the tissue volume of the first patient, particularly information related to water and / or fat volume; and / or

[0030] Receive tissue volume information from MRI scanning equipment, especially information related to water and / or fat volume;

[0031] A third conversion element is applied to simplified patient data and / or MRI patient scan data of the first patient to obtain total bone volume information;

[0032] The processor module is configured to process and / or analyze / evaluate MRI patient scan data, at least one set of synthetic CT data and / or simplified patient data, preferably combining the above data to determine at least one bone volume parameter of the first patient, particularly bone mineral density, fat percentage, calcium concentration, total density of intertrabecular bone, etc.

[0033] The core concept of this invention is that by converting and further evaluating MRI patient scan data, additional information can be obtained from the corresponding dataset, thereby enabling detailed analysis of bone composition, especially bone volume parameters such as bone mineral density and calcium concentration, as well as other bone component-related parameters.

[0034] Furthermore, since bone mineral density is related to bone strength, such as the pull-out strength of pedicle screws, another core concept of this invention is to optimize health care, such as optimizing screw placement in scenarios like unstable spinal fractures, spinal instability, degenerative scoliosis, and anterior support bone grafting spinal fusion (see European Radiology, June 2015, 25(6): 1714-1720, doi: 10.1007 / s00330-014-3529-7, published online on December 7, 2014, "Evaluation of Cancellous Bone Mineral Density of Pedicles Without a Phantom Based on Quantitative Dual-Energy CT: Correlation with Pedicle Screw Pull-out Strength", Wichmann JL, Booz C, Wesarg S, Bauer RW, Kerl JM, Fischer S, Lehnert T, Vogl TJ, Khan MF, Kafchitsas K.).

[0035] However, since MRI imaging data cannot provide the above information in an obvious and direct way, further processing and / or analysis of data collected during MRI imaging or by generating synthetic CT data is necessary to highlight specific features related to bone composition. In particular, additional information about the bone structure and its composition of the first patient can be obtained by combining these further processed data.

[0036] In particular, the present invention can provide multiple conversion elements to enable appropriate data processing from MRI patient scan data and / or synthetic CT data to obtain sufficient bone structure assessment information, namely bone volume parameters, such as mineral density assessment, and other relevant parameters of the first patient's bone structure.

[0037] MRI patient scan data specifically refers to the data acquired during a particular MRI scan of the first patient. Therefore, MRI patient scan data can refer to a specific part of the body of the first patient during an MRI scan.

[0038] Alternatively, synthetic CT data of specific parts of the first patient's body may have already been provided.

[0039] In the context of this invention, a processor module can be considered as a single processor unit, computer, etc., suitable for (corresponding) data processing and / or management.

[0040] In addition, the storage module can be any kind of storage device, such as a solid-state drive, a hard disk drive, or a database capable of storing and reading / forwarding data, thereby reading and writing data.

[0041] Furthermore, the processor module is configured to apply multiple (specifically three) different conversion elements to achieve appropriate data processing, particularly appropriate processing of MRI patient scan data, thereby providing bone volume parameter information, such as bone mineral density.

[0042] Specifically, different conversion elements can be provided as data processing elements and / or algorithms to appropriately process, analyze, and / or evaluate individual patient data, starting from an MRI patient scan dataset of an individual first patient. This can generate synthetic CT data, simplified patient data providing patient tissue volume-specific information, such as water and / or fat volume-related information, and total bone volume information, preferably where the total bone volume information is separable (or distinguishable) from the fat volume information in the simplified patient data.

[0043] For example, in the context of this invention, an MRI patient dataset can be used as the basis for the entire process, which is preferably acquired from a single MRI sequence. At least one three-dimensional image dataset can be obtained from this MRI sequence, which is then used to generate at least one synthetic CT dataset. Alternatively, multiple different synthetic CT datasets can be generated based on the MRI imaging sequences of individual patients.

[0044] In particular, the application of multiple conversion elements can be regarded as step-by-step data processing, and the system's processor module is reasonably (appropriately) configured for this step-by-step data processing.

[0045] For MRI patient scan data, in particular, three conversion elements are used to generate the corresponding information required for this invention.

[0046] If synthetic CT data of the first patient is provided, in particular, two conversion elements (i.e., the second and third conversion elements) can be applied to generate appropriate information related to the bone volume parameters of the first patient.

[0047] In particular, this step-by-step data processing / management allows bone volume information to be extracted from the raw MRI patient scan data.

[0048] Therefore, the present invention provides a reliable and more efficient process for determining bone volume information based on MRI patient scan data, namely bone volume assessment.

[0049] According to a preferred embodiment, the first conversion element is trained based on CT and MRI patient scan data from multiple patients to generate synthetic CT data. Furthermore, in another embodiment, training data used to train and / or adapt the first conversion element, particularly including CT and / or MRI patient scan data from multiple patients, is stored in the system's storage module and / or received by the system from a remote data storage device.

[0050] Specifically, the first conversion element is capable of generating synthetic CT patient data based on MRI patient scan data. The first conversion element can be trained and optimized based on datasets of multiple patients to improve the quality of the results provided by the first conversion element.

[0051] Therefore, the first conversion element can be configured and / or adapted using MRI patient scan data and corresponding CT patient scan data, enabling the first conversion element to generate synthetic CT patient data based solely on the MRI patient scan data of the first patient.

[0052] CT and / or MRI patient scan data from multiple patients can be stored in the storage module so that the processor module can request the training information for configuring and / or adapting the first conversion element.

[0053] Alternatively, data from multiple patients, such as MRI and CT scans used as training data, can be stored in a remote data storage device and retrieved from it, such as a central server.

[0054] According to another preferred embodiment, the first transformation element is provided in the form of a machine learning element / model, particularly a deep learning element / model, a neural network element / model, preferably a Transformer, encoder-decoder, diffusion model, U-shaped network, V-shaped network, recurrent neural network, recurrent interface network (RIN) or perceiver IO type neural network, or any other type of machine learning element / model.

[0055] Therefore, the first conversion element can be trained based on a given dataset, particularly MRI and CT patient scan data, so that it can be configured based on that given dataset to provide synthetic CT data of the first patient, starting from the corresponding MRI patient scan data of the first patient.

[0056] Furthermore, the first transformation element can also be provided as any other form of machine learning element.

[0057] In particular, as the quality and quantity of the provided learning data improve, the quality and effectiveness of the first conversion element can be continuously improved.

[0058] Therefore, the first conversion element can be further improved and configured to become more efficient over time and provide increasingly higher quality results.

[0059] In another embodiment of the invention, CT patient scan data of multiple patients are provided in the form of two or more sets of CT patient scan data corresponding to different energy levels (especially dual-energy CT patient scan data), or corresponding non-contrast multi-energy scan data.

[0060] Specifically, by providing multiple sets of CT patient scan data—that is, datasets acquired from the same patient at different energies—the quantity and quality of information provided by the CT patient scan data can be improved. This information can also be used to generate at least one set of synthetic CT scan data corresponding to a specific tube voltage energy level. Furthermore, by performing a specific calibration procedure on the generated synthetic CT scan dataset, a specific tube voltage energy level can also be achieved, thereby obtaining a specific synthetic CT scan dataset corresponding to a specific tube voltage energy.

[0061] Therefore, the configuration basis of the first conversion element, i.e. the corresponding training data basis, can have this additional multi-functional information to improve the configuration / adaptation effect of the first conversion element.

[0062] According to one embodiment, the second conversion element is provided in the form of an analysis element of the Dixon, IDEAL, MAGO, or MAGORIN type, as a machine learning based on water-lipid separation or in-phase and / or out-of-phase acquisition based on water-lipid separation, etc., to determine fat and / or water volume related information from MRI patient scan data and / or synthetic CT data.

[0063] Specifically, the second conversion element preferably provides separation of fat and water components, particularly based on MRI patient scan data from the first patient. Therefore, by applying the second conversion element to identify fat volume (information), the accuracy of total bone volume estimation can be improved.

[0064] Preferably, Dixon analysis can be used as the second transformation element, or, for example, the echo asymmetry and least squares estimation iterative decomposition (IDEAL) of water and fat can be used as the second transformation element.

[0065] In addition, based on the MRI patient scan data or synthetic CT data of the first patient, information related to water and / or fat volume can be determined. Other technical forms can provide a second conversion element, such as in-phase and out-of-phase acquisition, water-lipid separation algorithms based on machine learning, etc. (see, for example, US2021 / 0270919A1; Triay Bagur et al., “Solving the inherent water-lipid ambiguity of amplitude based on multi-peak fat modeling and multi-point search method”, Medical Magnetic Resonance, Vol. 82, No. 1, 460-475; Bray et al., “MAGORINO: Amplitude-only fat fraction and R*2 estimation based on Rayleigh noise modeling”, Medical Magnetic Resonance, Vol. 89, No. 3, 1173-1192; Liu et al., “Robust water-lipid separation based on deep learning model for exploring multi-echo gradient echo multi-echo characteristics”, Medical Magnetic Resonance, Vol. 85, No. 5, 2828-2841).

[0066] In summary, bone volume assessment can preferably be performed based on at least one set of synthetic CT patient data, generated by a conversion element trained from multiple patient CT scans of one type. Additionally, fat fraction images can be provided, acquired via MRI-based water-lipid separation conversion elements / algorithms such as DIXON, IDEAL, MAGO, MAGORINO, water-lipid separation-based machine learning, or water-lipid separation-based algorithms (based on in-phase and / or out-of-phase echo times).

[0067] In another embodiment of the invention, the third conversion element is provided in the form of a Nickoloff model element, specifically for determining the total bone volume based on MRI patient scan data and / or simplified patient data of the first patient.

[0068] In particular, the third conversion element can provide additional information on total bone volume to obtain corresponding results for bone volume parameters such as bone mineral density, fat percentage, calcium concentration, and total density of intertrabecular bone.

[0069] Therefore, in this invention, the Nickoloff model is preferably used as the third conversion element.

[0070] The Nickoloff model preferred in this invention can also be found in "Bone Mineral Assessment: A New Dual-Energy CT Technique" Nickoloff EL, Feldman F, Atherton JV, Radiology, July 1988, 168(1): 223-228, PMID: 3380964).

[0071] Alternatively, the third conversion element may take the form of other models or algorithms that enable appropriate data processing to further determine bone volume parameters based on MRI patient scan data and / or simplified patient data.

[0072] Specifically, the Nickoloff model combines the water component and red bone marrow together as a single component. However, by combining it with, for example, a Dixon element as a second conversion element, the red bone marrow component (mainly containing water) and the yellow bone marrow component (mainly containing fat) can be separated, especially by comparing the data provided by the second and third conversion elements.

[0073] According to a preferred embodiment, MRI patient scan data is provided by applying (dual-echo / multi-echo) gradient echo sequences (ME-GRE), (fast) spin echo sequences (TSE), dual-echo steady-state (DESS) sequences, multi-echo steady-state (MESS) sequences, balanced sequences (such as bSSFP), (multi-echo) ultrashort echo time (UTE) sequences, zero echo time (ZTE) sequences, synthetic MRI sequences, sequences based on magnetic resonance fingerprinting, sequences based on MR-STAT, etc., particularly to enhance the determination of fat and / or bone volume in the first patient by means of a conversion element.

[0074] Therefore, in the context of this invention, MRI patient scan data containing sequences / datasets focusing on different information can also be generated due to the different sequences applied.

[0075] By applying the aforementioned specific sequence techniques, such as multi-echo (e.g., dual-echo gradient echo), the overall generated MRI patient scan data can provide more information.

[0076] In another embodiment, MRI patient scan data and / or CT patient scan data are provided and / or stored in DICOM format, HDF5 format, NIfTI format, etc.

[0077] Therefore, the corresponding datasets are provided in a medical imaging standard format to ensure the highest quality and availability.

[0078] Furthermore, according to another embodiment of the present invention, the system includes at least one MRI scanning device, wherein the MRI scanning device is provided with a data connection, particularly a unidirectional data connection, between the MRI scanning device and at least one processor module and / or at least one storage module.

[0079] Therefore, the system of the present invention may also include the MRI scanning device itself, including direct data connection to the processor module and / or storage module.

[0080] Alternatively, the MRI scanning device used to collect MRI patient scan data can be located remotely so that the system of the present invention can obtain the corresponding MRI patient scan data through, for example, a central server, a remote data connection, etc.

[0081] Therefore, the system of the present invention does not necessarily require a dedicated MRI scanning device. Instead, the system of the present invention can be provided as a standalone system, acquiring MRI patient scan data from different MRI scanning devices via remote data connection.

[0082] Therefore, the system of the present invention can be regarded as either not containing (its own / independent / dedicated) MRI scanning equipment, or equipped with its own MRI scanning equipment.

[0083] In one embodiment, the system further includes at least one calibration phantom element, wherein the calibration phantom element is used to determine a universal calibration factor for MRI patient scan data of multiple patients, or to determine an individual calibration factor for each MRI patient scan data of multiple patients. Furthermore, according to a preferred embodiment, at least one calibration phantom element is provided in the form of multiple tubes and / or balls filled with a uniform or varying water-fat fraction, preferably ranging from 0% to 100%, specifically from 20% to 80%.

[0084] Therefore, the accuracy of at least one bone volume parameter can be improved by providing standardized reference values ​​with calibration phantom elements, particularly in the form of calibration phantom elements with multiple different water-fat fractions in tubes / balls.

[0085] Furthermore, in another aspect, the present invention relates to a method for bone volume assessment based on MRI patient scan data and / or at least a set of synthetic CT data, which applies the system according to any of the preceding claims, comprising the following steps:

[0086] Receive and / or request at least one set of synthetic CT data from the first patient;

[0087] or

[0088] Receive and / or request MRI patient scan data of a first patient from a storage module or MRI scanning device, and apply a first conversion element to the MRI patient scan data of the first patient to generate at least one set of synthetic CT data of the first patient;

[0089] The second conversion element is applied to at least one set of synthetic CT data and / or MRI patient scan data of the first patient to generate simplified patient data that provides tissue volume information of the first patient, particularly including water and / or fat volume related information, etc.; and / or tissue volume information, particularly including water and / or fat volume related information, etc., is received from the MRI scanning device (130);

[0090] A third conversion element is applied to simplified patient data and / or MRI patient scan data of the first patient to generate total bone volume information;

[0091] The processor module processes and / or analyzes and / or evaluates MRI patient scan data, at least one set of synthetic CT data, and / or simplified patient data, particularly by combining the above data to determine at least one bone volume parameter of the first patient based on the MRI patient scan data of the first patient, particularly including bone mineral density, fat percentage, calcium concentration, total density of intertrabecular bone, etc.

[0092] In particular, according to the method of the invention, bone volume assessment can be performed based on at least one set of synthetic CT data generated using a method trained with CT scan data from multiple patients.

[0093] In addition, a fat percentage image needs to be presented / provided, which can be obtained through MRI-based water-lipid separation algorithm components such as DIXON or IDEAL.

[0094] Subsequently, bone volume parameters, such as mineral density and / or other relevant parameters, can be obtained by using the Nickoloff method or a Nickoloff-based procedure.

[0095] Finally, since the CT value and fat volume (VF) are known, only a minimum number of parameters need to be determined, preferably only one equation needs to be used / solved to directly calculate the total bone volume (VT).

[0096] Furthermore, the method of the present invention can be fully performed and can be executed by the corresponding system configured according to the present invention.

[0097] According to another embodiment of the present invention, the method further includes the following steps:

[0098] Receive MRI patient scan data from multiple patients;

[0099] Receive CT patient scan data from multiple patients;

[0100] The first conversion element is trained using MRI patient scan data and CT patient scan data, preferably in the form of a machine learning element, deep learning element, neural network element, etc., to generate at least one set of synthetic CT data based on the MRI patient scan data of the first patient.

[0101] In particular, the method of the present invention ensures that the first conversion element used to generate synthetic CT data of the first patient from MRI patient scan data of the first patient can provide synthetic CT data results of appropriate and sufficient quality.

[0102] According to the method of the present invention, quality is ensured by appropriately training a first conversion element using MRI patient scan data and corresponding CT patient scan data from multiple patients, preferably in the form of a machine learning element, such as a deep learning element or a neural network element.

[0103] Therefore, as the amount of data from multiple patients increases, the performance of the first conversion element can be continuously improved.

[0104] According to another aspect of the present invention, a computer-readable medium is provided, comprising instructions that cause at least one computer (processor) or the like to execute the method according to the present invention.

[0105] In particular, the computer-readable medium may be a non-transitory computer-readable medium.

[0106] Therefore, computer-readable and executable instructions can be stored on any type of storage medium, such as optical discs, digital universal optical discs, flash memory devices, universal serial bus flash drives, hard disk drives, solid-state drives, etc. Attached Figure Description

[0107] Further details and advantages of the present invention are disclosed below with reference to the accompanying drawings.

[0108] Figure 1 A schematic diagram of one implementation of a bone volume assessment system;

[0109] Figure 2 This is a schematic flowchart of a bone volume assessment method.

[0110] Figure 3 for Figure 2 The diagram illustrates the method, including system calibration steps. Detailed Implementation

[0111] Figure 1 A schematic diagram of one embodiment of the bone volume assessment system 100 is shown, specifically determining at least one bone volume parameter (such as bone mineral density, fat percentage, calcium concentration, total density of intertrabecular bone, etc.).

[0112] according to Figure 1 The system 100 includes at least one processor module 110 and at least one storage module 120.

[0113] Optionally, the system 100 may further include at least one magnetic resonance imaging (MRI) scanning device 130.

[0114] The processor module 110 and the storage module 120 can be interconnected via a bidirectional data connection 140.

[0115] In addition, an optional MRI scanning device 130 may be connected to the processor module 110 and / or the storage module 120.

[0116] Specifically, the data connections 140 to the MRI scanning device 130 can be provided as unidirectional data connections 140.

[0117] Alternatively, system 100 may be provided without including at least one MRI scanning device 130.

[0118] Specifically, one or more MRI scanning devices 130 may be located, for example, at a remote location, and the MRI scanning devices 130 establish a data connection with the system 100. In particular, the system 100 has at least one processor module 110 and at least one storage module 120.

[0119] Alternatively, at least one processor module 110 and at least one storage module 120 may also be integrated into at least one MRI scanning device 130.

[0120] Regarding the MRI scanning device 130, the system 100 can be provided as an embedded system, preferably embedded in the corresponding MRI scanning device 130.

[0121] In addition, such as Figure 1 As shown, a remote data storage device and / or server 150 can be connected to system 100.

[0122] Specifically, the remote data storage device 150 may include a data connection 140 of a processor module 110 with system 100, preferably a bidirectional data connection 140.

[0123] Figure 2 A flowchart of a corresponding bone volume assessment method according to one embodiment is shown.

[0124] In the first step, processor module 110 preferably receives and / or requests 210 the MRI patient scan data 212 of the first patient from storage module 120, MRI scanning device 130 or remote data storage device 150.

[0125] Therefore, depending in particular whether the corresponding system 100 includes an MRI scanning device 130, MRI patient scan data 212 can be directly forwarded to the processor module 110, or requested / received from the storage module 120 or the (central) remote data storage device 150.

[0126] In the next step, processor module 110 applies / uses first conversion element 220 to MRI patient scan data 212 to generate and / or determine and / or process synthetic CT data 222.

[0127] Specifically, machine learning elements such as neural networks can be used as the first conversion element 220 to convert MRI patient scan data 212 into synthetic CT data 222.

[0128] In another step of method 200, a second conversion element 230 is applied to the MRI patient scan data 212 and / or synthetic CT data 222 of the first patient to generate and / or determine and / or process simplified patient data 232.

[0129] For example, the second conversion element 230 can be provided as a Dixon analysis element, etc., to identify fat volume information, especially to optimize subsequent total bone volume estimation.

[0130] Therefore, as an alternative, machine learning or other types of (analysis) sequences based on water-lipid separation, such as IDEAL, MAGO, MAGORINO, or acquisition-based water-lipid separation algorithms (based on in-phase and / or out-of-phase echo times) can also be applied.

[0131] In a further step, a third conversion element 240 may be applied to the MRI patient scan data 212 and / or simplified patient data 232 of the first patient to generate and / or determine and / or process at least one total bone volume information 242.

[0132] The third conversion element 240 may preferably be provided in the form of a Nickoloff model in order to identify total bone volume information 242.

[0133] Based on this, at least 250 bone volume parameters can be identified / determined.

[0134] Therefore, in summary, Figure 2 The method 200 shown can analyze and process MRI patient scan data 210, 212 in such a way that at least one bone volume parameter 250 of the first patient can be determined, such as bone mineral density, fat percentage, calcium concentration, total density of intertrabecular bone, etc.

[0135] also, Figure 2 The method 200 shown also provides a training basis for the first conversion element 220, namely configuration and adaptation.

[0136] Specifically, MRI patient scan data 202 and CT patient scan data 204 of multiple patients may be provided, for example, in a remote data storage device / server 150 or storage module 120 of system 100.

[0137] In the next step, the first conversion element 220 is trained based on MRI and CT patient scan data 202, 204 from multiple patients to generate improved synthetic CT data 222 from a single MRI patient scan dataset 210.

[0138] Specifically, the first conversion element 220 can be trained 206 once to achieve at least temporarily fixed settings based on one or more sets of currently available MRI and CT patient scan data 202, 204.

[0139] Alternatively, the first conversion element 220 may be trained periodically, such as once a month, on demand, such as when a large amount of additional new MRI / CT patient scan data 202, 204 is available, or (roughly) continuously, such as whenever a new MRI / CT patient scan dataset is available.

[0140] Preferably, the first conversion element 220 can be provided in the form of a machine learning element.

[0141] For example, machine learning components can be provided in the form of deep learning components, neural network components, etc.

[0142] Furthermore, in the context of this invention, machine learning elements, such as the first conversion element 220, can also be trained based on the output of Nickoloff model elements applied to CT patient scan data of multiple patients.

[0143] Based on MRI and CT patient scan data 202, 204 from multiple patients, the quality of synthetic CT data generated by applying the first conversion element 220 can be efficiently and advantageously improved.

[0144] In particular, according to one embodiment of the method of the present invention, bone volume assessment may be performed based on a set of synthetic CT data 222 of a first patient, which is generated by using a method of training 206 with one type of CT scan data of multiple patients 202, 204.

[0145] In addition, a fat percentage image needs to be presented / provided, which can be obtained using MRI-based water-lipid separation algorithm elements, such as DIXON, IDEAL, machine learning elements, etc., for example, in the form of applying a second conversion element 230.

[0146] Subsequently, by using the Nickoloff method, for example by applying the third transformation element 240, bone mineral density and other parameters related to bone volume information can be obtained.

[0147] Finally, since the CT value and fat volume (VF) are both known, the total bone volume (VT) of 250 can be directly calculated by using / solving only one equation.

[0148] More specifically, the method of the present invention can utilize Nickoloff's method 240 based on the following equation, which has exemplary parameter values ​​according to Koch et al., "Accuracy and precision of volumetric bone mineral density assessment based on dual-source dual-energy computed tomography and quantitative computed tomography: a phantom study," *European Journal of Experimental Radiology* (2021) 5:43, (https: / / doi.org / 10.1186 / s41747-021-00241-1):

[0149] X 90 Hu=(μ 90 -γ 90 g)·V TB +(β 90 t-γ 90 g)·V F +γ 90 g+δ (1)

[0151] X 150 Hu=(μ 150 -γ 150 g)·V TB +(β 150 t-γ 150 g)·V F +γ 150 g+δ (2);

[0153] These exemplary equations (1) and (2) correlate exemplary intensities X90 and X150 (in HU) in at least one set of (synthetic) CT sequences obtained at 90 and / or 150 kV tube voltages with the volume fraction VTB of the matrix material (bone minerals + collagen) and the volume VF of the adipose tissue. Thus, the synthetic CT sequence / set is not obtained at a specific tube voltage, but is reconstructed as a representation of a CT image obtained at a specific tube voltage.

[0154] However, regarding the present invention, the parameter VF can already be obtained from the MRI patient scan (data) sequence of the first patient. Therefore, according to the present invention, only one set of synthetic CT datasets is required to solve the above equations accordingly.

[0155] Furthermore, according to the present invention, the tube voltage of at least one, preferably two (synthesized) CT sequences can be 20 to 150 kV, preferably 90 to 150 kV.

[0156] The values ​​of t and g can be 0.92 g / cm³ and 1.02 g / cm³, respectively, while other variables (μ90, Y90, β90, μ150, Y150, β150, etc.) are energy-related constants. For example, the constant value of parameter α is -1000 HU. The values ​​of other variables can be obtained from Table 1 below (which contains parameter values ​​for different energy levels / photon energies):

[0157]

[0158] The values ​​of VTB and VF can be obtained by calculating the average strength of the trabeculae in the two sets of CT datasets. Finally, the bone mineral density (BMD) value ρBM (in g / cm3) can be calculated from VTB according to equation (3) by applying the material constants l = 3.06 g / cm3 and λ = 2.11.

[0159]

[0160] Other parameters can be determined by equations (4) to (6):

[0161]

[0162] ωρ W +θρ M =γ(ρ W +ρ M )≡γρ T (5)

[0163]

[0164] Therefore, the method of the present invention can efficiently determine and evaluate bone volume, particularly bone mineral density and / or other relevant bone volume parameters, especially in terms of cost, time and resources.

[0165] Specifically, the results of this bone mineral density assessment can be considered as absolute values. However, if calibration is incomplete, these values ​​can be considered as relative values, which may still be very useful in individual patients. This relative value is preferably correlated with a bone mineral density map, but may not perfectly align with the reference value.

[0166] Furthermore, according to another exemplary embodiment of the invention, it is also possible to generate / receive (a set / series) of synthetic CT data with a specific tube voltage. Moisture and / or fat fractions can be derived from (corresponding) MRI data.

[0167] Given the fat fraction and a single CT value, bone mineral density can be calculated as follows:

[0168] Specifically, the CT value (CTn) can typically be determined based on a specific voltage using the following formula, where the parameter δ is -1000 (i.e., a negative value), and the parameter E is the water offset value:

[0169] CT n =(μ-γg)V TB +(β T -γg)V F +γg+δ+∈ (7);

[0170] The parameters VF and VTB are usually unknown, with VF being obtained from results such as those from Dixon analysis or IDEAL analysis.

[0171] In addition, the CT value parameter CTn can be obtained from synthetic CT or conventional CT.

[0172] Therefore, based on formula (7), the total bone volume VTB can be calculated as follows:

[0173]

[0174] Subsequently, similar to the scheme described by Nickoloff, bone mineral density ρBM can be calculated according to formula (9) as follows:

[0175]

[0176] Therefore, this invention also allows for the efficient determination and evaluation of bone mineral density based on the generated synthetic CT data.

[0177] Figure 3 It shows Figure 2 The schematic diagram of the method shown includes the calibration steps of system 100, specifically MRI scanning equipment 130.

[0178] In particular, Figure 3 The process shown is the same as Figure 2 The difference in the illustrated implementation lies in the training 206 of the first conversion element and the calibration step 260 of the system 100, preferably an MRI scanning device 130.

[0179] The method of the present invention can be provided as including Figure 2 Training step 206 of the first conversion element shown. Figure 3 The calibration step 260 shown may include both training step 206 and calibration step 260.

[0180] according to Figure 3 It can provide a calibration phantom element 260, which preferably makes the images generated by fat volume imaging and post-processing steps more accurate, so as to allow for improved estimation of fat percentage.

[0181] The calibration phantom elements can be provided in the form of multiple tubes and / or spheres. Different elements may contain different water-lipid fractions, preferably ranging from 0 to 100%, particularly 20 to 80%. The calibration phantom may also contain bone-mimicking materials such as hydroxyapatite, calcium, collagen, etc., at concentrations ranging from 0 to 100%.

[0182] In addition, the obtained T1 and T2 signals can be weighted and taken into account to optimize the obtained MRI patient scan data accordingly.

[0183] Based on the calibration 270 of the system 100 using the provided calibration phantom element 260, MRI patient scan data of the first patient can be received / generated 210.

[0184] Alternatively, the calibration of system 100, especially the MRI scanning device 130, can also be performed in... Figure 2 and / or Figure 3 The following steps of the method 200 provided are for calibrating the post-processing of the MRI patient scan data.

[0185] Specifically, calibration step 270 can be provided / performed / repeated within a (predetermined) time frame, such as once a day, once a week, once a month, once every six months, or it can be provided / performed individually for each patient.

[0186] Furthermore, one or more calibration phantoms may also contain other related substances, such as calcium and collagen.

[0187] The system 100, particularly the MRI scanning device 130, and / or image post-processing can be calibrated in an appropriate manner to achieve optimized image data quality.

[0188] In summary, the present invention specifically discloses a system and method to allow bone volume assessment / evaluation based on a single medical imaging technique, namely, MRI patient scan data of a first patient.

[0189] Furthermore, by applying different conversion elements, such as machine learning elements as the first conversion element, MRI patient scan data can be processed to identify specific information, such as fat volume and total bone volume information, and ultimately, in particular, at least one bone volume parameter of the first patient can be determined by combining the processed data.

[0190] Figure Labels

[0191] 100 System

[0192] 110 Processor Module

[0193] 120 storage module

[0194] 130 MRI scanning equipment

[0195] 140 Data Connection

[0196] 150 Remote data storage devices / servers

[0197] 200 methods

[0198] 202 Receives MRI patient scan data from multiple patients

[0199] 204 Receives CT scan data from multiple patients

[0200] 206 Training the first conversion element

[0201] 210 Receive the MRI patient scan data of the first patient.

[0202] 212 MRI scan data of the first patient

[0203] 220 Application of the first conversion element

[0204] 222 Synthetic CT Data

[0205] 230 Application of the second conversion element

[0206] 232 Simplify patient data

[0207] 240 Application of a third conversion element

[0208] 242 Total Bone Volume Information

[0209] 250 Determining bone volume parameters

[0210] 260 provides phantom components

[0211] 270 Perform system calibration

Claims

1. A bone volume assessment system (100) based on magnetic resonance (MRI) patient scan data and / or at least one set of synthetic computed tomography (CT) data, comprising: at least one processor module (110) for processing and analyzing the MRI patient scan data; at least one storage module (120) for storing patient data of at least a first patient, in particular the MRI patient scan data and / or the synthetic CT data and / or associated data; wherein there is at least one data connection (140), in particular a bidirectional data connection (140), between the processor module (110) and the at least one storage module (120) for data transfer for processing and / or storage; wherein the processor module (110) is further configured to: receive at least one set of synthetic CT data of the first patient; and / or receive and / or request MRI patient scan data of the first patient from the storage module (120) or an MRI scan device (130) and apply a first conversion element to the MRI patient scan data of the first patient to generate at least one set of synthetic CT data of the first patient; apply a second conversion element to the at least one set of synthetic CT data and / or the MRI patient scan data of the first patient to generate simplified patient data providing information on tissue volume of the first patient, in particular water and / or fat volume related information etc.; and / or receive tissue volume information, in particular water and / or fat volume related information etc., from the MRI scan device (130); apply a third conversion element to the simplified patient data and / or the MRI patient scan data of the first patient to obtain total bone volume information; wherein the processor module (110) is configured to process and / or analyze the MRI patient scan data, the at least one set of synthetic CT data and / or the simplified patient data, preferably combining the data sets for determining at least one bone volume parameter of the first patient, in particular bone mineral density, fat percentage, calcium concentration, total density of trabecular bone spacing etc.

2. The system (100) according to claim 1, characterized in that the first conversion element is trained based on CT patient scan data and MRI patient scan data of a plurality of patients to generate synthetic CT data.

3. The system (100) according to claim 2, characterized in that training data, in particular CT patient scan data and / or MRI patient scan data of a plurality of patients, for training and / or adapting the training data of the first conversion element are stored in the storage module (120) of the system (100) and / or are received by the system (100) from a remote data storage device (150).

4. The system (100) according to any of the preceding claims, characterized in that The first conversion element is provided in the form of a machine learning element, in particular in the form of a deep learning element, a neural network element, preferably a neural network of the type Transformer, Encoder-Decoder, Diffusion Model, U-Net, V-Net, Recurrent Neural Network, Recurrent Interface Network (RIN) or Perceiver IO, or in the form of any other type of machine learning element.

5. The system (100) according to any one of claims 2 to 4, characterized in that The CT patient scan data of the plurality of patients are provided in the form of two sets of CT patient scan data involving different energy levels, in particular dual-energy CT patient scan data, or unenhanced multi-energy scan data, respectively.

6. The system (100) according to any one of the preceding claims, characterized in that The second conversion element is provided in the form of a Dixon, IDEAL, MAGO, MAGORIN type analysis element, as a water-fat separation based machine learning or water-fat separation based in- and / or out-of-phase acquisition, etc., in order to determine fat and / or water volume related information from the MRI patient scan data and / or the synthetic CT data.

7. The system (100) according to any one of the preceding claims, characterized in that The third conversion element is provided in the form of a Nickoloff model element for determining total bone volume information, in particular based on the MRI patient scan data and / or the reduced patient data of the first patient.

8. The system (100) according to any one of the preceding claims, characterized in that The MRI patient scan data are provided by applying a (multi-) echo gradient echo sequence, a (fast) spin echo sequence, a dual echo steady state (DESS) sequence, a multi-echo steady state (MESS) sequence, a balanced sequence (e.g. bSSFP), a (multi-) ultra-short echo time (UTE) sequence, a zero echo time (ZTE) sequence, a synthetic MRI sequence (based on magnetic resonance fingerprinting, based on MR-STAT), etc., in particular for further determining the fat volume information of the first patient by the conversion element provided.

9. The system (100) according to any one of the preceding claims, characterized in that The MRI patient scan data and / or the CT patient scan data are provided and / or stored in DICOM format, HDF5 format, NIfTI format, etc.

10. The system (100) according to any one of the preceding claims, characterized in that The system (100) comprises at least one MRI scanning device (130), wherein the data connection (140), in particular a unidirectional data connection (140), is provided between the MRI scanning device and the at least one processor module and / or the at least one storage module.

11. The system (100) according to any one of the preceding claims, characterized in that The system (100) further comprises a calibration phantom element, wherein the calibration phantom element is configured to determine a general calibration factor for a plurality of MRI patient scan data of a plurality of patients, or to determine an individual calibration factor for each MRI patient scan data of a plurality of patients, respectively.

12. The system (100) according to claim 10, characterized in that the calibration phantom element is provided in the form of a plurality of tubes and / or spheres filled with a uniform or varying water-fat fraction, preferably a water-fat fraction in the range of 20% to 80%.

13. A method (200) for bone volume assessment based on magnetic resonance (MRI) patient scan data and / or at least one set of synthetic computed tomography (CT) data, applying the system (100) according to any of the preceding claims, comprising the following steps: receiving and / or requesting at least one set of synthetic CT data of the first patient; or receiving and / or requesting (210) MRI patient scan data (212) of the first patient from the storage module or the MRI scan device and applying a first conversion element (220) to the MRI patient scan data of the first patient to generate at least one set of synthetic CT data (222) of the first patient; applying a second conversion element (230) to the at least one set of synthetic CT data (222) and / or the MRI patient scan data (210) of the first patient in order to generate simplified patient data (232) providing information on tissue volumes of the first patient, in particular water and / or fat volume related information, etc.; and / or receiving tissue volume information, in particular water and / or fat volume related information, etc., from the MRI scan device (130); applying a third conversion element (240) to the simplified patient data (232) and / or the MRI patient scan data (210) of the first patient to generate total bone volume information (242); wherein the processor module processes and / or analyzes and / or evaluates the MRI patient scan data (210), the at least one set of synthetic CT data (222) and / or the simplified patient data (232), in particular combines the aforementioned data sets, in order to determine at least one bone volume parameter (250) of the first patient based on the MRI patient scan data (210) of the first patient, in particular bone mineral density, fat percentage, calcium concentration, total density of trabecular bone spacing, etc.

14. The method (200) according to claim 13, characterized in that the method (200) further comprises the following steps: receiving MRI patient scan data (202) of a plurality of patients; receiving CT patient scan data (204) of a plurality of patients; applying the MRI patient scan data (202) and the CT patient scan data (204) to train the first conversion element (206), preferably in the form of a deep learning element, a neural network element, etc., to generate at least one set of synthetic CT data (222) based on the MRI patient scan data (210) of the first patient.

15. A computer readable medium comprising instructions to cause at least one computer or the like to perform the method according to claim 13 or 14.

Images