A multi-voxel spectral reconstruction method and system with ROI adaptation

Abstract

This invention belongs to the field of magnetic resonance imaging technology, and particularly relates to a multi-voxel spectral reconstruction method and system with ROI adaptive capability. The method includes acquiring a structural image of the scanned object; performing multi-voxel spectral scanning on the scanned object to acquire K-space spectral data; based on the location information of the region of interest (ROI) and interference regions, combined with the point spread function of the spectral scanning system; performing target reconstruction processing on the K-space spectral data according to the optimal voxel localization parameters to generate reconstructed spectral data; and extracting spectral information corresponding to the ROI from the reconstructed spectral data. This invention, through a strategy of multiple adaptive reconstructions in a single scan, can independently calculate the optimal localization parameters and perform targeted reconstruction for each ROI within the field of view based on the same set of K-space data. This allows different anatomical structures to obtain the maximum volume proportion in their dedicated reconstructed voxels without compromising the localization effect between different regions.

Description

Technical Field

[0001] This invention belongs to the field of magnetic resonance imaging technology, and particularly relates to a multi-voxel spectral reconstruction method and system with ROI adaptive. Background Technology

[0002] Magnetic Resonance Spectroscopy Imaging (MRSI), also known as chemical shift imaging, is a technique that combines the spatial localization capabilities of magnetic resonance imaging (MRI) with the metabolite analysis capabilities of magnetic resonance spectroscopy (MRS), providing non-invasive metabolic and biochemical information within living tissues.

[0003] In existing multi-voxel spectral scanning techniques, the typical workflow is to first locate, then scan, and finally reconstruct. That is, the operator first sets the field of view (FOV) and divides the image into voxel grids based on high-resolution structural images (such as T1 or T2 weighted images). The aim is to match the voxel positions within the grids to the anatomical structures as closely as possible, so that the spectra of individual voxels can represent the metabolic information of the region of interest (ROI) to the greatest extent possible, i.e., maximizing the volume fraction of the ROI within the voxel.

[0004] However, existing technologies are limited in practical applications by their inherently low spatial resolution (voxel size is typically greater than 1 cm) and uniform, fixed voxel grids, making it difficult to simultaneously accommodate multiple anatomical structures in terms of spatial coverage and geometry. Specifically, since the voxel grid as a whole only allows for translation and rotation adjustments, when the grid achieves optimal alignment for a specific region (such as the midbrain), other regions of interest (such as the basal ganglia or hippocampus) often fail to achieve ideal coverage simultaneously, meaning it is difficult to guarantee that these regions simultaneously have the largest volume fraction in individual voxels. Furthermore, many regions of interest are irregularly shaped, further increasing the difficulty of precise localization. On the other hand, existing localization processes mainly rely on the operator's subjective experience to position the structure on the image, lacking an objective, quantitative indicator based on signal component contribution to evaluate the localization effect, and failing to accurately determine whether the current grid position truly maximizes the signal of the region of interest and minimizes the signal of the interfering region. Summary of the Invention

[0005] To address the technical problems in existing technologies where multi-voxel spectral scanning and localization is limited by a fixed grid, making it difficult to simultaneously achieve optimal coverage of multiple regions of interest (ROIs) and lacking objective quantitative evaluation indicators for localization performance, this invention proposes an ROI-adaptive multi-voxel spectral reconstruction method and system.

[0006] This invention is achieved through the following technical solution:

[0007] A multi-voxel spectral reconstruction method with ROI adaptation includes the following steps:

[0008] Step S1: Obtain the structural image of the scanned object, and determine the location information of at least one region of interest and the spatial location information of at least one interference region based on the structural image;

[0009] Step S2: Perform multi-voxel spectral scanning on the scanned object to obtain K-space spectral data;

[0010] Step S3: Based on the location information of the region of interest and the interference region, and combined with the point spread function of the spectral scanning system, calculate the optimal voxel localization parameters of the target region of interest under the current scanning conditions;

[0011] Step S4: Based on the optimal voxel localization parameters, perform target reconstruction processing on the K-space spectral data to generate reconstructed spectral data, so that the voxel grid center of the reconstructed spectral data is aligned with the optimal voxel position of the target region of interest.

[0012] Step S5: Extract spectral information corresponding to the target region of interest from the reconstructed spectral data.

[0013] Furthermore, in step S3, the method for determining the optimal voxel localization parameters specifically includes:

[0014] Construct a spatial scoring map, where the value of each pixel in the spatial scoring map is used to characterize the reconstruction quality when the center of the spectral voxel is located at that point;

[0015] The calculation logic of the spatial scoring map is as follows: the point spread function of the spectral scanning system is used to perform convolution operation on the spatial distribution of the target region of interest to obtain the signal contribution distribution of the region of interest, and the weighted signal contribution distribution of the interference region is subtracted from it;

[0016] The maximum value is searched in the spatial scoring map, and the spatial location coordinates corresponding to the maximum value are determined as the optimal voxel localization parameters.

[0017] Furthermore, the calculation of the spatial scoring map satisfies the following logical relationship:

[0018] ;

[0019] in Represents convolution; Represents the diffusion function; A spatial mask representing the region of interest; Indicates the first A spatial mask for a region of non-interest; These are the weighting coefficients corresponding to the non-interest regions, used to control the suppression strength of interference regions. Indicates the region of non-interest. Indicates the subscript index.

[0020] Furthermore, in step S4, the target reconstruction process is specifically implemented by performing a summation operation based on the inverse discrete Fourier transform:

[0021] Based on the optimal voxel localization parameters, determine the phase correction term for each sampling point in K-space. ;

[0022] Using the summation formula that includes the phase correction term, from K-space spectral data Directly calculate spatial domain signals ;

[0023] The calculation logic performed by the summation formula is as follows:

[0024] ;

[0025] That is, for any point in space Its signal For all K-space sampling points of signal The result of the weighted summation, where the weighting term is a complex exponential function, is given by an exponential part that includes the spatial location-coded phase. With the phase correction term The superposition of.

[0026] Furthermore, the phase correction term The optimal voxel localization parameters satisfy a linear relationship as follows:

[0027] ;

[0028] in, These represent the coordinates in the k-space in the three directions, respectively; It is the geometric offset of the scan block center relative to the gradient center, calculated based on the optimal voxel localization parameters; this phase correction term is introduced into the exponent term of the summation formula. This enables the translation of the reconstructed spectral voxel grid in the spatial domain, aligning it with the optimal voxel positioning parameters.

[0029] Furthermore, in step S4, the target reconstruction process employs a sub-voxel resolution reconstruction method, specifically including:

[0030] The K-space spectral data is zero-padded to increase the dimension of the K-space matrix;

[0031] A Fourier transform is performed on the expanded K-space matrix to generate a sub-voxel grid image with higher apparent resolution.

[0032] Based on the optimal voxel positioning parameters, locate the subvoxel in the subvoxel grid image that is spatially closest to the optimal voxel position;

[0033] The data of this subvoxel is used as the optimal reconstructed spectrum for the target region of interest.

[0034] Furthermore, the structural image contains multiple spatially independent regions of interest;

[0035] The method executes steps S3 to S5 independently for each region of interest.

[0036] For each region of interest, an independent set of optimal voxel localization parameters is calculated, and an independent target reconstruction process is performed for each, thereby obtaining multiple sets of reconstructed spectral data. Each set of reconstructed spectral data contains only the optimal spectral information for the corresponding region of interest.

[0037] Furthermore, the multi-voxel spectral scanning employs either Cartesian sampling or non-Cartesian sampling;

[0038] When Cartesian sampling is used, the point spread function is in the form of a Sinc function;

[0039] When non-Cartesian sampling is used, the point spread function is determined according to the specific sampling trajectory, and in step S4, the data is first gridded before the target reconstruction process is performed, or phase modulation is directly applied to the non-Cartesian K-space data.

[0040] This invention also proposes a ROI-adaptive multi-voxel spectral reconstruction system, comprising:

[0041] The image and localization module is used to acquire structural images of the scanned object and determine the location information of at least one region of interest and interference region;

[0042] The data acquisition module is used to control the magnetic resonance equipment to perform multi-voxel spectral scanning and acquire K-space spectral data;

[0043] The parameter optimization module is used to calculate the optimal voxel localization parameters of the target region of interest based on the location information of the region of interest and the point spread function of the system.

[0044] The reconstruction module is used to perform target reconstruction processing on the K-space spectral data based on the optimal voxel positioning parameters, using inverse discrete Fourier transform summation operation with phase correction term or K-space zero-adding operation.

[0045] The analysis module is used to extract the optimal spectrum from the reconstructed data.

[0046] The beneficial effects of this invention are:

[0047] (1) The present invention uses a strategy of one scan and multiple adaptive reconstructions to independently calculate the best localization parameters and perform targeted reconstruction for each region of interest in the field of view based on the same set of K-space data, so that different anatomical structures can obtain the maximum volume ratio in their dedicated reconstruction voxels without compromising the localization effect in different regions.

[0048] (2) This invention uses the point spread function of the system to perform convolution operation with the spatial mask of the region of interest / interference region to construct a spatial scoring map. This method can quantitatively evaluate the degree of inclusion of ROI signal and the degree of suppression of ROA signal at any spatial location, thereby transforming the traditional subjective placement that relies on human experience into objective optimization based on mathematical model, ensuring the accuracy and repeatability of voxel center localization.

[0049] (3) When calculating the optimal voxel position, the present invention introduces a weight suppression mechanism for the interference region, so that the reconstructed voxel not only contains the signal of the target tissue to the greatest extent, but also avoids the signal aliasing of high lipid, bone or other interfering tissues to the greatest extent. Compared with traditional grid positioning, this method significantly reduces the spectral pollution caused by part of the volume effect and improves the accuracy of subsequent metabolite quantitative analysis.

[0050] (4) The present invention achieves grid translation by post-processing algorithms such as linear phase modulation or subvoxel zero-padding of K-space data. It does not require changing the magnetic resonance hardware equipment, nor does it require repeated scanning to obtain the best spectrum in different regions. Without increasing the patient's scanning time burden, the algorithm significantly improves the spatial resolution performance and usability of clinical spectral data. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a schematic diagram of the overall process of a multi-voxel spectral reconstruction method with ROI adaptation proposed in this invention;

[0053] Figure 2 This is a schematic diagram of the Mode A flow of a multi-voxel spectral reconstruction method for ROI adaptive proposed in this invention;

[0054] Figure 3 This is a schematic diagram of the Mode B process of a multi-voxel spectral reconstruction method for ROI adaptive proposed in this invention.

[0055] Figure 4 This is a schematic diagram of the initial localization state of multiple regions of interest in an adaptive multi-voxel spectral reconstruction method for ROI proposed in this invention.

[0056] Figure 5 This is a schematic diagram of the reconstructed mesh optimized for a specific region of interest in a multi-voxel spectral reconstruction method for ROI adaptive proposed in this invention.

[0057] Figure 6 This is a schematic diagram of the initial reconstruction grid containing the interference region of a multi-voxel spectral reconstruction method for ROI adaptive proposed in this invention.

[0058] Figure 7 This is a schematic diagram of the adaptive reconstruction mesh after suppressing the interference region in the ROI adaptive multi-voxel spectral reconstruction method proposed in this invention.

[0059] Figure 8 This is a schematic diagram of the localization effect at the original resolution of the ROI adaptive multi-voxel spectral reconstruction method proposed in this invention.

[0060] Figure 9 This is a schematic diagram of the optimal localization based on sub-voxel resolution reconstruction of a multi-voxel spectral reconstruction method for ROI adaptive ROI proposed in this invention.

[0061] Figure 10 This is a schematic diagram of a terminal device for a ROI-adaptive multi-voxel spectral reconstruction method proposed in this invention.

[0062] Figure 11 This is a schematic diagram of a readable storage medium for a ROI-adaptive multi-voxel spectral reconstruction method proposed in this invention.

[0063] In the diagram, 200 is the terminal device, 210 is the memory, 211 is the RAM, 212 is the cache, 213 is the ROM, 214 is the program / utility, 215 is the program module, 220 is the processor, 230 is the bus, 240 is the external device, 250 is the I / O interface, 260 is the network adapter, and 300 is the program product. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of this invention are only for explaining this invention and are not intended to limit this invention.

[0065] Example 1

[0066] This embodiment details a method and system for ROI adaptive multi-voxel spectral reconstruction based on a magnetic resonance imaging system. The core inventive concept is to overcome the technical bottleneck of traditional multi-voxel spectral imaging (MRSI) where positioning is performed before scanning and the voxel grid is fixed after scanning. By employing a strategy of one scan and multiple adaptive reconstructions, signal processing algorithms are used to perform sub-voxel-level information recombination in the K-space domain or image domain, thereby achieving optimal voxel center matching for each region of interest (ROI). In existing multi-voxel spectral scanning techniques, the usual process relies on operators to set the imaging range and divide the voxel grid based on high-resolution structural images, in order to maximize the representation of metabolic information of the region of interest by the spectrum of individual voxels. However, due to the inherently low spatial resolution of spectral scanning and the use of a uniform, fixed voxel grid, it is difficult to simultaneously meet the precise localization requirements of multiple anatomical structures in a single scan. The voxel grid as a whole can only be adjusted by translation and rotation. When the grid achieves optimal alignment for a specific region, regions of interest in other locations are often located at the grid boundaries or off-center, causing the signal to be dispersed into adjacent voxels—a phenomenon known as partial volume effect—which severely impacts the accuracy of spectral analysis. This invention aims to resolve this fundamental contradiction by establishing a quantitative evaluation system based on the point spread function and a reconstruction system based on linear phase modulation, thereby achieving precise spatial manipulation of spectral signals.

[0067] refer to Figure 1 The specific implementation process of this invention is as follows:

[0068] S1. Obtain the structural image of the scanned object, and determine the location information of at least one region of interest and the spatial location information of at least one interference region based on the structural image;

[0069] S2. Perform multi-voxel spectral scanning on the object to obtain K-space spectral data;

[0070] S3. Based on the location information of the region of interest and the interference region, and combined with the point spread function, calculate the optimal voxel localization parameters of the target region of interest;

[0071] S4. Based on the optimal voxel localization parameters, perform target reconstruction processing on the K-space spectral data to generate reconstructed spectral data.

[0072] S5. Extract spectral information corresponding to the target region of interest from the reconstructed spectral data.

[0073] The system can take two approaches: one is to directly utilize the phase correction term. The system employs two main approaches: first, it reconstructs the region of interest (ROI) using the IDFT summation formula; second, it first adds zeros to the K-space to improve apparent resolution, and then locates the optimal subvoxels within a finer mesh. Regardless of the approach used, it ultimately obtains a set of optimal spectral data for that specific ROI. More importantly, this invention supports parallel processing of multiple spatially independent ROIs. For example, in a single brain scan, the midbrain, basal ganglia, and hippocampus can be simultaneously analyzed. The system only needs to scan the K-space data once, then calculates and reconstructs the offsets for the midbrain, the basal ganglia, and the hippocampus, respectively. This strategy of multiple calculations in a single scan allows each ROI to achieve the maximum volume fraction and optimal signal-to-noise ratio in its dedicated reconstruction version, thus completely solving the problem of simultaneously considering multiple anatomical structures in existing technologies. Furthermore, it effectively utilizes the ROA weighting mechanism to suppress interfering signals such as lipids, significantly enhancing the clinical application value of magnetic resonance spectroscopy.

[0074] This embodiment is not limited to Cartesian acquisition. For non-Cartesian acquisition, the point spread function needs to be analyzed according to the acquisition and reconstruction method. For example, gridding can be performed first and then reconstruction can be carried out using the Cartesian acquisition method. The core idea is the same as the above embodiment.

[0075] Specifically, before delving into the specific implementation steps of this invention, it is necessary to clarify the essence of magnetic resonance spectral signal acquisition and reconstruction. This is the theoretical basis for the establishment of the technical solution of this invention. In magnetic resonance spectral imaging, the spectral information contained within a reconstructed voxel does not merely originate from the material inside the geometric boundary of that voxel. Its signal composition is profoundly mathematically related to the system's signal acquisition method and reconstruction method. After magnetic resonance scanning excites the sample, it can acquire and reconstruct signals at any point in space. There exists a time-varying... A changing time-domain signal, denoted as Taking the most common clinical three-dimensional multi-voxel spectral scanning Cartesian sampling as an example, the system performs phase encoding by applying linear gradients in three spatial directions, thereby acquiring signals in the frequency domain K-space. From the perspective of continuous signals, the acquired K-space signals... Physically, this is expressed as the integral form of the Fourier transform of a signal in the spatial domain. According to the principle of the Fourier transform, this integral relationship can be precisely expressed as:

[0076]

[0077] The above formula indicates that for each K-space data point In fact, it refers to all spin signals within the entire imaging volume. At a specific spatial frequency In actual image or spectral reconstruction, it is necessary to recover the spatial domain signal from the acquired finite discrete K-space data. This process is usually achieved through three-dimensional discrete Fourier transform (IDFT).

[0078] In a standard, non-adaptive reconstruction, for each voxel mesh center position Its time-domain signal varies with time The formula for discrete summation reconstruction is:

[0079]

[0080] However, due to limitations in physical equipment and scanning time, the K-space sampling range is inevitably finite, and the sampling is discrete and equally spaced. This truncated sampling process in the frequency domain, according to signal and system theory, is equivalent in the image domain to performing a convolution operation on the real continuous spatial distribution using a specific point spread function (PSF). For Cartesian sampling, this PSF is the Sinc function. Since the Sinc function has infinitely extended sidelobes, its main lobe width is approximately two pixels. That is, the reconstructed signal of any voxel, while primarily derived from the sample signal within the corresponding geometric space of that voxel, cannot be... This also inevitably introduces leakage signals from other spatial locations outside the voxel, which decay according to the Sinc function. From a frequency domain perspective, a voxel can be viewed as a frequency filter with a specific center frequency and bandwidth. Spatial signals with the exact same center frequency as the voxel will be completely and losslessly reconstructed within the voxel. Signals with frequency differences are weighted into the voxel according to the Sinc function decay ratio, indicating that the voxel actually contains signals beyond its nominal spatial bandwidth, and their contribution weight decays according to the Sinc function as the spatial distance increases. For example, the signal contribution of a signal at the voxel's edge is attenuated by approximately 36% compared to the signal at the center.

[0081] Therefore, the relative spatial position of the region of interest (ROI) within the voxel grid has a decisive influence on the final spectral reconstruction quality. When the center of the ROI is strictly aligned with the geometric center of a voxel, the main lobe peak of the Sinc function is located exactly at the center of the ROI, and the signal energy is highly concentrated within the target voxel. In this case, the signal of the ROI will be reconstructed to the maximum extent within that voxel, and interference from the surrounding background tissue is minimized, which corresponds to the optimal reconstruction condition. Conversely, when the center of the ROI deviates significantly from the voxel center, for example, by about half the voxel bandwidth, the ROI is located exactly on the boundary between two voxels. The main lobe peak of the Sinc function will be located between the two adjacent voxels, causing the volume and signal energy of the ROI to be evenly distributed between the two adjacent voxels. For this ROI, this voxel partitioning constitutes the worst reconstruction parameter, resulting in severe signal dispersion, a significant reduction in the effective volume fraction within the target voxel, and a significant increase in the signal contribution from non-target regions. The ability of the reconstructed spectrum to characterize the ROI characteristics is reduced to the minimum.

[0082] Based on the analysis of the above physical principles, this invention proposes a localization evaluation method based on voxel information components, aiming to quantitatively evaluate the merits of using any spatial location as a reconstruction center. To address the problem of optimizing the voxel center frequency of the target region of interest, this embodiment establishes the following generalized criteria:

[0083] An optimal ROI optimization acquisition and reconstruction scheme should include as much effective spectral information as possible from the region of interest (ROI), while suppressing the spectral contribution from interfering regions (ROA) as much as possible.

[0084] Based on this criterion, this embodiment constructs an evaluation matrix called the spatial rating map and defines the overall rating function. This scoring function quantifies the weighted difference between the ROI signal and the ROA signal at a specific voxel location, and its mathematical expression is defined as:

[0085]

[0086] in, For the first The weighting coefficients for each interference region are used to control the inhibition intensity of different types of interference regions according to clinical needs. The definition can simultaneously reflect the proportion of volumetric information resulting from the spectral weighting contribution of each region due to the acquisition method, providing a unified evaluation framework for optimizing ROI spectral reconstruction. In the specific calculation process, in order to find the optimal location across the entire field of view, the system uses the point spread function (PSF) of the spectral scanning system to create a high-resolution ROI information map, i.e., to calculate the spatial scoring map. This calculation process involves complex spatial convolution operations, and the specific calculation formula is as follows:

[0087]

[0088] in Represents convolution; Represents the diffusion function; A spatial mask representing the region of interest; Indicates the first A spatial mask for a region of non-interest; These are the weighting coefficients corresponding to the non-interest regions, used to control the suppression strength of interference regions. Indicates the region of non-interest. This indicates the subscript index. By performing the above convolution operation, the system effectively simulates the signal point diffusion effect during the scanning process, generating a map reflecting the signal quality distribution when any point in the entire space is used as the voxel center. Subsequently, the system... Perform a full-space search to find the pixel with the largest value, and determine the spatial location or geometric offset of this maximum value relative to the original scan center. This is the optimal voxel localization parameter for the ROI under the current acquisition and reconstruction conditions. This position represents the most effective suppression of aliasing and interference from the ROA while maximizing the ROI signal contribution. It is theoretically the most favorable spatial position for subsequent spectral reconstruction and quantitative analysis.

[0089] The optimal voxel localization parameters, i.e., the optimal geometric offset, for a specific ROI were obtained through the above evaluation algorithm. The next core technology of this invention lies in how to use these parameters to move the voxel grid to the optimal position through mathematical means without re-scanning the physical grid.

[0090] This embodiment employs an adaptive reconstruction method based on K-space linear phase modulation. Its physical principle is based on the shift property of the Fourier transform; spatial domain translation is equivalent to linear phase modulation in the frequency domain (K-space). Phase encoding is achieved by applying a linear gradient in space. The zero point of this linear gradient is fixed and unchangeable in physical space. When the positioning center needs to deviate from the gradient center, in order to ensure the reconstructed image is in the correct position, a phase change caused by the deviation of the scan block center from the gradient center needs to be applied in K-space. Specifically, this includes the following steps:

[0091] refer to Figure 2 First, structural image scanning is performed, and based on this, two parallel processes are started simultaneously. On the one hand, the structural image is used to accurately locate the ROI / ROA to obtain spatial migration parameters; on the other hand, multi-voxel spectral scanning is carried out to acquire raw K-space data.

[0092] Phase encoding is achieved by applying a linear gradient in space. The zero point of this linear gradient is fixed and unchangeable in physical space. When the ROI / ROA positioning center needs to be offset from the gradient center, in order to make the reconstructed spectrum in the correct position, the phase change caused by the offset of the scan block center from the gradient center needs to be applied in K space.

[0093] To achieve this goal, this embodiment innovatively modifies the aforementioned standard discrete reconstruction formula in the ROI spectral reconstruction stage by introducing a key phase correction term. The revised reconstruction formula, which is the inverse discrete Fourier transform summation formula including phase modulation, is expressed as follows:

[0094]

[0095] The above formula shows that for any point in space The time-domain signal that is ultimately reconstructed For all K-space sampling points of signal The weighted summation result differs from traditional reconstruction in that the weighting term here is a complex exponential function, whose exponential part includes not only the original spatial location encoded phase. It also includes a phase correction term calculated based on the ROI / ROA positioning parameters. This superposition process mathematically precisely achieves a sub-voxel-level translation of the reconstructed mesh in the spatial domain. To ensure that the translation process does not produce geometric distortion, a phase correction term is used. It must maintain a strict linear relationship with the K-space coordinates. Based on the derivation in this embodiment, the specific definition formula of this phase correction term is as follows:

[0096]

[0097] in, These represent the coordinate variables in the K-space in three directions; while It is precisely based on the previous step The optimal geometric offset of the entire scan block center relative to the original gradient center is obtained from the maximum value location calculation. When the system calculates the optimal offset... Substituting into the above formula and applying the phase correction term during ROI spectral reconstruction, the reconstructed voxel center will precisely coincide with the optimal voxel position of the region of interest, thereby obtaining the optimal reconstructed spectrum of the ROI. After reconstruction, the system enters the final ROI spectral analysis stage to quantitatively calculate metabolites.

[0098] In addition to the above Figure 2In addition to the phase modulation-based reconstruction mode (Mode A), this embodiment also provides another sub-voxel resolution-based reconstruction mode (Mode B), which utilizes K-space zero-padding technology. Although the true resolution of magnetic resonance refers to the smallest spatial feature size that the system hardware and acquisition parameters can truly distinguish, determined by the K-space acquisition parameters, reconstruction or post-processing can make the image or spectrum appear to have higher resolution, i.e., increase the apparent resolution.

[0099] From a signal processing perspective, zero-filling in K-space is essentially a reweighting of subvoxel information, applying a rational frequency offset to K-space. The linear phase of the reconstructed first phase Each pixel is equivalent to The first after adding zero The fact that zero-addition reconstruction and phase modulation reconstruction are theoretically inherently consistent is that of zero-addition reconstruction and phase modulation reconstruction.

[0100] The specific steps are as follows:

[0101] refer to Figure 3 First, structural image scanning is performed, and ROI / ROA localization is completed based on this. At the same time, multi-voxel spectral scanning is performed on the target area to obtain raw spectral data.

[0102] In the sub-voxel resolution spectral reconstruction stage, K-space zero-filling technology is used to process the acquired data. Although the true resolution of magnetic resonance refers to the smallest spatial feature size that the system hardware and acquisition parameters can truly distinguish, which is determined by the K-space acquisition parameters, reconstruction or post-processing can make the image or spectrum appear to have higher resolution, that is, increase the apparent resolution. From the perspective of signal processing, K-space zero-filling is essentially a reweighting of sub-voxel information, applying a linear phase with a rational number frequency shift in K-space.

[0103] Finally, in the data analysis phase, the location information is combined with the reconstructed high apparent resolution data to proceed to the spectral analysis stage to obtain more accurate metabolic information.

[0104] In practice, the system first performs zero-padding expansion on the original K-space data, that is, adds zero values ​​around the data matrix to increase the matrix dimension. Then, it performs a Fourier transform on the expanded matrix to generate a spectral image with higher pixel density. Next, according to... The calculated optimal positioning parameters are used to find the sub-voxel with the closest spatial distance to the optimal location within this high apparent resolution sub-voxel grid. Finally, the data from this sub-voxel is extracted as the optimal reconstructed spectrum for the ROI. This method, within the limits of computational resources, intuitively approximates the optimal reconstruction result through interpolation.

[0105] Example 2

[0106] This embodiment is based on Embodiment 1, and is described in detail with reference to specific scanning examples.

[0107] A two-dimensional multi-voxel scanning environment is set up, and its original image matrix is ​​set as a 4×4 grid, within which exist... Four independent regions of interest (ROIs). (Reference) Figure 4 In a conventional, unadaptive initial scan and reconstruction (labeled Reconstruction 1), the voxel mesh is typically positioned based on the geometric center of the field of view or a major organ. In this example, the initial mesh position precisely makes the ROI... It achieved relatively ideal coverage.

[0108] Specifically, ROI Primarily located within the voxel in row 1, column 2, ROI The voxel occupies the second row and third column, while the ROI Located in row 4, column 1, since the centers of these ROIs are relatively close to the geometric centers of their respective voxels, according to the aforementioned point spread function (PSF) theory, the main lobe energy of the Sinc function is highly concentrated within these voxels. Therefore, the spectral data of these three regions have high signal-to-noise ratio and tissue specificity. However, for ROIs... However, the situation is completely different. Under the mesh generation of Reconstruction 1, the ROI It precisely crosses the boundary lines of two voxels, one in row 4 and the other in column 3. According to the attenuation characteristics of the Sinc function, when the target object is located at the edge of a voxel, its signal energy will be severely dispersed into adjacent voxels, resulting in the voxel in row 4 and column 3 containing only [missing information - likely a specific type of voxel]. The partial volumetric effect, which includes a portion of the signal and may be mixed with background noise, makes it impossible to obtain a representative ROI in the reconstruction 1 result. To address the issue of obtaining accurate spectra, this embodiment does not require rescanning. Instead, it initiates an adaptive reconstruction process, where the system first utilizes... Algorithm targeting ROI Convolution calculations are performed on the spatial location and shape to determine how to make... Optimal spatial offset required to be located at the voxel center The calculations show that alignment can be achieved simply by performing a specific subvoxel translation of the mesh in the row or column direction. The system then substitutes this offset into the phase correction formula. ;refer to Figure 5 Phase modulation is applied to the original K-space data, and a second reconstruction is performed, labeled Reconstruction 2. In the new mesh generated by Reconstruction 2, the original mesh lines are shifted, so that the center of the new voxel in the 4th row and 4th column falls precisely on the ROI. At the center of the anatomy. At this moment, ROI The signal energy is reconstructed to the maximum extent possible within this single voxel. Ultimately, doctors or analysis software can extract the signal energy from the reconstructed dataset. The spectrum, and extract from the reconstructed 2 dataset. The spectrum is obtained, thus perfectly taking into account all spatially irregularly distributed regions of interest in a single scan.

[0109] Example 3

[0110] This embodiment is based on Embodiment 1, and is described in detail with reference to specific scanning examples.

[0111] refer to Figure 6 The second implementation scenario involves the complex situation of suppressing the region of interest (ROA) in the interference area. Also within a 4×4 grid, it focuses on a specific region of interest (ROI). However, there is an interference region ROA located in its immediate vicinity. For example, in a cranial scan, scalp lipids or bones, in the default localization scheme, the voxels in the second row and second column cover most of the ROI. However, due to the randomness of the mesh position, the edge of this voxel also cuts into the ROA. Within this range. The consequence of this is that the reconstructed spectrum of this voxel not only includes... The metabolic information is also severely mixed with information from... Interference signals. If It is a high-intensity lipid signal, and its spectral peaks may completely mask it. To address the issue of weak metabolite peaks in [the system], this invention introduces a weighted scoring mechanism. The objective function of the system is:

[0112] ;

[0113] in It is a large weighting factor used to penalize any content that includes... The system uses the point spread function (PSF) to determine the reconstructed location of the signal. and The system performs convolution operations on the spatial mask to generate a ScoreMap. During the search process, the system will detect if the current voxel mesh is moved away from the target area. A slight translation is made in the direction of ROI, although The proportion within voxels may decrease slightly, but ROA The signal contribution will decrease sharply because the translation utilizes the zeros or low sidelobe regions of the Sinc function to shield the interference source. The system calculates this, which makes the overall score... The optimal positioning parameters, i.e., the optimal offset, are maximized, and the K-space phase modulation parameters are calculated based on these. Subsequently, the K-space data is weighted and phase-modulated and reconstructed. In the reconstructed new data, a reference is used... Figure 7 The new voxel in the second row and second column maintains its position. While achieving optimal coverage, it successfully excluded This invention eliminates interference that cannot be avoided during physical scanning in the spectral reconstruction stage, significantly improving the purity of the spectrum and the accuracy of quantitative analysis.

[0114] Example 4

[0115] This embodiment is based on Embodiment 1, and is described in detail with reference to specific scanning examples.

[0116] refer to Figure 8 The third implementation scenario, namely the sub-voxel resolution-based reconstruction strategy, still uses a 4×4 raster matrix for the original image. In this scenario, the Region of Interest (ROI) appears. The problem of poor positioning, namely Located on the boundary of the original coarse grid, in addition to using the aforementioned phase modulation for continuous domain translation, this embodiment also provides a discrete domain approximation method. The system reads the original K-space data matrix and fills its high-frequency components with zeros, expanding its matrix dimension to twice or more than the original. According to the Fourier transform property, zero-filling in the K-space domain is equivalent to sinc interpolation in the spatial domain. This makes the reconstructed image matrix an 8×8 fine grid. Although this operation does not increase the actual physical resolution, it significantly improves the apparent resolution, subdividing the original large voxel into four sub-voxels. In these sub-voxels, the signal components contained in different locations are different, each corresponding to a weighted average of different small regions in space. The system uses the optimal physical coordinates calculated by ScoreMap to search and match in this 8×8 sub-voxel grid. The results show that the sub-voxel center at the 7th row and 6th column of the subdivided grid corresponds to a specific offset position at the original scale and the ROI. The optimal position is closest. For example, Figure 9 As shown in the black box, the sampling range of this subvoxel precisely defines the ROI. This avoids the surrounding blank areas. Therefore, the system directly selects the data of this subvoxel as the ROI. The advantage of this method is that the calculation is intuitive and it can generate a high-density spectral distribution map, which makes it easy for doctors to find the hidden optimal sampling point in a seemingly continuous image with a simple mouse click on the post-processing workstation.

[0117] Example 5

[0118] refer to Figure 10 Based on Example 1, this example proposes a terminal device for a multi-voxel spectral reconstruction method with ROI adaptation. The terminal device 200 includes at least one memory 210, at least one processor 220, and a bus 230 connecting different platform systems.

[0119] The memory 210 may include a readable medium in the form of volatile memory, such as RAM 211 and / or cache memory 212, and may further include ROM 213.

[0120] The memory 210 also stores a computer program that can be executed by the processor 220, causing the processor 220 to perform any of the above-described ROI adaptive multi-voxel spectral reconstruction methods in the embodiments of this application. The specific implementation method and the achieved technical effects are consistent with those described in the embodiments of the above applications, and some details will not be repeated here. The memory 210 may also include a program / utility 214 having a set (at least one) of program modules 215. Such program modules include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment.

[0121] Accordingly, processor 220 can execute the aforementioned computer program, as well as executable program / utility 214.

[0122] Bus 230 can represent one or more of several types of bus structures, including a memory bus or memory controller, peripheral bus, graphics acceleration port, processor, or a local bus using any of the various bus structures.

[0123] Terminal device 200 can also communicate with one or more external devices 240, such as keyboards, pointing devices, Bluetooth devices, etc., and with one or more devices capable of interacting with it, and / or with any device that enables it to communicate with one or more other computing devices (e.g., routers, modems, etc.). This communication can be performed via I / O interface 250. Furthermore, terminal device 200 can communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via network adapter 260. Network adapter 260 can communicate with other modules of terminal device 200 via bus 230. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with terminal device 200, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0124] Example 6

[0125] This embodiment proposes a readable storage medium for a ROI-adaptive multi-voxel spectral reconstruction method. The computer-readable storage medium stores instructions that, when executed by a processor, implement any of the aforementioned ROI-adaptive multi-voxel spectral reconstruction methods. The specific implementation method and the achieved technical effects are consistent with those described in the above-mentioned application embodiments, and some details will not be repeated.

[0126] Figure 11 The present embodiment illustrates a program product 300 for implementing the above-described applications. This product may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a terminal device, such as a personal computer. However, the program product 300 of the present invention is not limited thereto. In this embodiment, the readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device. The program product 300 may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0127] Computer-readable storage media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable storage medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof. Program code for performing operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on a user computing device, partially on a user device, as a standalone software package, partially on a user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing devices can be connected to user computing devices via any type of network, including local area networks (LANs) or wide area networks (WANs), or they can be connected to external computing devices (e.g., via the Internet using an Internet service provider).

[0128] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A multi-voxel spectral reconstruction method with ROI adaptation, characterized in that, Includes the following steps: Step S1: Obtain the structural image of the scanned object, and determine the location information of at least one region of interest and the spatial location information of at least one interference region based on the structural image; Step S2: Perform multi-voxel spectral scanning on the scanned object to obtain K-space spectral data; Step S3: Based on the location information of the region of interest and the interference region, and combined with the point spread function of the spectral scanning system, calculate the optimal voxel localization parameters of the target region of interest under the current scanning conditions; Step S4: Based on the optimal voxel localization parameters, perform target reconstruction processing on the K-space spectral data to generate reconstructed spectral data, so that the voxel grid center of the reconstructed spectral data is aligned with the optimal voxel position of the target region of interest. Step S5: Extract spectral information corresponding to the target region of interest from the reconstructed spectral data.

2. The ROI-adaptive multi-voxel spectral reconstruction method according to claim 1, characterized in that, In step S3, the method for determining the optimal voxel localization parameters specifically includes: Construct a spatial scoring map, where the value of each pixel in the spatial scoring map is used to characterize the reconstruction quality when the center of the spectral voxel is located at that point; The calculation logic of the spatial scoring map is as follows: the point spread function of the spectral scanning system is used to perform convolution operation on the spatial distribution of the target region of interest to obtain the signal contribution distribution of the region of interest, and the weighted signal contribution distribution of the interference region is subtracted from it; The maximum value is searched in the spatial scoring map, and the spatial location coordinates corresponding to the maximum value are determined as the optimal voxel localization parameters.

3. The ROI-adaptive multi-voxel spectral reconstruction method according to claim 2, characterized in that, The calculation of the spatial scoring map satisfies the following logical relationship: ； in Represents convolution; Represents the diffusion function; A spatial mask representing the region of interest; Indicates the first A spatial mask for a region of non-interest; These are the weighting coefficients corresponding to the non-interest regions, used to control the suppression strength of interference regions. Indicates the region of non-interest. Indicates the subscript index.

4. The ROI-adaptive multi-voxel spectral reconstruction method according to claim 1, characterized in that, In step S4, the target reconstruction process is specifically implemented by performing a summation operation based on the inverse discrete Fourier transform: Based on the optimal voxel localization parameters, determine the phase correction term for each sampling point in K-space. ; Using the summation formula that includes the phase correction term, from K-space spectral data Directly calculate spatial domain signals ; The calculation logic performed by the summation formula is as follows: ； That is, for any point in space Its signal For all K-space sampling points of signal The result of the weighted summation, where the weighting term is a complex exponential function, is given by an exponential part that includes the spatial location-coded phase. With the phase correction term The superposition of.

5. The ROI-adaptive multi-voxel spectral reconstruction method according to claim 4, characterized in that, The phase correction item The optimal voxel localization parameters satisfy a linear relationship as follows: ； in, These represent the coordinates in the k-space in the three directions, respectively; It is the geometric offset of the scan block center relative to the gradient center, calculated based on the optimal voxel localization parameters; this phase correction term is introduced into the exponent term of the summation formula. This enables the translation of the reconstructed spectral voxel grid in the spatial domain, aligning it with the optimal voxel positioning parameters.

6. The ROI-adaptive multi-voxel spectral reconstruction method according to claim 1, characterized in that, In step S4, the target reconstruction process employs a sub-voxel resolution reconstruction method, specifically including: The K-space spectral data is zero-padded to increase the dimension of the K-space matrix; A Fourier transform is performed on the expanded K-space matrix to generate a sub-voxel grid image with higher apparent resolution. Based on the optimal voxel positioning parameters, locate the subvoxel in the subvoxel grid image that is spatially closest to the optimal voxel position; The data of this subvoxel is used as the optimal reconstructed spectrum for the target region of interest.

7. The ROI-adaptive multi-voxel spectral reconstruction method according to claim 1, characterized in that, The structural image contains multiple spatially independent regions of interest; The method executes steps S3 to S5 independently for each region of interest. For each region of interest, an independent set of optimal voxel localization parameters is calculated, and an independent target reconstruction process is performed for each, thereby obtaining multiple sets of reconstructed spectral data. Each set of reconstructed spectral data contains only the optimal spectral information for the corresponding region of interest.

8. The ROI-adaptive multi-voxel spectral reconstruction method according to claim 1, characterized in that, The multi-voxel spectral scanning employs either Cartesian sampling or non-Cartesian sampling. When Cartesian sampling is used, the point spread function is in the form of a Sinc function; When non-Cartesian sampling is used, the point spread function is determined according to the specific sampling trajectory, and in step S4, the data is first gridded before the target reconstruction process is performed, or phase modulation is directly applied to the non-Cartesian K-space data.

9. A multi-voxel spectral reconstruction system with ROI adaptive capability, characterized in that, include: The image and localization module is used to acquire structural images of the scanned object and determine the location information of at least one region of interest and interference region; The data acquisition module is used to control the magnetic resonance equipment to perform multi-voxel spectral scanning and acquire K-space spectral data; The parameter optimization module is used to calculate the optimal voxel localization parameters of the target region of interest based on the location information of the region of interest and the point spread function of the system. The reconstruction module is used to perform target reconstruction processing on the K-space spectral data based on the optimal voxel positioning parameters, using inverse discrete Fourier transform summation operation with phase correction term or K-space zero-adding operation. The analysis module is used to extract the optimal spectrum from the reconstructed data.

Images