System and method for automatic volume of interest prescription for multi-voxel brain proton spectroscopy acquisition and post-processing

Abstract

The invention relates to a system and method for automatic volume of interest prescription for multi-voxel brain proton spectroscopy acquisition and post-processing. A method for performing multi-voxel spectroscopy includes obtaining structural magnetic resonance imaging data of a brain of a subject acquired with a magnetic resonance imaging scanner. The method also includes performing skull dissection on the structural magnetic resonance imaging data to generate a skull dissection brain image. The method further includes generating a lesion core mask from the brain image using a trained deep learning based segmentation model. The method also includes locating a slice having a maximum lesion volume present in the lesion core mask. The method includes calculating a voxel volume that avoids aliasing from the slice based on the field of view. The method includes automatically selecting a volume of interest in the brain having both the lesion and normal brain tissue for a multi-voxel spectral scan by the magnetic resonance imaging scanner based on the brain image, the lesion core mask, and the voxel volume.

Description

Background Art

[0001] The subject matter disclosed herein relates to medical imaging, and more particularly, to systems and methods for automatic volume of interest prescription for multi-voxel brain proton spectroscopy acquisition and post-processing.

[0002] Non-invasive imaging techniques allow images of internal structures or features of a patient / subject to be obtained without performing an invasive procedure on the patient / subject. Specifically, such non-invasive imaging techniques rely on various physical principles (such as differential transmission of X-rays through a target volume, reflection of acoustic waves within a volume, paramagnetism of different tissues and materials within a volume, decomposition of target radionuclides within the body, etc.) to acquire data and construct an image or otherwise represent the observed internal features of the patient / subject.

[0003] During magnetic resonance imaging (MRI), when a substance such as human tissue is subjected to a uniform magnetic field (polarization field B0), the individual magnetic moments of the spins in the tissue attempt to align with the polarization field, but precess around it in a random order at their characteristic Larmor frequency. If the substance or tissue is subjected to a magnetic field (excitation field B1) that lies in the xy plane and is close to the Larmor frequency, the net alignment moment or "longitudinal magnetization" M z can be rotated or "tilted" into the xy plane to produce a net transverse magnetic moment M t After excitation signal B1 terminates, a signal is emitted by the excited spins, and this signal can be received and processed to form an image.

[0004] When these signals are used to generate images, magnetic field gradients (G x , G y and G z Typically, the area to be imaged is scanned in a series of measurement cycles in which these gradient fields are varied according to the particular positioning method being used. The resulting set of received nuclear magnetic resonance (NMR) signals is digitized and processed to reconstruct an image using one of the well-known reconstruction techniques.

[0005] Proton ( 1MRI spectroscopy is an important sequence for studying metabolites and distinguishing tumor lesions from non-tumor lesions in brain tumor imaging. Currently, two types of spectral acquisition are available: single-voxel and multi-voxel techniques. Using single-voxel techniques, users can acquire data from a single region of interest in the brain. Using multi-voxel techniques, users can acquire large volumes of interest (VOIs). However, acquiring such large VOIs requires additional post-processing to generate the individual spectrograms. The primary advantage of multi-voxel techniques is that metabolic information from normal brain regions can also be calculated from the acquired data, allowing users to compare metabolite profiles of lesions with those of normal brain tissue. The main challenge in acquiring good spectral data is planning the VOI. The operator must interpret all structural sequences used to identify lesions and then select appropriate slices for acquiring a multi-voxel spectral VOI that encompasses both the abnormality and adjacent or contralateral normal brain tissue. Care should be taken to ensure that the VOI does not include bone or air-tissue interfaces, as inclusion of these interfaces can degrade the spectral data due to B0 inhomogeneities. Therefore, planning multi-voxel spectroscopy sequences and post-processing multi-voxel spectroscopy require a high level of expertise. This expertise extends to the interpretation of the entire structural data to detect abnormalities and select appropriate slices from the structural sequence (e.g., T1, T2, T2 Flair, post-contrast T1, etc.). Inappropriate MR spectroscopy planning will degrade the metabolite profiles required to distinguish between various brain lesions and ultimately lead to rescanning of the patient. Summary of the Invention

[0006] The following is an overview of certain embodiments disclosed herein. It should be understood that these aspects are provided merely to provide the reader with a brief overview of these specific embodiments, and these aspects are not intended to limit the scope of the present disclosure. In fact, the present disclosure may encompass various aspects that may not be shown below.

[0007] In one embodiment, a computer-implemented method for performing multi-voxel spectroscopy is provided. The computer-implemented method includes obtaining structural magnetic resonance imaging data of a subject's brain acquired using a magnetic resonance imaging scanner at a processor. The computer-implemented method includes performing skull stripping on the structural magnetic resonance imaging data via the processor to generate a skull-stripped brain image. The computer-implemented method also includes generating a lesion core mask from the skull-stripped brain image using a trained deep learning-based segmentation model via the processor. The computer-implemented method even includes locating, via the processor, a slice with the largest lesion volume present in the lesion core mask. The computer-implemented method still includes calculating, via the processor, a voxel volume that avoids aliasing from the slice based on the field of view. The computer-implemented method further includes automatically selecting, via the processor, a volume of interest in the brain having both lesions and normal brain tissue based on the skull-stripped brain image, the lesion core mask, and the voxel volume for multi-voxel spectroscopy scanning performed by the magnetic resonance imaging scanner.

[0008] In another embodiment, a computer-implemented method for post-processing multi-voxel spectroscopy is provided. The computer-implemented method includes obtaining, at a processor, susceptibility-weighted image data of a subject's brain acquired during a multi-voxel spectroscopy scan using a magnetic resonance imaging scanner. The computer-implemented method also includes performing, via the processor, skull stripping on the susceptibility-weighted image data to generate a skull-stripped brain image. The computer-implemented method also includes generating, via the processor, a hemorrhage and calcification mask, a lesion core mask, and a normal brain mask from the skull-stripped brain image using a trained deep learning-based segmentation model. The computer-implemented method further includes automatically locating, via the processor, a maximum voxel with pathology lacking hemorrhage and calcification within a defined volume of interest based on both the lesion core mask and the hemorrhage and calcification mask, wherein the defined volume of interest includes both lesions and normal brain tissue. The computer-implemented method still further includes automatically locating, via the processor, corresponding voxels of healthy regions of the brain within the defined volume of interest based on both the normal brain mask and the hemorrhage and calcification mask. The computer-implemented method further includes providing, via the processor, a request to the user to complete a selection of voxels, the voxels comprising a largest voxel having pathology lacking hemorrhage and calcifications and corresponding voxels of a healthy region. The computer-implemented method further includes generating, via the processor, corresponding metabolite maps for both the largest voxel having pathology lacking hemorrhage and calcifications and the corresponding voxels of the healthy region upon completing the selection of the voxels.

[0009] In another embodiment, a system for performing multi-voxel spectroscopy is provided. The system includes a memory that encodes a processor-executable routine. The system also includes a processor configured to access the memory and execute the processor-executable routine, wherein the routine, when executed by the processor, causes the processor to perform an action. The action includes obtaining structural magnetic resonance imaging (MRI) data of a subject's brain acquired using a magnetic resonance imaging (MRI) scanner. The action also includes performing skull stripping on the structural MRI data to generate a skull-stripped brain image. The action also includes generating a lesion core mask from the skull-stripped brain image using a trained deep learning-based segmentation model. The action further includes locating a slice with the largest lesion volume within the lesion core mask. The action still further includes calculating a voxel volume that avoids aliasing from the slice based on the field of view. The action further includes automatically selecting a volume of interest in the brain containing both lesions and normal brain tissue for multi-voxel spectroscopy scanning performed by the MRI scanner based on the skull-stripped brain image, the lesion core mask, and the voxel volume. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] These and other features, aspects, and advantages of the present subject matter will be better understood upon reading the following detailed description with reference to the accompanying drawings, in which like characters represent like parts throughout, and in which:

[0011] Figure 1 shows a schematic diagram of a magnetic resonance imaging (MRI) system suitable for use with the presently disclosed technology;

[0012] Figure 2 shows a flow chart of a method for performing multi-voxel spectroscopy according to aspects of the present disclosure;

[0013] Figure 3 is a schematic diagram illustrating a process for prescribing a volume of interest for multi-voxel spectroscopy according to aspects of the present disclosure;

[0014] Figure 4 A flow chart illustrating a method for selecting a volume of interest according to aspects of the present disclosure is shown;

[0015] Figure 5 shows utilization of a volume of interest selection algorithm according to aspects of the present disclosure (e.g., utilizing a horizontal volume);

[0016] Figure 6 shows utilization of a volume of interest selection algorithm according to aspects of the present disclosure (e.g., utilizing a vertical volume);

[0017] Figure 7 A flow chart illustrating a method for post-processing of multi-voxel spectra according to aspects of the present disclosure is shown;

[0018] Figure 8 is a schematic diagram illustrating a process for post-processing of multi-voxel spectra according to aspects of the present disclosure;

[0019] Figure 9 shows a display of a metabolite map and corresponding locations of voxels placed in a contrast-enhanced portion of a pathology in accordance with aspects of the present disclosure; and

[0020] Figure 10 Shown are displays of metabolite maps and corresponding locations of contralateral voxels in a healthy brain region according to aspects of the present disclosure. DETAILED DESCRIPTION

[0021] One or more specific embodiments are described below. In order to provide a concise description of these embodiments, not all features of an actual implementation are necessarily described in the specification. It should be understood that in the development of any such actual implementation, as in any engineering or design project, many implementation-specific decisions must be made to achieve the developer's specific goals, such as complying with system-related and business-related constraints that may vary from implementation to implementation. Furthermore, it should be understood that such development efforts may be complex and time-consuming, but remain a routine task of design, fabrication, and manufacturing for those of ordinary skill having the benefit of this disclosure.

[0022] When introducing elements of various embodiments of the present subject matter, the articles "a," "an," "the," and "said" are intended to indicate that there are one or more elements. The terms "comprising," "including," and "having" are intended to be inclusive and mean that there may be additional elements besides the listed elements. Furthermore, any numerical examples in the following discussion are intended to be non-limiting, and thus the appended numerical values, ranges, and percentages are within the scope of the disclosed embodiments.

[0023] The deep learning (DL) methods discussed herein may be based on artificial neural networks and may therefore encompass one or more of the following: deep neural networks, fully interconnected networks, convolutional neural networks (CNNs), unfolded neural networks, perceptrons, codecs, recurrent networks, wavelet filter banks, u-nets, generative adversarial networks (GANs), dense neural networks, or other neural network architectures. Neural networks may include shortcuts, activations, batch normalization layers, and / or other features. These techniques are referred to herein as DL techniques, although the term may also be used with particular reference to the use of deep neural networks, which are neural networks with multiple layers.

[0024] As discussed herein, DL techniques (which may also be referred to as deep machine learning, hierarchical learning, or deep structured learning) are a branch of machine learning techniques that employ mathematical representations of data and artificial neural networks for learning and processing such representations. For example, DL methods can be characterized as using one or more algorithms to extract or model highly abstract concepts of a class of data of interest. This can be accomplished using one or more processing layers, where each layer typically corresponds to a different level of abstraction and may therefore employ or utilize different aspects of the initial data or the output of the previous layer (i.e., a hierarchical or cascaded structure of layers) as the target of the process or algorithm at a given layer. In the context of image processing or reconstruction, this can be characterized as different layers corresponding to different feature levels or resolutions in the data. Generally speaking, the processing of one representation space to the next level representation space can be considered a "stage" of the process. Each stage of the process can be performed by a separate neural network or by different parts of a larger neural network.

[0025] The present disclosure provides systems and methods for performing multivoxel spectroscopy. Specifically, the present disclosure provides systems and methods for automatically identifying and selecting slices with lesions from structural magnetic resonance imaging data (e.g., T1, T2, T2 Flair, post-contrast T1, etc.), and prescribing a volume of interest that includes both the abnormality and adjacent or contralateral brain tissue (excluding bone and air-tissue interfaces). Exclusion of bone and air-tissue interfaces reduces magnetic field inhomogeneities, which can degrade metabolite map quality.

[0026] For example, the disclosed systems and methods include obtaining, at a processor, structural magnetic resonance imaging data (e.g., T1-weighted images, T2-weighted images, fluid-attenuated inversion recovery (FLAIR) images, and post-contrast T1-weighted images) of a subject's brain acquired using a magnetic resonance imaging scanner. The disclosed systems and methods also include performing skull stripping on the structural magnetic resonance imaging data via the processor to generate a skull-stripped brain image. The disclosed systems and methods also include generating, via the processor, a lesion core mask from the skull-stripped brain image using a trained deep learning-based segmentation model. The disclosed systems and methods even include locating, via the processor, a slice with the largest lesion volume present in the lesion core mask. The disclosed systems and methods still further include calculating, via the processor, a voxel volume that avoids aliasing from the slice based on a field of view. The disclosed systems and methods also include automatically selecting, via the processor, a volume of interest in the brain having both lesions and normal brain tissue based on the skull-stripped brain image, the lesion core mask, and the voxel volume for multi-voxel spectroscopy scanning performed by the magnetic resonance imaging scanner.

[0027] In some embodiments, automatically selecting the volume of interest includes identifying, via the processor, using a brain mask, in which hemisphere of the brain the core of the lesion is located. In some embodiments, automatically selecting the volume of interest also includes determining, via the processor, whether to use a horizontal volume of interest or a vertical volume of interest in the brain tissue.

[0028] In some embodiments, a brain tissue mask is used to determine whether to utilize a horizontal volume of interest or a vertical volume of interest. This process includes, via a processor, examining a graph of a horizontal line as thick as a voxel volume, the voxel volume originating from the brain tissue mask of the hemisphere where the core of the lesion is located, and passing through the center of the core to the end of the brain tissue mask to determine whether the intensity decreases along the horizontal line. In some embodiments, determining whether to utilize a horizontal volume of interest or a vertical volume of interest in the brain tissue mask further includes, via the processor, determining that the horizontal volume of interest can be utilized when the intensity does not decrease along the horizontal line. In some embodiments, determining whether to utilize a horizontal volume of interest or a vertical volume of interest in the brain tissue mask further includes, via the processor, rotating the horizontal line 90 degrees to become a vertical line when the intensity does decrease along the horizontal line. In some embodiments, determining whether to utilize a horizontal volume of interest or a vertical volume of interest in the brain tissue mask further includes, via the processor, determining whether the intensity decreases along the vertical line. In some embodiments, determining whether to utilize a horizontal volume of interest or a vertical volume of interest in the brain tissue mask further includes, via the processor, determining that the vertical volume of interest can be utilized when the intensity does not decrease along the vertical line.

[0029] In certain embodiments, the disclosed systems and methods include providing, via the processor, a notification recommending a single-voxel spectroscopy scan of the brain rather than a multi-voxel spectroscopy scan when intensity decreases along both horizontal and vertical lines. In certain embodiments, automatically selecting a volume of interest further includes determining, via the processor, the number of voxels to utilize in the volume of interest. In certain embodiments, the disclosed systems and methods include causing, via the processor, acquisition of susceptibility-weighted image data of the brain of a subject for the volume of interest during a multi-voxel spectroscopy scan using a magnetic resonance imaging scanner.

[0030] The disclosed systems and methods include obtaining, at a processor, susceptibility-weighted image data of a subject's brain acquired during a multi-voxel spectroscopy scan using a magnetic resonance imaging scanner. The disclosed systems and methods also include performing, via the processor, skull stripping on the susceptibility-weighted image data to generate a skull-stripped brain image. The disclosed systems and methods also include utilizing, via the processor, a trained deep learning-based segmentation model to generate a hemorrhage and calcification mask, a lesion core mask, and a normal brain mask from the skull-stripped brain image. The disclosed systems and methods further include automatically locating, via the processor, a maximum voxel having pathology lacking hemorrhage and calcification within a defined volume of interest based on both the lesion core mask and the hemorrhage and calcification mask, wherein the defined volume of interest comprises both lesions and normal brain tissue. The disclosed systems and methods still further include automatically locating, via the processor, corresponding voxels of a healthy region of the brain within the defined volume of interest based on both the normal brain mask and the hemorrhage and calcification mask. The disclosed systems and methods further include providing, via the processor, a request to the user to complete a selection of voxels comprising a maximum voxel having pathology lacking hemorrhage and calcifications and corresponding voxels of healthy regions. In certain embodiments, the request includes presenting to the user an option to edit or add more voxels. The disclosed systems and methods further include generating, via the processor, corresponding metabolite maps for both the maximum voxel having pathology lacking hemorrhage and calcifications and the corresponding voxels of healthy regions upon completing the selection of voxels. In certain embodiments, the disclosed systems and methods include causing the corresponding metabolite maps and corresponding positions of both the maximum voxel having pathology lacking hemorrhage and calcifications and the corresponding voxels of healthy regions to be displayed on a reference image of the subject's brain, the reference image being used for prescription of a volume of interest for a multi-voxel spectral scan.

[0031] The disclosed embodiments assist users in detecting pathologies and selecting slices for spectral planning. The disclosed embodiments also reduce planning time and avoid unnecessary spectral rescans due to lack of expertise among first-time MRI users. The disclosed embodiments help avoid intravoxel magnetic field inhomogeneity artifacts by excluding bone and air-tissue interfaces when prescribing spectral volumes of interest. The disclosed embodiments enable automatic selection of the most appropriate voxels for study, thereby facilitating MRI reporting.

[0032] Although the techniques described below are used to acquire multivoxel spectra of brain tumors, the techniques can be used to acquire multivoxel spectra of other pathologies. For example, the disclosed techniques can be used to acquire spectra of the bilateral hippocampus in the case of epilepsy. The disclosed techniques can also be used for metabolic disorders, demyelinating diseases, neurodegenerative diseases, and other applications. For example, a user can select a region of interest to utilize these techniques from a drop-down menu (e.g., the hippocampus in epilepsy, the parietal cortex or basal ganglia in the case of a suspected metabolic disorder in the brain, etc.).

[0033] Considering the above, Figure 1 , a magnetic resonance imaging (MRI) system 100 is schematically illustrated as including a scanner 102, scanner control circuitry 104, and system control circuitry 106. According to embodiments described herein, the MRI system 100 is generally configured to perform MR imaging.

[0034] The system 100 also includes a remote access and storage system or device, such as a picture archiving and communication system (PACS) 108, or other devices, such as teleradiology equipment, that enable on-site or off-site access to data acquired by the system 100. In this way, magnetic resonance data can be acquired and then processed and evaluated on-site or off-site. Although the magnetic resonance imaging system 100 can include any suitable scanner or detector, in the illustrated embodiment, the system 100 includes a whole-body scanner 102 having a housing 120 through which an aperture 122 is formed. An examination table 124 can be moved into the aperture 122 to allow a patient 126 (e.g., a subject) to be positioned therein for imaging selected anatomical structures within the patient's body.

[0035] The scanner 102 includes a series of associated coils for generating a controlled magnetic field for exciting gyromagnetic material within the anatomy of the patient being imaged. Specifically, a primary magnetic coil 128 is provided for generating a primary magnetic field B0 that is generally aligned with the orifice 122. A series of gradient coils 130, 132, and 134 allow for the generation of controlled gradient magnetic fields during an examination sequence for position encoding of certain gyromagnetic nuclei within the patient 126. A radio frequency (RF) coil 136 (e.g., an RF transmit coil) is configured to generate radio frequency pulses for exciting certain gyromagnetic nuclei within the patient. In addition to the coils that may be located locally within the scanner 102, the system 100 also includes a set of receive coils or RF receive coils 138 (e.g., a coil array) configured for placement proximal to (e.g., against) the patient 126. For example, the receive coils 138 may include a cervical / thoracic / lumbar (CTL) coil, a head coil, a single-sided spine coil, and the like. Generally, receive coil 138 is placed near or on top of patient 126 to receive the weak RF signals (relative to the transmit pulses generated by the scanner coils) generated by certain gyromagnetic nuclei in the patient's body when patient 126 returns to his or her relaxed state.

[0036] The various coils of the system 100 are controlled by external circuitry to generate the desired fields and pulses and to read the emissions from the gyromagnetic material in a controlled manner. In the illustrated embodiment, a main power supply 140 provides power to the primary field coil 128 to generate the primary magnetic field Bo. A power input (e.g., power from a utility or grid), a power distribution unit (PDU), a power supply (PS), and a drive circuit 150 can collectively provide power to pulse the gradient field coils 130, 132, and 134. The drive circuit 150 can include amplification and control circuitry for supplying current to the coils as defined by the digitized pulse sequence output by the scanner control circuitry 104.

[0037] Another control circuit 152 is provided for regulating the operation of the RF coil 136. Circuit 152 includes a switching device for alternating between an active mode of operation and a passive mode of operation, in which the RF coil 136 transmits and does not transmit signals, respectively. Circuit 152 also includes amplification circuitry configured to generate RF pulses. Similarly, the receive coil 138 is connected to a switch 154 that switches the receive coil 138 between a receive mode and a non-receive mode. Thus, in the receive mode, the receive coil 138 resonates with RF signals generated by gyromagnetic nuclei released from the patient 126, and in the non-receive mode, they do not resonate with RF energy from the transmit coil (i.e., coil 136) to prevent undesirable operation. Additionally, the receive circuit 156 is configured to receive data detected by the receive coil 138 and may include one or more multiplexing and / or amplification circuits.

[0038] It should be noted that while the scanner 102 and the control / amplification circuitry are illustrated as being coupled by a single line, in practice, multiple such lines may exist. For example, separate lines may be used for control, data communication, power transmission, and the like. Furthermore, appropriate hardware may be provided along each type of line to properly process data and current / voltage. In practice, various filters, digitizers, and processors may be provided between the scanner and either or both of the scanner control circuitry 104 and the system control circuitry 106.

[0039] As shown, the scanner control circuit system 104 includes an interface circuit 158 that outputs signals for driving the gradient field coils and the RF coils and for receiving data representing magnetic resonance signals generated during an examination sequence. The interface circuit 158 is coupled to a control and analysis circuit 160. Based on a defined protocol selected via the system control circuit 106, the control and analysis circuit 160 executes commands for driving the circuits 150 and 152.

[0040] The control and analysis circuitry 160 is also used to receive magnetic resonance signals and perform subsequent processing before transmitting the data to the system control circuitry 106. The scanner control circuitry 104 also includes one or more memory circuits 162 that store configuration parameters, pulse sequence descriptions, examination results, etc. during operation.

[0041] Interface circuitry 164 is coupled to control and analysis circuitry 160 for exchanging data between scanner control circuitry 104 and system control circuitry 106. In certain embodiments, control and analysis circuitry 160, although illustrated as a single unit, may include one or more hardware devices. System control circuitry 106 includes interface circuitry 166 that receives data from scanner control circuitry 104 and transmits data and commands back to scanner control circuitry 104. Control and analysis circuitry 168 may include a CPU in a general-purpose or special-purpose computer or workstation. Control and analysis circuitry 168 is coupled to memory circuitry 170 to store programming code for operating MRI system 100 and to store processed image data for later reconstruction, display, and transmission. The programming code may execute one or more algorithms that, when executed by a processor, are configured to generate a variety of data for training a deep learning-based segmentation model as described below. In some embodiments, the memory circuit 170 can store one or more neural networks (e.g., a deep learning-based segmentation network for generating a brain image (or skull-stripped image), a lesion core mask, a hemorrhage and calcification mask, and a healthy brain mask (i.e., one without brain abnormalities). In some embodiments, the disclosed techniques can occur on a separate computing device having processing circuitry and memory circuitry.

[0042] Additional interface circuitry 172 may be provided for exchanging image data, configuration parameters, and the like with external system components, such as the remote access and storage device 108. Finally, the system control and analysis circuitry 168 may be communicatively coupled to various peripheral devices for facilitating operator interface and generating hard copies of reconstructed images. In the illustrated embodiment, these peripheral devices include a printer 174, a monitor 176, and a user interface 178, which may include devices such as a keyboard, a mouse, a touch screen (e.g., integral to the monitor 176), and the like.

[0043] Figure 2 A flow chart of a method 180 for performing multi-voxel spectroscopy is shown. One or more steps of the method 180 may be performed by Figure 1 One or more of the steps of method 180 may be performed simultaneously or sequentially. Figure 2One or more (and in some cases) all of the steps of method 180 may be performed automatically.

[0044] Method 180 includes obtaining structural magnetic resonance imaging data (e.g., T1-weighted images, T2-weighted images, FLAIR images, post-contrast T1-weighted images) of a subject's brain acquired using a magnetic resonance imaging scanner (box 182). Method 180 also includes normalizing the structural magnetic resonance imaging data (box 184). In some embodiments, normalizing the structural magnetic resonance imaging data includes performing an N4 bias field correction to correct for low-frequency intensity non-uniformities (e.g., bias fields or gain fields) present in the structural magnetic resonance imaging data. Method 180 also includes aligning the structural magnetic resonance imaging data to a standard brain atlas (e.g., to have a common coordinate system) (box 186). Method 180 even includes performing skull stripping (i.e., removing or extracting skull and other non-brain tissue) on the structural magnetic resonance imaging data to generate a brain image (or a skull stripped image) (box 188).

[0045] Method 180 also includes generating a lesion core mask from the brain image using the trained deep learning-based segmentation model 189 (box 190). In some embodiments, in the lesion core mask, gadolinium-enhanced tumors or lesions are marked, and necrotic and non-enhanced tumor cores and edema are also marked separately. Method 180 even includes locating the slice with the largest lesion volume present in the lesion core mask (box 192). Method 180 also includes calculating a voxel volume (e.g., for prescription) that avoids aliasing from the slice based on the field of view (FOV) (box 194). Method 180 also includes automatically selecting a volume of interest (VOI) in the brain having both lesions and normal brain tissue based on the skull-stripped brain image, the lesion core mask, and the voxel volume for a multi-voxel spectroscopy scan performed by a magnetic resonance imaging scanner (box 196). For example, as described in more detail below, a volume of interest selection algorithm can be used to select the volume of interest. The method 180 includes acquiring susceptibility-weighted (SWI) image data of a brain of a subject for a volume of interest (eg, a prescribed volume of interest) during a multi-voxel spectroscopy scan using a magnetic resonance imaging scanner (block 198 ).

[0046] Figure 3 is a schematic diagram illustrating the process for prescribing a volume of interest for multi-voxel spectroscopy. Figure 3As shown, structural MRI data 200 of a subject's brain acquired using a magnetic resonance imaging scanner is obtained. As depicted, the structural MRI data 200 includes a T1-weighted image volume 202, a T2-weighted image volume 204, a FLAIR image volume 206, and a post-contrast T1-weighted image volume 208. As shown by reference numeral 210, an N4 bias field correction is performed on the structural MRI data 200 to correct for low-frequency intensity non-uniformities (e.g., bias fields or gain fields) present in the structural MRI data 200. As shown by reference numeral 212, the normalized structural MRI data is registered to a standard brain atlas (e.g., to have a common coordinate system). As shown by reference numeral 214, skull stripping is performed on the normalized and registered structural MRI data 200 to generate a brain image (or a skull-stripped mask).

[0047] The brain image is input into a deep learning-based segmentation model 216, which outputs a lesion core mask 218. As depicted in the lesion core mask 218, gadolinium-enhancing tumors or lesions are marked, and necrotic and non-enhancing tumor cores and edema are also marked separately. As shown in reference numeral 220, a slice with the largest lesion volume present in the lesion core mask is located or selected. As shown in reference numeral 222, a voxel volume that avoids aliasing from the slice is selected or calculated based on the field of view (FOV) (e.g., for prescription) (box 194). A volume of interest (VOI) selection algorithm is then used to automatically select a volume of interest in the brain having both lesions and normal brain tissue based on the skull-stripped brain image, the lesion core mask, and the voxel volume for multi-voxel spectroscopy scanning by a magnetic resonance imaging scanner, as shown in reference numeral 226. The selected volume of interest (or prescribed volume of interest) is then sent to a host for multi-voxel spectroscopy scanning (e.g., by a magnetic resonance scanner).

[0048] Figure 4 A flow chart of a method 228 for selecting a volume of interest (e.g., by a volume of interest selection algorithm) in a brain having both lesions and normal brain tissue for multi-voxel spectroscopy is shown. One or more steps of the method 228 may be performed by Figure 1 One or more of the steps of method 228 may be performed simultaneously or sequentially. Figure 4 One or more (and in some cases) all of the steps of method 228 may be performed automatically.

[0049] Method 228 includes utilizing a brain mask (e.g., Figure 2The method 228 also includes determining whether to utilize a horizontal volume of interest or a vertical volume of interest in the brain tissue mask, as shown by reference numeral 232. Specifically, in determining whether to utilize a horizontal volume of interest or a vertical volume of interest in the brain tissue mask, the method 228 includes studying the voxel volume (e.g., in FIG. 230 ) and determining whether to utilize a horizontal volume of interest or a vertical volume of interest in the brain tissue mask. Figure 2 A graph is drawn of a horizontal line as thick as the voxel volume determined in block 194 of method 188 of FIG. 1 , which voxel volume originates from a brain tissue mask from the hemisphere of the brain where the core of the lesion is located and passes through the center of the core to the end of the brain tissue mask to determine whether the intensity decreases along the horizontal line (block 234). In some embodiments, upon determining whether to utilize a horizontal volume of interest or a vertical volume of interest in the brain tissue mask, method 228 includes determining whether the horizontal volume of interest can be utilized when the intensity does not decrease along the horizontal line (block 236). In some embodiments, upon further determining whether to utilize a horizontal volume of interest or a vertical volume of interest in the brain tissue mask, method 236 includes rotating the horizontal line 90 degrees to become a vertical line when the intensity does decrease along the horizontal line (block 238). In some embodiments, upon determining whether to utilize a horizontal volume of interest or a vertical volume of interest in the brain tissue mask, method 228 includes determining whether the intensity decreases along the vertical line (block 240). In some embodiments, upon determining to utilize a horizontal volume of interest or a vertical volume of interest in the brain tissue mask, method 228 includes determining that a vertical volume of interest can be utilized when intensity does not decrease along a vertical line. (Block 242) In some embodiments, method 228 includes providing a notification (e.g., on an operator console or any device utilized by the user) recommending a single-voxel spectral scan of the brain rather than a multi-voxel spectral scan when intensity decreases along both the horizontal and vertical lines (Block 244).

[0050] Method 228 also includes determining the number of voxels to be utilized in the volume of interest (block 246). Specifically, the number of voxels is determined in both the x-direction and the y-direction. The x-direction and the y-direction are relative to the line (regardless of whether the line is oriented vertically or horizontally). For example, the x-direction is along the line, while the y-direction intersects the line. To identify the number of voxels along the y-direction, both the length of the plotted or drawn line and the voxel volume are utilized. To identify the number of voxels along the x-direction, initially select one voxel and successively add another voxel until no intensity drop is seen for the thickness of the line. The thickness of the line is the number of voxels multiplied by the voxel volume.

[0051] Figure 5 and Figure 6 Utilization of a volume of interest selection algorithm is shown. Figure 5Utilization of a volume of interest selection algorithm (e.g., utilizing a horizontal volume) is illustrated. In brain image 248, a dashed line 250 divides the two hemispheres of the brain, and an ellipse 252 indicates the lesion core. In brain image 254, a horizontal line 256 is drawn that is as thick as the voxel volume, originating from the brain image of the hemisphere where the lesion core is located, and passing through the center of the core to the end of the brain image, to determine whether the intensity decreases along the horizontal line. In this case, the intensity does not decrease, and a horizontal volume of interest is utilized, and the number of voxels to be utilized in volume of interest 258 is determined, as marked on brain image 260.

[0052] Figure 6 Utilization of a volume of interest selection algorithm is shown (e.g., utilizing a vertical volume). In brain image 262, a dotted line 264 divides the two hemispheres of the brain, and an ellipse 266 indicates the core of the lesion. In brain image 268, a horizontal line 270 is drawn that is as thick as the voxel volume, which originates from the brain image of the hemisphere of the brain where the core of the lesion is located, and passes through the center of the core to the end of the brain image to determine whether the intensity decreases along the horizontal line. In this case, the intensity decreases along horizontal line 270 due to air. Therefore, as shown in brain image 272, the horizontal line is rotated 90 degrees to become a vertical line 274, and the line is studied to see whether the intensity decreases along vertical line 274. In this case, a vertical volume of interest is utilized, and the number of voxels to be utilized in volume of interest 276 is determined, as marked on brain image 278.

[0053] Figure 7 A flow chart of a method 280 for post-processing a multi-voxel spectrum is shown. One or more steps of the method 280 may be performed by Figure 1 One or more of the steps of method 280 may be performed simultaneously or sequentially. Figure 7 One or more (and in some cases) all of the steps of method 280 may be performed automatically.

[0054] Method 280 includes obtaining susceptibility-weighted image data of a subject's brain acquired during a multi-voxel spectroscopy scan using a magnetic resonance imaging scanner (block 282). Method 280 also includes normalizing the susceptibility-weighted image data (block 284). In some embodiments, normalizing the susceptibility-weighted image data includes performing an N4 bias field correction to correct for low-frequency intensity inhomogeneities (e.g., bias or gain fields) present in the susceptibility-weighted image data. Method 280 also includes registering the susceptibility-weighted image data to a standard brain atlas (e.g., to have a common coordinate system) (block 286). Method 280 further includes performing skull stripping (i.e., removing or extracting skull bone and other non-brain tissue) on the susceptibility-weighted image data to generate a brain image (or skull-stripped image) (block 288).

[0055] The method 280 also includes generating a hemorrhage and calcification mask using the trained deep learning-based segmentation model 289. In this method, the lesion core mask and the abnormality-free brain mask of the brain image from the segmentation model 189 are reused. The method 280 further includes automatically locating the largest voxel with pathology lacking hemorrhage and calcification within a specified volume of interest based on the lesion core mask and the hemorrhage and calcification mask, wherein the specified volume of interest has both lesions and normal brain tissue (block 292). The method 280 still further includes automatically locating a corresponding voxel (e.g., a contralateral voxel) of a healthy region of the brain (also lacking hemorrhage and calcification) within the specified volume of interest based on both the abnormality-free brain mask and the hemorrhage and calcification mask (block 294).

[0056] The method 280 further includes providing a request to the user to complete the selection of voxels, the voxels including the largest voxel with pathology lacking hemorrhage and calcifications and the corresponding voxels of the healthy region (box 296). In some embodiments, the request includes presenting the user with an option to edit or add more voxels. The method 280 also includes generating corresponding metabolite maps for both the largest voxel with pathology lacking hemorrhage and calcifications and the corresponding voxels of the healthy region upon completing the selection of the voxels (box 298). In some embodiments, the method 280 includes causing the corresponding metabolite maps and the corresponding locations of both the largest voxel with pathology lacking hemorrhage and calcifications and the corresponding voxels of the healthy region to be displayed on a reference image of the brain of the subject for the volume of interest specified for the multi-voxel spectroscopy scan (box 300).

[0057] Figure 8 is a schematic diagram showing the process for post-processing of multi-voxel spectra. Figure 8 As shown, a multispectral scan is obtained using a magnetic resonance imaging scanner (e.g., for Figure 2Susceptibility-weighted image data 302 of a subject's brain acquired during a period of 30 days (a predetermined volume of interest determined in method 180 of FIG. 1 ) is obtained. As shown by reference numeral 304, an N4 bias field correction is performed on the susceptibility-weighted image data 302 to correct for low-frequency intensity non-uniformities (e.g., bias fields or gain fields) present in the susceptibility-weighted image data 302. As shown by reference numeral 306, the normalized susceptibility-weighted image data is registered to a standard brain atlas (e.g., to have a common coordinate system). As shown by reference numeral 308, skull stripping is performed on the normalized and registered susceptibility-weighted image data to generate a brain image (or skull-stripped image).

[0058] A brain image is input into a deep learning-based segmentation model 310, which outputs a hemorrhage and calcification mask 312, a lesion core mask 314, and a mask 316 of the normal brain (i.e., a healthy brain mask). The lesion core mask 314 is used to generate the normal brain mask 316. As shown in reference numeral 318, based on both the lesion core mask and the hemorrhage and calcification mask, the largest voxel with pathology lacking hemorrhage and calcification is automatically located within a specified volume of interest, where the specified volume of interest contains both lesions and normal brain tissue. Specifically, the lesion core mask 314 is multiplied by the hemorrhage and calcification mask 312. As shown in reference numeral 318, based on both the normal brain mask and the hemorrhage and calcification mask, the corresponding voxel (e.g., the contralateral voxel) of a healthy region of the brain (also lacking hemorrhage and calcification) is automatically located within the specified volume of interest. Specifically, the normal brain mask 316 is multiplied by the hemorrhage and calcification mask 312.

[0059] As shown in reference numeral 322, a request is provided to the user to complete the selection of voxels, including the largest voxel with pathology lacking hemorrhage and calcification and the corresponding voxel in the healthy region. In some embodiments, the request includes an option presented to the user to edit or add more voxels. As shown in reference numeral 324, upon completion of the voxel selection, corresponding metabolite maps are generated for both the largest voxel with pathology lacking hemorrhage and calcification and the corresponding voxel in the healthy region, and then displayed on a reference image of the subject's brain along the corresponding positions of both the largest voxel with pathology lacking hemorrhage and calcification and the corresponding voxel in the healthy region, the reference image being used to prescribe a volume of interest for a multi-voxel spectroscopy scan.

[0060] Figure 9 and Figure 10 Post-processing automation for multi-voxel spectra is shown. Figure 9A display of a metabolite map 326 and the corresponding positions of voxels placed in the contrast-enhanced portion of the pathology are shown. Magnetic resonance images 328, 330, 332, and 334 are reference images of the subject's brain used for prescribing a volume of interest for a multi-voxel spectroscopy scan. Magnetic resonance images 328 and 330 are axial views. Magnetic resonance images 330 and 332 are sagittal and coronal views, respectively. A prescribed volume of interest 336 is indicated in magnetic resonance images 328 and 330. Crosses 338 indicate the positions of voxels placed in the contrast-enhanced portion of the pathology. A grid 340 represents the voxels within the volume of interest 336. The prescribed volume of interest 336 includes both the pathology and adjacent or contralateral healthy brain tissue.

[0061] Figure 10 A display of the metabolite map 342 is shown, along with the contralateral voxels in the healthy brain region (relative to Figure 9 Magnetic resonance images 344, 346, 348, and 350 are reference images of the subject's brain for prescribing a volume of interest for a multi-voxel spectroscopy scan. Magnetic resonance images 344 and 346 are axial views. Magnetic resonance images 348 and 350 are sagittal and coronal views, respectively. A prescribed volume of interest 352 is indicated in magnetic resonance images 344 and 346. Cross 338 indicates the location of the voxel placed in the contrast-enhanced portion of the pathology. Grid 356 represents the voxels within volume of interest 352. The prescribed volume of interest 352 includes both the pathology and adjacent or contralateral healthy brain tissue.

[0062] Technical benefits of the disclosed subject matter include assisting users in detecting pathologies and selecting slices for spectral planning. The disclosed embodiments also reduce planning time and avoid unnecessary spectral rescans due to a lack of expertise among first-time MRI users. The disclosed embodiments help avoid intravoxel magnetic field inhomogeneity artifacts by excluding bone and air-tissue interfaces when prescribing spectral volumes of interest. The disclosed embodiments enable automatic selection of the most appropriate voxels for study, thereby aiding MRI reporting.

[0063] The technology presented and claimed herein is referred to and applied to physical and concrete examples of a practical nature that clearly improves upon the state of the art and, therefore, is not abstract, intangible, or purely theoretical. Furthermore, if any claim appended to the end of this specification contains one or more elements designated as "means for [performing] the function of ..." or "steps for [performing] the function of ...," it is intended that such elements be construed under 35 U.S.C. § 112(f). However, for any claim containing elements designated in any other manner, it is not intended that such elements be construed under 35 U.S.C. § 112(f).

[0064] This written description uses examples to disclose the subject matter, including the best mode, and also to enable any person skilled in the art to practice the subject matter, including making and using any devices or systems and performing any included methods. The patentable scope of the subject matter is defined by the claims, and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims if they have structural elements that do not differ from the literal language of the claims, or if they include equivalent structural elements with insignificant differences from the literal language of the claims.

Claims

1. A computer-implemented method for performing multi-voxel spectroscopy, the computer-implemented method comprising: obtaining, at a processor, structural magnetic resonance imaging data of a subject's brain acquired using a magnetic resonance imaging scanner; performing, via the processor, skull stripping on the structural magnetic resonance imaging data to generate a skull stripped brain image; generating, via the processor, a lesion core mask from the skull-stripped brain image using a trained deep learning-based segmentation model; locating, via the processor, a slice having a maximum lesion volume present in the lesion core mask; calculating, via the processor, a voxel volume that avoids aliasing from the slice based on a field of view; Based on the skull-stripped brain image, the lesion core mask and the voxel volume, a volume of interest in the brain having both the lesion and normal brain tissue is automatically selected via the processor for multi-voxel spectroscopy scanning by the magnetic resonance imaging scanner. 2 . The computer-implemented method of claim 1 , wherein automatically selecting the volume of interest comprises identifying, via the processor, using a brain mask, in which hemisphere of the brain the core of the lesion is located. 3 . The computer-implemented method of claim 2 , wherein automatically selecting the volume of interest further comprises determining, via the processor, whether to utilize a horizontal volume of interest or a vertical volume of interest in a brain tissue mask.

4. The computer-implemented method of claim 3 , wherein determining whether to utilize the horizontal volume of interest or the vertical volume of interest in the brain tissue mask comprises studying, via the processor, a graph of a horizontal line as thick as the voxel volume originating from the brain tissue mask from the hemisphere of the brain in which the core of the lesion is located and passing through the center of the core to an end of the brain tissue mask to determine whether intensity decreases along the horizontal line. 5 . The computer-implemented method of claim 4 , wherein determining to utilize the horizontal volume of interest or the vertical volume of interest in the brain tissue mask further comprises determining, via the processor, that the horizontal volume of interest can be utilized when the intensity does not decrease along the horizontal line.

6. The computer-implemented method of claim 4, wherein determining to utilize the horizontal volume of interest or the vertical volume of interest in the brain tissue mask further comprises rotating, via the processor, the horizontal line 90 degrees to become a vertical line when the intensity does decrease along the horizontal line. 7 . The computer-implemented method of claim 6 , wherein determining whether to utilize the horizontal volume of interest or the vertical volume of interest in the brain tissue mask further comprises determining, via the processor, whether the intensity decreases along the vertical line.

8. The computer-implemented method of claim 7, wherein determining to utilize the horizontal volume of interest or the vertical volume of interest in the brain tissue mask further comprises determining, via the processor, that the vertical volume of interest can be utilized when the intensity does not decrease along the vertical line.

9. The computer-implemented method of claim 7, further comprising providing, via the processor, a notification recommending a single-voxel spectral scan of the brain rather than a multi-voxel spectral scan when the intensity decreases along both the horizontal line and the vertical line.

10. The computer-implemented method of claim 3, wherein automatically selecting the volume of interest further comprises determining, via the processor, a number of voxels to utilize in the volume of interest.

11. The computer-implemented method of claim 1 , further comprising causing, via the processor, acquisition of susceptibility-weighted image data of the brain of the subject for the volume of interest during a multi-voxel spectroscopy scan using the magnetic resonance imaging scanner.

12. The computer-implemented method of claim 11 , further comprising: obtaining, via the processor, the susceptibility-weighted image data; performing, via the processor, skull stripping on the susceptibility-weighted image data to generate a second skull stripped brain image; generating, via the processor, a hemorrhage and calcification mask, a second lesion core mask, and an abnormality-free brain mask from the second skull-stripped brain image using the trained deep learning-based segmentation model; automatically locating, via the processor, within the volume of interest, a largest voxel having pathology lacking hemorrhage and calcification based on both the second lesion core mask and the hemorrhage and calcification mask; automatically locating, via the processor, corresponding voxels of a healthy region of the brain within the volume of interest based on both the abnormality-free brain mask and the hemorrhage and calcification mask; providing, via the processor, a request to a user to complete selection of voxels including the largest voxel having pathology lacking hemorrhage and calcification and the corresponding voxels of the healthy region; as well as Upon completion of the selection of voxels, corresponding metabolite maps are generated via the processor for both the largest voxel having pathology lacking hemorrhage and calcification and the corresponding voxel of the healthy region.

13. The computer-implemented method of claim 12 , further comprising causing, via the processor, display of the corresponding metabolite map and the corresponding locations of both the largest voxel with pathology lacking hemorrhage and calcification and the corresponding voxel of the healthy region on a reference image of the brain of the subject, the reference image being used for prescription of the volume of interest for the multi-voxel spectroscopy scan.

14. A computer-implemented method for post-processing of a multi-voxel spectrum, the computer-implemented method comprising: obtaining, at a processor, susceptibility-weighted image data of a subject's brain acquired during a multi-voxel spectroscopy scan using a magnetic resonance imaging scanner; performing, via the processor, skull stripping on the susceptibility-weighted image data to generate a skull stripped brain image; generating, via the processor, a hemorrhage and calcification mask, a lesion core mask, and an abnormality-free brain mask from the skull-stripped brain image using a trained deep learning-based segmentation model; automatically locating, via the processor, the largest voxel having pathology lacking hemorrhage and calcification within a defined volume of interest based on both the lesion core mask and the hemorrhage and calcification masks, wherein the defined volume of interest has both lesions and normal brain tissue; automatically locating, via the processor, corresponding voxels of a healthy region of the brain within a specified volume of interest based on both the abnormality-free brain mask and the hemorrhage and calcification masks; providing, via the processor, a request to a user to complete selection of voxels including the largest voxel having pathology lacking hemorrhage and calcification and the corresponding voxels of the healthy region; as well as Upon completion of the selection of voxels, corresponding metabolite maps are generated via the processor for both the largest voxel having pathology lacking hemorrhage and calcification and the corresponding voxel of the healthy region.

15. The computer-implemented method of claim 14 , further comprising causing, via the processor, display of the corresponding metabolite map and the corresponding locations of both the largest voxel with pathology lacking hemorrhage and calcification and the corresponding voxel of the healthy region on a reference image of the brain of the subject, the reference image being used for prescription of a specified volume of interest for the multi-voxel spectroscopy scan.

16. The computer-implemented method of claim 14, wherein the request includes an option presented to the user to edit or add more voxels.

17. A system for performing multi-voxel spectroscopy, the system comprising: a memory encoding processor-executable routines; and a processor configured to access the memory and to execute the processor-executable routine, wherein the processor-executable routine, when executed by the processor, causes the processor to: obtaining structural magnetic resonance imaging data of a subject's brain acquired using a magnetic resonance imaging scanner; performing skull stripping on the structural magnetic resonance imaging data to generate a first skull stripped brain image; generating a first lesion core mask from the first skull-stripped brain image using a trained deep learning-based segmentation model; locating a slice having a maximum lesion volume present in the lesion core mask; calculating a voxel volume that avoids aliasing from the slice based on the field of view; A volume of interest in the brain having both the lesion and normal brain tissue is automatically selected for multi-voxel spectroscopy scanning by the magnetic resonance imaging scanner based on the skull-stripped brain image, the lesion core mask, and the voxel volume.

18. The system of claim 17, wherein the processor-executable routine, when executed by the processor, further causes the processor to acquire susceptibility-weighted image data of the brain of the subject for the volume of interest during a multi-voxel spectroscopy scan using the magnetic resonance imaging scanner.

19. The system of claim 18, wherein the processor-executable routine, when executed by the processor, further causes the processor to: obtaining the susceptibility-weighted image data; performing skull stripping on the susceptibility-weighted image data to generate a second skull stripped brain image; generating a hemorrhage and calcification mask, a second lesion core mask, and an abnormality-free brain mask from the second skull-stripped brain image using the trained deep learning-based segmentation model; automatically locating the largest voxel having pathology lacking hemorrhage and calcification within the volume of interest based on both the second lesion core mask and the hemorrhage and calcification mask; automatically locating corresponding voxels of a healthy region of the brain within the volume of interest based on both the abnormality-free brain mask and the hemorrhage and calcification masks; providing a request to a user to complete selection of voxels including the largest voxel having pathology lacking hemorrhage and calcification and the corresponding voxels of the healthy region; and Upon completion of the selection of voxels, corresponding metabolite maps are generated for both the largest voxel with pathology lacking hemorrhage and calcification and the corresponding voxel of the healthy region.

20. The system of claim 19, wherein the processor-executable routine, when executed by the processor, further causes the processor to display the corresponding metabolite profile and the pathology having a lack of hemorrhage and calcification on a reference image of the brain of the subject. the respective positions of both the largest voxel and the corresponding voxel of the healthy area, The reference image is used for prescription of the volume of interest for the multi-voxel spectroscopy scan.

Images