Workflow management for labeling subject anatomy

Abstract

Systems and methods for workflow management for labeling subject anatomy are provided. The method includes obtaining at least one scout image of a subject anatomy using a low resolution medical imaging device. The method further includes labeling at least one anatomical point within the at least one scout image. The method further includes extracting a mask of the at least one scout image including the at least one anatomical point label using a machine learning module. The method further includes labeling at least one anatomical point on a high resolution image of the subject anatomy based on the at least one anatomical point within the scout image using the mask.

Description

Technical Field

[0001] This disclosure generally relates to improved medical imaging systems and methods, and more specifically to improved workflow management systems and methods for marking subject anatomy. Background Technology

[0002] Various medical imaging systems and methods are used to obtain images of a subject's anatomy for the purpose of diagnosing medical conditions. Magnetic resonance imaging (MRI) is a known medical imaging modality used to image different parts of the body and provides detailed images of soft tissues, abnormal tissues such as tumors, and other structures.

[0003] The entire workflow for an MRI scan of a subject's body can take anywhere from fifteen to ninety minutes, depending on the part of the subject's body being imaged. Obtaining high-resolution MR images requires the use of high-resolution MRI equipment. Furthermore, accurately labeling the parts of the subject's anatomy within the MR images is a manual and time-consuming process. Any errors in the manual labeling of MR images may necessitate relabeling the images, which further delays anatomical labeling.

[0004] A typical MR imaging workflow may involve acquiring several primary high-resolution images of the subject's anatomy using an MRI scanner, conducting a primary assessment of the subject's anatomy, and obtaining a final high-resolution MR image based on the primary assessment of the anatomy. The primary MR images of the subject require manual identification and labeling of the subject's anatomy, a process that is inconvenient for several reasons. Some of these inconveniences may include the time required for manual labeling of the subject's anatomy and operator errors in labeling parts of the anatomy. Another significant inconvenience is the use of image segmentation techniques to divide the image into several parts and label each part individually. Therefore, traditional MR imaging workflows are time-consuming, operator-centric, and based on segmentation techniques.

[0005] Artificial intelligence-based machine learning modules are used to automatically segment, identify, and label anatomical regions within a subject's body to reduce the amount of time required for labeling. Artificial intelligence (AI)-based machine learning techniques employ representation learning methods that allow the machine to acquire raw data and determine the representations needed for data classification. Machine learning uses the backpropagation algorithm to determine the structure within the dataset. Machine learning can leverage various multi-layered architectures and algorithms. Furthermore, the machine learning (ML) module can be trained on a set of expert classification data and used to identify features within the target dataset. The ML module can further label the target dataset, and the accuracy of labeling the target dataset determines the accuracy of training the ML module. In traditional MR imaging, the acquired primary high-resolution images are used to train the ML module for primary anatomical assessment; however, training a machine learning module using a machine learning module is not very cost-effective.

[0006] An improved workflow is needed to process medical images in reduced time for subject anatomy identification and labeling, without requiring high-resolution images for primary evaluation of anatomy and segmentation techniques. Summary of the Invention

[0007] This invention presents concepts described in more detail in specific embodiments. It should not be used to determine the essential features of the claimed subject matter, nor should it be used to limit the scope of the claimed subject matter. Its sole purpose is to present the concepts in a simplified form as a prelude to the more detailed description that follows.

[0008] According to one aspect of this disclosure, a method is disclosed. The method includes acquiring at least one locator image of a subject's anatomy using a low-resolution medical imaging device. The method further includes labeling at least one anatomical point within the at least one locator image. The method further includes extracting a mask of the at least one locator image including the labeling of the at least one anatomical point using a machine learning module. The method further includes using the mask to label at least one anatomical point on a high-resolution image of the subject's anatomy based on the at least one anatomical point within the locator image.

[0009] According to one aspect of this disclosure, a method for automated spine mapping is disclosed. The method includes acquiring at least one locator image of a subject's spine using a low-resolution medical imaging device. The method further includes tagging at least one anatomical point within the at least one locator image of the subject's spine. The method further includes extracting a mask of the at least one locator image of the subject's spine including the at least one anatomical point tagging using a machine learning module. The method further includes using the mask to tag at least one anatomical point in a high-resolution image of the subject's spine based on the at least one anatomical point within the locator image of the subject's spine.

[0010] According to another aspect of this disclosure, a system is disclosed. The system includes a low-resolution medical imaging device configured to acquire at least one locator image of a subject's anatomy. The system further includes a computer processor configured to receive the at least one locator image from the low-resolution medical imaging device. The computer processor is further configured to label at least one anatomical point within the at least one locator image. The computer processor is further configured to extract a mask of the at least one locator image containing the at least one anatomical point label using a machine learning module. The computer processor is further configured to use the mask to label at least one anatomical point on a high-resolution image of the subject's anatomy based on the at least one anatomical point within the locator image. Attached Figure Description

[0011] Figure 1 A schematic diagram of a magnetic resonance imaging (MRI) system according to one aspect of this disclosure is shown.

[0012] Figure 2 A human spinal cord or spine is shown according to one aspect of this disclosure.

[0013] Figure 3 A method for generating vertebral markers from a 2D locator image is shown according to one aspect of this disclosure.

[0014] Figure 4 This illustrates the point determination using a machine learning module according to one aspect of this disclosure.

[0015] Figure 5 A method for generating a mask overlay from a 2D spine image is shown according to one aspect of this disclosure.

[0016] Figure 6 This illustrates the use of an ML module to evaluate the error of marker points according to one aspect of this disclosure. Detailed Implementation

[0017] In the following specification and claims, references shall be defined as several terms having the following meanings.

[0018] The singular forms “a,” “one,” and “the” include multiple references unless the context clearly indicates otherwise.

[0019] As used herein, the term "non-transitory computer-readable medium" is intended to mean any tangible computer-based device implemented in any method or technology for short-term and long-term storage of information, such as computer-readable instructions, data structures, program modules and submodules, or other data in any device. Therefore, the methods described herein can be encoded as executable instructions embodied in tangible non-transitory computer-readable media, including but not limited to storage devices and / or memory devices. When executed by a processor, such instructions cause the processor to perform at least a portion of the methods described herein. Furthermore, as used herein, the term "non-transitory computer-readable medium" includes all tangible computer-readable media, including but not limited to non-transitory computer storage devices, including but not limited to volatile and non-volatile media, and removable and non-removable media such as firmware, physical and virtual storage, CD-ROMs, DVDs, and any other digital source (such as a network or the Internet), as well as digital devices not yet developed, with the sole exception of transient propagation signals.

[0020] As used herein, the terms “software” and “firmware” are used interchangeably and include any computer program stored in memory for execution by devices including, but not limited to, mobile devices, clusters, personal computers, workstations, clients and servers.

[0021] As used herein, the term “computer” and related terms (e.g., “computing device”, “computer system”, “processor”, “controller”) are not limited to integrated circuits referred to as computers in the art, but are broadly used to refer to at least one microcontroller, microcomputer, programmable logic controller (PLC), application-specific integrated circuit and other programmable circuits, and these terms are used interchangeably herein.

[0022] As used herein throughout the specification and claims, approximate language may be applied to modify any quantitative expression that allows for variation without altering its underlying function. Therefore, values ​​modified by one or more terms such as “about” and “substantially” are not limited to the specified precise values. In at least some cases, approximate language may correspond to the precision of the instrument used to measure the value. Scope limitations may be combined and / or interchanged herein and throughout the specification and claims; such scopes may be identified and include all subscopes contained herein, unless otherwise indicated by context or language.

[0023] An image contains several points that together form a complete image. The resolution of an image is affected by the number of points present per unit distance. For example, if an image contains 300 or more points per inch, it can be considered high-resolution. More than 300 points per inch will display all elements of the image. Conversely, if an image contains fewer than 300 points per inch, it is considered low-resolution. However, in the case of medical imaging, image resolution is defined according to the rules of in-plane and out-of-plane resolution. When the thickness of an image slice is less than six millimeters, a high-resolution image in medical imaging can be an image with an in-plane resolution of less than three millimeters.

[0024] Although this invention is explained in relation to magnetic resonance imaging (MRI) devices and images obtained therefrom, it can be applied to images obtained using other medical imaging devices, including but not limited to computed tomography (CT) systems, X-ray systems, positron emission tomography (PET) systems, and single-photon emission computed tomography (SPECT) systems.

[0025] According to one aspect of this disclosure, a method is disclosed. The method includes acquiring at least one locator image of a subject's anatomy using a low-resolution medical imaging device. The method further includes labeling at least one anatomical point within the at least one locator image. The method further includes extracting a mask of the at least one locator image including the labeling of the at least one anatomical point using a machine learning module. The method further includes using the mask to label at least one anatomical point on a high-resolution image of the subject's anatomy based on the at least one anatomical point within the locator image.

[0026] According to one aspect of this disclosure, a method for automated spine mapping is disclosed. The method includes acquiring at least one locator image of a subject's spine using a low-resolution medical imaging device. The method further includes tagging at least one anatomical point within the at least one locator image of the subject's spine. The method further includes extracting a mask of the at least one locator image of the subject's spine including the at least one anatomical point tagging using a machine learning module. The method further includes using the mask to tag at least one anatomical point in a high-resolution image of the subject's spine based on the at least one anatomical point within the locator image of the subject's spine.

[0027] According to another aspect of this disclosure, a system is disclosed. The system includes a low-resolution medical imaging device configured to acquire at least one locator image of a subject's anatomy. The system further includes a computer processor configured to receive the at least one locator image from the low-resolution medical imaging device. The computer processor is further configured to label at least one anatomical point within the at least one locator image. The computer processor is further configured to extract a mask of the at least one locator image including the at least one anatomical point label using a machine learning module. The computer processor is further configured to use the mask to label at least one anatomical point on a high-resolution image of the subject's anatomy based on the at least one anatomical point within the locator image.

[0028] Embodiments of this disclosure will now be described by way of example with reference to the accompanying drawings, wherein Figure 1This is a schematic diagram of a magnetic resonance imaging (MRI) system 10. The system can be controlled from an operator console 12, which includes an input device 13, a control panel 14, and a display screen 16. The input device 13 may be a mouse, joystick, keyboard, trackball, touchscreen, light bar, voice controller, and / or other input device. The input device 13 can be used for interactive geometry specification. The console 12 communicates with a computer system 20 via a link 18, which enables the operator to control the generation and display of images on the display screen 16. The link 18 can be a wireless or wired connection. The computer system 20 may include modules that communicate with each other via a backplane 20a. The modules of the computer system 20 may include, for example, an image processor module 22, a central processing unit (CPU) module 24, and a memory module 26, which may include, for example, a frame buffer for storing image data arrays. The computer system 20 may be linked to archive media devices, permanent or backup storage, or a network for storing image data and programs, and communicates with the MRI system control 32 via a high-speed signal link 34. Programs stored in the memory of computer system 20 may include artificial intelligence-based machine learning modules. Medical imaging workflows and the devices involved in these workflows can be configured, monitored, and updated throughout the operation of the medical imaging workflows and devices. One or more machine learning methods can be used to assist in configuring, monitoring, and updating medical imaging workflows and devices. For example, machine learning techniques (whether deep learning networks or other experiential / observational learning systems) can be used to locate objects in images, understand speech and convert speech to text, and improve the relevance of search engine results. Deep learning is a subset of machine learning that uses a set of algorithms to model high-level abstractions in data using depth maps with multiple processing layers (including linear and nonlinear transformations). While many machine learning systems first embed initial features and / or network weights and then modify them through learning and updating of the machine learning network, deep learning networks identify the features being analyzed by training themselves. When using a multi-layered architecture, machines employing deep learning techniques can process raw data better than machines using conventional machine learning techniques. Using different layers of evaluation or abstraction facilitates data examination of highly correlated values ​​or distinctive topics.

[0029] The MRI system control 32 can be separate from or integrated with the computer system 20. The computer system 20 and the MRI system control 32 together form the "MRI controller" 33.

[0030] In this exemplary embodiment, the computer system 20 includes a user interface that receives at least one input from a user. The user interface may include a keyboard 806 that enables the user to input relevant information. The user interface may also include, for example, a pointing device, a mouse, a stylus, a touch-sensitive panel (e.g., a touchpad and a touchscreen), a gyroscope, an accelerometer, a position detector, and / or an audio input interface (e.g., including a microphone).

[0031] Furthermore, in this exemplary embodiment, the computer system 20 includes a presentation interface that presents information to the user, such as input events and / or verification results. The presentation interface may also include a display adapter coupled to at least one display device. More specifically, in this exemplary embodiment, the display device may be a visual display device, such as a cathode ray tube (CRT), liquid crystal display (LCD), light-emitting diode (LED) display, and / or an "electronic ink" display. Alternatively, the presentation interface may include an audio output device (e.g., an audio adapter and / or a speaker) and / or a printer.

[0032] Computer system 20 also includes processor module 22 and memory module 26. Processor module 22 is coupled to user interface, presentation interface, and memory module 26 via a system bus. In an exemplary embodiment, processor module 26 communicates with the user, for example by prompting the user through the presentation interface and / or by receiving user input through the user interface. The term "processor" generally refers to any programmable system, including systems and microcontrollers, reduced instruction set computers (RISC), complex instruction set computers (CISC), application-specific integrated circuits (ASIC), programmable logic circuits (PLC), and any other circuitry or processor capable of performing the functions described herein. The examples above are merely exemplary and are therefore not intended to limit the definition and / or meaning of the term "processor" in any way.

[0033] In this exemplary embodiment, memory module 26 includes one or more devices that enable the storage and retrieval of information such as executable instructions and / or other data. Furthermore, memory module 26 includes one or more computer-readable media, such as, but not limited to, dynamic random access memory (DRAM), static random access memory (SRAM), solid-state disks, and / or hard disks. In this exemplary embodiment, memory module 26 stores, but is not limited to, application source code, application object code, configuration data, additional input events, application state, assertion statements, validation results, and / or any other type of data. In this exemplary embodiment, computer system 20 may also include a communication interface coupled to processor module 22 via a system bus. Furthermore, the communication interface is communicatively coupled to a data acquisition device.

[0034] In this exemplary embodiment, processor module 22 can be programmed by encoding operations using one or more executable instructions and by providing executable instructions in memory module 26. In this exemplary embodiment, processor 814 is programmed to select multiple measurements received from a data acquisition device.

[0035] In operation, computer system 20 executes computer-executable instructions embodied in one or more computer-executable components stored on one or more computer-readable media to implement aspects of the invention described and / or illustrated herein. Unless otherwise specified, the order of execution or implementation of operations in the embodiments of the invention shown and described herein is not required. That is, unless otherwise specified, these operations may be performed in any order, and embodiments of the invention may include more or fewer operations than those disclosed herein. For example, it is contemplated that a particular operation may be performed before, simultaneously with, or after another operation within the scope of various aspects of the invention.

[0036] In an exemplary embodiment, the MRI system control 32 includes modules connected together by a base plate 32a. These modules include a CPU module 36 and a pulse generator module 38. The CPU module 36 is connected to the operator console 12 via a data link 40. The MRI system control 32 receives commands from the operator via the data link 40 to indicate the scan sequence to be executed. The CPU module 36 operates the system components to execute the desired scan sequence and generates data indicating the timing, intensity, and shape of the generated RF pulses, as well as the timing and length of the data acquisition window. The CPU module 36 is connected to components operated by the MRI controller 33, including the pulse generator module 38 (controlling the gradient amplifier 42, further discussed below), the physiological acquisition controller (PAC) 44, and the scan chamber interface circuitry 46.

[0037] In one example, CPU module 36 receives subject data from physiological acquisition controller 44, which receives signals from sensors connected to the subject, such as ECG signals from electrodes attached to the patient. CPU module 36 receives signals from sensors associated with the patient and magnet system status via scan chamber interface circuitry 46. Scan chamber interface circuitry 46 also enables MRI controller 33 to command patient positioning system 48 to move the patient to the desired location for scanning.

[0038] The whole-body RF coil 56 is used to transmit waveforms toward the subject's anatomical structures. The whole-body RF coil 56 can be a body coil (such as...) Figure 1(As shown). The RF coil can also be a local coil 57, which can be positioned closer to the subject anatomy than the body coil. The RF coil 57 can be a surface coil. A surface coil 57 containing receiving channels can be used to receive signals from the subject anatomy. A typical surface coil 57 will have eight receiving channels; however, different numbers of channels are possible. It is known that using a combination of both the body coil 56 and the surface coil 57 can provide better image quality. Using the body coil 56 as both the transmitting and receiving coils can reduce the cost of the magnetic resonance (MR) system. However, when only the body coil 56 is used, the signal-to-noise ratio (SNR) of the MR system decreases.

[0039] Pulse generator module 38 operates gradient amplifier 42 to achieve the desired timing and shape of gradient pulses generated during scanning. The gradient waveform generated by pulse generator module 38 is applied to gradient amplifier system 42, which has Gx, Gy, and Gz amplifiers. Each gradient amplifier excites a corresponding physical gradient coil in gradient coil assembly 50 to generate a magnetic field gradient for spatial encoding of the acquired signal. Gradient coil assembly 50 forms part of magnet assembly 52, which also includes polarized magnet 54 (which, in operation, provides a longitudinal magnetic field B0 throughout the target space 55 surrounded by magnet assembly 52) and whole-body RF coil 56 (which, in operation, provides a transverse magnetic field B1, which is substantially perpendicular to B0 throughout the target space 55). Transceiver module 58 in MRI system control 32 generates pulses that can be amplified by RF amplifier 60, which is coupled to RF coil 56 via transmit / receive switch 62. The signal emitted by the stimulated nuclei in the subject's anatomy can be sensed by the same RF coil 56 and provided to the preamplifier 64 via the transmit / receive switch 62. The amplified MR signal is demodulated, filtered, and digitized in the receiver section of the transceiver 58. The transmit / receive switch 62 is controlled by a signal from the pulse generator module 38 to electrically connect the RF amplifier 60 to the coil 56 during transmit mode and to connect the preamplifier 64 to the coil 56 during receive mode. The transmit / receive switch 62 also allows a separate RF coil (e.g., a surface coil) to be used in either transmit or receive mode.

[0040] After the RF coil 56 picks up the MR signal generated by the target excitation, the transceiver module 58 digitizes these signals. The MR system control 32 then processes the digitized signals via Fourier transform to generate k-space data, which is then transmitted via the MRI system control 32 to the memory module 66 or other computer-readable medium. "Computer-readable medium" may include, for example, structures configured such that electrical, optical, or magnetic states can be perceptibly and reproducibly fixed by a conventional computer (e.g., text or images printed on paper or displayed on a screen, optical disc, or other optical storage medium; "flash memory," EEPROM, SDRAM, or other electrical storage media; floppy disks or other magnetic disks, magnetic tape, or other magnetic storage media).

[0041] The scan is complete when an array of raw k-space data is acquired in computer-readable medium 66. For each image to be reconstructed, the raw k-space data is rearranged into separate k-space data arrays, and each of these k-space data arrays is input to array processor 68, which operates to reconstruct an array of image data using reconstruction algorithms such as Fourier transform. The image data is transmitted to computer system 20 via data link 34 and stored in memory. In response to commands received from operator console 12, the image data may be archived in long-term storage or further processed by image processor module 22 and transmitted to operator console 12 and displayed on display 16.

[0042] In some examples, the MRI controller 33 includes an example image processing / image quality controller, which is implemented using at least one of the CPU 24 and / or other processors of the computer system 20A and / or other processors of the system control 32A and / or a separate computing device that communicates with the MRI controller 33.

[0043] In some examples, the MRI system 10 can be used to image various parts of the subject's anatomy. In one example, Figure 2A human spinal cord or spine is shown. MRI system 10 can be used to image the spinal cord 200 or spine of a subject. Although other aspects of this disclosure are explained with respect to imaging the spine 200 of a subject, it will be apparent to those skilled in the art that the disclosed systems and methods can be used to image and label other parts of the subject's anatomy. MRI system 10 can be a low-resolution scanner or a high-resolution scanner. A typical MR imaging workflow may include obtaining several primary high-resolution images of the anatomy using MRI system 10 for a primary assessment of the subject's anatomy and storing the images on the memory of computer system 20. Processor 22 can access the images stored on the computer memory and process these images to identify and label the various parts of the human anatomy within the images. The MR imaging workflow may further include obtaining a final high-resolution MR image based on the primary assessment of the anatomy. The primary MR images of the subject require manual identification and labeling of the subject's anatomy, and this process is inconvenient for several reasons. Some of the reasons for inconvenience may include the time required for manually labeling the subject's anatomy and due to the probability of operator error when labeling the anatomy. Another important reason for inconvenience is the use of image segmentation techniques to divide the image into several parts and label each part individually. Therefore, traditional MR imaging workflows are time-consuming, operator-centric, and based on segmentation techniques.

[0044] The human spine 200 in this example consists of interconnected vertebrae 201, 202, 203, 204, 205 that form the central portion of the body and extend from the cranium (not shown) to the sacrum 210. Although only a few vertebrae 201, 202, 203, 204, 205 have been shown in this current example, it is readily understood that many more vertebrae form the spinal cord. The spinal cord 200, or spine, can be divided into three broad sections: the cervical section, which connects to the cranium; the thoracic section, which forms the sternum; and the lumbar section, which connects to the sacrum. Imaging the spine 200 using MRI equipment 10 includes obtaining images of vertebrae 201, 202, 203, 204, 205 and labeling vertebrae 201, 202, 203, 204, 205 according to their position within the spinal images. Spinal labeling is an important task in planning and reporting spinal examination reports to diagnose the condition of the spine 200.

[0045] Existing methods for labelling the spine involve obtaining a high-resolution image of the spine (200), manually counting vertebrae (201, 202, 203, 204, 205) and inserting markers. Vertebrae (201, 202, 203, 204, 205) have similar appearances, and labelling technicians can mislabel them due to this similarity. Any errors in labeling vertebrae (201, 202, 203, 204, 205) will require recounting and mitigation efforts that will delay the labeling process. Another method for labelling vertebrae (201, 202, 203, 204, 205) involves obtaining a high-resolution two-dimensional (2D) or dedicated 3D image of the spine (200) and using machine learning algorithms to generate vertebral segmentation. After segmenting vertebrae (201, 202, 203, 204, 205), vertebral masks can be generated either manually by labeling the vertebrae or through automated landmark detection techniques. However, this technique requires accurate segmentation of the vertebrae and labeling of baseline ground truth data for training the machine learning module. Therefore, existing methods that segment vertebrae 201, 202, 203, 204, and 205 and use them as training data are not only time-consuming but also computationally burdensome.

[0046] According to one aspect of this disclosure, Figure 3 A method 300 for generating vertebral markers from 2D locator images is illustrated. 2D locator images, also known as 2D triplane locators, can be obtained using a low-resolution MRI scanner without requiring further scanning and segmentation of the vertebrae. Method 300 includes obtaining at least one locator 2D image of the spine of a subject 310 using a low-resolution MRI device 10. Method 300 further includes marking vertebrae 320 by providing location points on the image. Location points may include one or more points on the vertebrae indicating their location. Furthermore, the location of the markers may be indicated by shapes such as squares, rectangles, or circles; however, any other shape may be used to indicate the location of the markers. In one example, a point may be... Figure 2The location is indicated by the 12th thoracic vertebra (T12), the 4th lumbar vertebra (L4), or the 1st sacral vertebra (S1). Labeling of a 2D image can be performed manually or using a machine learning module stored in computer memory and can be used to further label future images for labeling. According to an aspect of this disclosure, computer processor 22 can be configured to perform one or more of the other steps of method 300. According to an aspect of this disclosure, method 300 includes extracting 330 labeled points from the labeled image and storing the point locations for future labeling. Extraction (330) can be performed using known machine learning techniques to label the points, such as two stacked U-Net architectures that together form a WNET architecture with size-weighted dicing and a per-mask shape encoder. The U-Net architecture is a convolutional neural network used for medical image segmentation. According to another aspect of this disclosure, extraction 330 can be performed using a regression network to process 2D locator images. For example, a machine learning module of a locator image quality module (LocalizerIQ) can be adapted to regress and identify the coordinates of labeled points in 2D or 3D space. Method 300 further includes labeling 340 future high-resolution images using point locations stored on the 2D locator images to be labeled in the future. Labeling 340 includes using a machine learning module configured to use the stored point locations and apply them to identify points on the high-resolution images.

[0047] According to one aspect of this disclosure, such as Figure 4 As shown, point determination 400 can be performed using a machine learning module configured to identify and label the locations of points 410, 420, and 430 on a low-resolution image. The machine learning (ML) module can be stored in computer memory, and the computer processor 22 can be configured to execute ML instructions. In one example, points 410, 420, and 430 correspond to points T12, L4, and S1. When a high-resolution 2D or 3D spinal image 400 is presented to the ML module, the ML module will locate these points 410, 420, and 430 on the high-resolution image and insert labels T12, L4, and S1 at the points corresponding to the locations of points T12, L4, and S1 in the low-resolution image. This labeling method provides an improved workflow for processing MR images, with a reduced amount of time required to identify and label subject anatomy. Furthermore, the method does not require obtaining high-resolution images to generate anatomical labels and can completely eliminate the segmentation requirement of vertebrae, reducing not only time but also imaging costs.

[0048] According to one aspect of this disclosure, Figure 5A method 500 for generating a mask overlay 520 from a 2D spine image 510 is illustrated. Method 500 includes obtaining at least one locator 2D image 510 of the spine of a subject 530 using a low-resolution MRI device. Method 500 further includes labeling the vertebrae 540 by providing location points 541, 542, 543 on the image. In one example, location points 541, 542, 543 may be T12, L4, and S1. However, a greater number of different points can be labeled. Method 500 also includes extracting a mask 520 including points 541, 542, 543 using a machine learning module. The mask 520 can be overlaid onto other images obtained from a high-resolution scanner to label corresponding points within such other images. This method 500 can substantially reduce the time required to label the spine and allows for labeling as the spine is segmented into regions as needed.

[0049] According to one aspect of this disclosure, Figure 6 This illustrates the use of an ML module for the lumbar spine to estimate the error of the marker points. Points L4 and S1 of the lumbar vertebrae are marked using the ML module against fifty images 610, resulting in a mask 620. The marking accuracy of the mask 620 is tested. In one example, fifty test images of the lumbar spine are tested for marking accuracy. The average error for L4 is approximately 4.5 mm, and for S1, the average error is approximately 4.1 mm. When the length and width of the vertebra are approximately fifteen to eighteen mm, the error indicates reliable determination of the vertebral marker center within the vertebra. In one aspect, the error can be reduced by using a larger number of markers on the spine to create what is known as dense marking. A larger number of closed markers will act as closer anchor points for adjacent markers. Therefore, any error in marking a vertebra can be compensated for by using other vertebral markers closer to a vertebra with a marking error, and self-correcting error codes can be developed to correct the error. This method facilitates rapid localization and marking of vertebrae using a standard triplane locator and can integrate with and improve existing workflows for MR imaging. Therefore, automation of spinal marking using 2D locator images can be achieved.

[0050] This written description uses examples to disclose the invention, including the best mode, and also enables those skilled in the art to practice the invention, including making and using any computing system or system and performing any included methods. The patent scope of the invention is defined by the claims and may include other examples that would occur to those skilled in the art. Such other examples are intended to fall within the scope of the claims if they have structural elements that are not indistinguishable from the literal language of the claims, or if they include equivalent structural elements that have minor differences from the literal language of the claims.

Claims

1. A method for automatic tagging, the method comprising: Use low-resolution medical imaging equipment to obtain at least one locator image of the subject's anatomy; Mark at least one anatomical point within the at least one locator image, wherein the at least one anatomical point within the at least one locator image is marked without image segmentation; A mask for the at least one locator image, including the at least one anatomical point marker, is extracted using a machine learning module, wherein extracting the mask further includes identifying the spatial coordinates of the at least one anatomical point; and Based on the spatial coordinates of the at least one anatomical point within the locator image, the at least one anatomical point on a high-resolution image of the subject's anatomy is marked using the mask.

2. The method of claim 1, wherein obtaining at least one locator image of the subject anatomy using the low-resolution medical imaging device comprises obtaining a two-dimensional 2D or three-dimensional 3D image of the subject anatomy.

3. The method of claim 1, wherein marking the at least one anatomical point within the locator image comprises manual or automatic marking using the machine learning module.

4. The method of claim 1, wherein marking at least one anatomical point within the locator image comprises identifying the at least one anatomical point within the locator image and inserting a shape indicating the location of the mark at the at least one anatomical point.

5. The method of claim 1, wherein extracting the mask of the locator image containing the at least one anatomical point marker using a machine learning module comprises using a localizerIQ module containing a regression network to locate the at least one anatomical point.

6. The method of claim 1, further comprising densely marking the at least one locator image of the subject's anatomy and using a self-correcting error code to automatically correct the position of the at least one anatomical point within the locator image.

7. The method of claim 1, wherein the subject anatomy is a human spinal cord or spine, and marking at least one anatomical point within the locator image includes marking at least one vertebra within the spine.

8. The method of claim 7, wherein the at least one anatomical point within the locator image comprises at least one anatomical point in the thoracic portion, lumbar portion, or sacral portion of the spine.

9. The method according to claim 1, wherein the medical imaging device includes magnetic resonance imaging (MRI) equipment, ultrasound equipment, computed tomography (CT) system, X-ray system, positron emission tomography (PET) system, and single-photon emission computed tomography (SPECT) system.

10. A method for automatic spine marking, the method comprising: At least one locator image of the subject's spine is obtained using a low-resolution medical imaging device, wherein the at least one anatomical point within the at least one locator image is marked without image segmentation. Mark at least one anatomical point within the at least one locator image of the subject's spine; The machine learning module is used to extract a mask of the at least one locator image of the subject's spine, including the at least one anatomical point marker, wherein extracting the mask also includes identifying the spatial coordinates of the at least one anatomical point; as well as Based on the spatial coordinates of the at least one anatomical point within the locator image of the subject's spine, the mask is used to mark at least one anatomical point on a high-resolution image of the subject's spine.

11. The method of claim 10, wherein the at least one anatomical point within the locator image comprises at least one anatomical point in the thoracic portion, lumbar portion, or sacral portion of the spine.

12. The method of claim 10, wherein obtaining the at least one locator image of the subject's spine using a low-resolution medical imaging device comprises obtaining a two-dimensional 2D or three-dimensional 3D image of the spine.

13. The method of claim 10, wherein marking at least one anatomical point within the locator image of the subject's spine comprises manually or automatically marking using the machine learning module.

14. The method of claim 13, wherein marking at least one anatomical point within the locator image of the subject's spine comprises inserting the spine marker into the thoracic, lumbar, or sacral portion of the spine.

15. The method of claim 10, further comprising densely marking the at least one locator image of the spine and using a self-correcting error code to automatically correct the position of the at least one anatomical point within the locator image.

16. A system for automatic labeling, the system comprising: A low-resolution medical imaging device, the low-resolution medical imaging device being configured to acquire at least one locator image of a subject's anatomy; and A computer processor configured to receive the at least one locator image from the low-resolution medical imaging device, wherein the computer processor is further configured to: Mark at least one anatomical point within the at least one locator image, wherein the at least one anatomical point within the at least one locator image is marked without image segmentation; A machine learning module is used to extract a mask of the at least one locator image containing the at least one anatomical point marker, wherein extracting the mask further includes identifying the spatial coordinates of the at least one anatomical point; and Based on the spatial coordinates of the at least one anatomical point within the locator image, the mask is used to mark at least one anatomical point on a high-resolution image of the subject's anatomy.

17. The system of claim 16, wherein the at least one locator image of the subject anatomy comprises a two-dimensional 2D or three-dimensional 3D image of the subject anatomy.

18. The system of claim 16, wherein marking at least one anatomical point within the locator image comprises manual or automatic marking using a machine learning module.

19. The system of claim 16, wherein the computer processor is further configured to densely label the at least one locator image of the subject's anatomy and use a self-correcting error code to automatically correct the position of the at least one anatomical point within the locator image.

20. The system of claim 16, wherein the medical imaging equipment includes magnetic resonance imaging (MRI) equipment, ultrasound equipment, computed tomography (CT) system, X-ray system, positron emission tomography (PET) system, and single-photon emission computed tomography (SPECT) system.

Images