System and method for monitoring the health of gradient coils

Abstract

The present invention provides a system (96) for monitoring the health status (110) of a gradient coil (86, 88, 90) disposed in a magnetic resonance imaging system (10). The system (96) includes one or more sensors (98, 100, 102) and a controller (106). The one or more sensors (98, 100, 102) operate to obtain one or more parameter readings (107) of the gradient coil (86, 88, 90), wherein the one or more parameter readings (107) include at least one of an acoustic measurement value and a back electromotive force measurement value. The controller (106) is in electronic communication with the one or more sensors (98, 100, 102) and operates to generate the health status (110) based on at least one of the acoustic measurement value and the back electromotive force measurement value.

Description

Technical Field

[0001] The present disclosure relates to systems and methods for monitoring the health of gradient coils disposed in a magnetic resonance imaging system.

[0002] Discussion in this area

[0003] Magnetic resonance imaging ("MRI") is a widely recognized, commercially available technology for obtaining digital visual images representing the internal structure of an object having a large population of atomic nuclei that are sensitive to nuclear magnetic resonance ("NMR"). Many MRI systems use superconducting magnets to scan a subject / patient by applying a strong main magnetic field on the subject's nuclei. The nuclei are excited by radio frequency ("RF") signals / pulses that are transmitted via RF coils at characteristic NMR (Larmor) frequencies. When the excited protons relax back to their normal state of lower energy, a map or image of these nuclear responses is generated and displayed based on their spatial locations by spatially perturbing the local magnetic field around the subject and analyzing the RF responses from the nuclei (hereinafter also referred to as "MR signals"). The images of the nuclear responses (hereinafter also referred to as "MRI images" and / or simply "images") provide a non-invasive view of the subject's internal structure.

[0004] Many MRI systems utilize large electromagnetic coils, commonly referred to as gradient coils, to generate a gradient magnetic field within a target space containing a subject by exciting / energizing the gradient coils via an electric current. However, long-term continuous / repeated excitation of the gradient coils may damage the gradient coils, which in turn may cause the gradient coils to fail, for example, generating a degraded gradient magnetic field and / or not generating a gradient magnetic field at all. Typically, the failure of the gradient coils results in useless data being obtained from the MRI procedure / scan. As will be appreciated, many MRI procedures are typically resource-intensive. Therefore, it is typically very expensive for patients and MRI system operators, such as hospitals, to perform an MRI procedure / scan using an undetected failed gradient coil.

[0005] For various reasons, it is often difficult and / or impossible to predict when a particular gradient coil will fail via manual inspection. For example, a gradient coil may be difficult to inspect manually because it is typically located / enclosed in a magnet assembly. Thus, manual inspection of the gradient coil typically requires the MRI system to be taken offline, i.e., out of service, which reduces the availability of the MRI system for the patient. In addition, since many conventional gradient coil diagnostic systems are limited in their ability to detect / identify symptoms indicating an impending failure, manual inspection of the gradient coil may not accurately predict the failure of the gradient coil. Although there are automatic methods for detecting failed gradient coils, many such methods are only effective after a gradient coil fails. In addition, such systems may be limited in their ability to detect / identify symptoms indicating an impending failure.

[0006] Therefore, there is a general need for an improved system and method for monitoring the health of gradient coils disposed in an MRI system. Summary of the invention

[0007] In one embodiment, the present disclosure provides a system for monitoring the health status of a gradient coil disposed in a magnetic resonance imaging system. The system includes a controller and one or more sensors. The one or more sensors are operable to obtain one or more parameter readings of the gradient coil, wherein the one or more parameter readings include at least one of an acoustic measurement value and a back electromotive force measurement value. The controller is in electronic communication with the one or more sensors and is operable to generate a health status based on at least one of the acoustic measurement value and the back electromotive force measurement value.

[0008] In another embodiment, the present disclosure provides a method for monitoring the health status of a gradient coil in a magnetic resonance imaging system. The method includes obtaining one or more parameter readings of the gradient coil via one or more sensors, wherein the one or more parameter readings include at least one of an acoustic measurement and a back electromotive force measurement. The method also includes generating a health status based on at least one of the acoustic measurement and the back electromotive force measurement using a controller in electronic communication with the one or more sensors.

[0009] In yet another embodiment, the present disclosure provides a method for training a neural network. The method includes feeding a training data set to the neural network. The training data set includes a plurality of pairs, each pairing including a parameter reading and a known health state of a gradient coil, wherein the parameter reading is at least one of an acoustic measurement and a back electromotive force measurement. The method also includes training the neural network in a supervised manner based on the training data set, so that for one or more of the pairs, the neural network generates a health state that substantially matches the known health state. The method also includes outputting one or more weights of the neural network after training the neural network.

[0010] In yet another embodiment, the present disclosure provides a system for monitoring the health status of a gradient coil. The system includes a sensor and a controller. The sensor is operable to obtain one or more parameter readings of the gradient coil. The controller is in electronic communication with the sensor and is operable to generate a health status based at least in part on the one or more parameter readings. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Various aspects of the present disclosure may be better understood by reading the following detailed description and referring to the accompanying drawings, in which:

[0012] Figure 1 is a block diagram of an MRI system including a system for monitoring the health of gradient coils according to an exemplary embodiment;

[0013] Figure 2 According to an exemplary embodiment Figure 1 A schematic cross-sectional view of a magnet assembly of an MRI system;

[0014] Figure 3 According to an exemplary embodiment Figure 2 a diagram of a gradient coil assembly of a magnet assembly;

[0015] Figure 4 According to an exemplary embodiment Figure 3 A diagram of an x-axis gradient coil of a gradient coil assembly;

[0016] Figure 5 According to an exemplary embodiment Figure 3 A diagram of a y-axis gradient coil of a gradient coil assembly;

[0017] Figure 6 According to an exemplary embodiment Figure 3 A diagram of a z-axis gradient coil of a gradient coil assembly;

[0018] Figure 7 According to an exemplary embodiment, a method for monitoring a device arranged in Figure 1A block diagram of a system for analyzing the health status of a gradient coil in an MRI system;

[0019] Figure 8 is a depiction according to an exemplary embodiment applied to Figure 3 a diagram of excitation currents on gradient coils of a gradient coil assembly;

[0020] Fig. 9 is a depiction of an exemplary embodiment according to which an Figure 8 A graph of the back EMF voltage induced in the gradient coil after the excitation current of ;

[0021] Fig.10 is a diagram according to an exemplary embodiment of a method for generating a plurality of excitation currents each having a different frequency within a period of time. Figure 3 A graph of the peak back EMF voltage induced in the gradient coils of the gradient coil assembly;

[0022] Fig.11 is a depiction according to an exemplary embodiment for a plurality of excitation currents each having a different frequency, Figure 3 A graph of a peak back EMF voltage of a gradient coil of a gradient coil assembly;

[0023] Fig.12 According to an exemplary embodiment Figure 7 A block diagram of the neural network of the system;

[0024] Fig.13 According to an exemplary embodiment Fig.12 A block diagram of a neuron in a neural network;

[0025] Fig.14 According to an exemplary embodiment, as shown in Figure 1 The MRI system's monitor Figure 7 A diagram of the health status of the system;

[0026] Fig.15 According to an exemplary embodiment, the use Figure 7 The system monitoring settings are Figure 1 A flow chart of a method for determining the health status of a gradient coil in an MRI system;

[0027] Fig.16 is a depiction training according to an exemplary embodiment Fig.12 A flowchart of a method of a neural network; and

[0028] Fig.17 is a depiction of an exemplary embodiment for training Fig.12 Block diagram of a neural network server, wherein the server is networked to one or more MRI systems.

[0029] The accompanying drawings illustrate certain aspects of the described systems and methods for monitoring the health of gradient coils. Together with the following description, the accompanying drawings illustrate and explain the structural principles, methods, and principles described herein. In the accompanying drawings, the sizes of components may be exaggerated or otherwise modified for clarity. Well-known structures, materials, or operations may not be shown or described in detail to avoid obscuring various aspects of the described components, systems, and methods. DETAILED DESCRIPTION

[0030] One or more specific embodiments of the present disclosure are described below to provide a thorough understanding. These described embodiments are merely examples of systems and methods for monitoring the health of gradient coils. In addition, as will be appreciated, embodiments of the present invention are not limited to neural networks, and therefore may include other forms of artificial intelligence. Those skilled in the art will appreciate that the specific details described in the embodiments may be modified in practice without departing from the essence of the present disclosure.

[0031] When introducing the elements of various embodiments of the present disclosure, the words "one", "a kind of" and "the" are intended to mean one or more elements in these elements. The terms "first", "second", etc. do not represent any order, amount or importance, but are used to distinguish one element from another element. The terms "comprise", "include" and "have" are intended to be inclusive, and mean that additional elements may also exist in addition to the listed elements. As used herein, the terms "connected to", "coupled to", etc., an object (e.g., material, element, structure, member, etc.) can be connected to or coupled to another object, regardless of whether the object is directly connected or coupled to another object, or whether there are one or more intervening objects between the object and another object. As used herein, "electrical coupling", "electrical connection" and "electrical communication" mean that the referenced element is directly or indirectly connected so that electric current can flow from one to another. The connection may include direct conductive connection (i.e., without intervening capacitance, inductance or active element), inductive connection, capacitive connection and / or any other suitable electrical connection. There may be intervening components.

[0032] Furthermore, it should be understood that references to "one embodiment" or "an embodiment" of the present disclosure are not intended to be interpreted as excluding the existence of additional embodiments that also incorporate the recited features. Furthermore, the following claim limitations are not written in means-plus-function form and are not intended to be so interpreted unless and until such claim limitations expressly use the phrase "means for..." followed by a recitation of functionality without other structure.

[0033] Additionally, it should be understood that embodiments of the present invention may be applicable to positron emission tomography ("PET") / MRI and / or any other system having components that are susceptible to failure and / or degradation due to use-induced stresses. For example, while the present invention is discussed herein as monitoring the health of gradient coils, it should be understood that the systems and methods disclosed herein are equally applicable to other components in an MRI system, such as body coils, superconducting magnets, gradient amplifiers, etc.

[0034] Referring generally to the accompanying drawings, the present disclosure will provide systems and methods for monitoring the health of gradient coils disposed in MRI systems. In some embodiments, the systems and methods disclosed herein generate the health of the gradient coils based on statistical deviations between one or more parameter readings of the gradient coils and historical standards of the same parameter readings of the gradient coils that have experienced little or almost no structural degradation. As used herein with respect to the gradient coils, the term "parameter reading" refers to a measurement of a physical and / or chemical property / measurement of the gradient coils. The parameter readings can be various measurements of the gradient coils, for example, acoustic waves, back electromotive force ("back EMF") measurements, and / or other measurements related to structural degradation of the gradient coils. As used herein, the terms "back electromotive force" and "back EMF" refer to the back electromotive force generated in the gradient coils after the applied excitation current is removed. As used herein with respect to the gradient coils, the term "structural degradation" refers to changes in the physical and / or chemical structure of the material forming the gradient coils.

[0035] In some embodiments, the controller may generate a health state based on a predetermined / known correlation / scale / model that maps / correlates one or more statistical differences / variances of gradient coil parameter readings from historical standards to a known level / amount of gradient coil structural degradation. In an embodiment, the correlation between the statistical deviation of the parameter readings and the structural degradation may be determined in part by passing parameter readings obtained from one or more gradient coils to a neural network, i.e., the neural network may be trained based on a historical data set of parameter readings collected from gradient coils of multiple MRI systems. By analyzing the data set of historical parameter readings, the neural network of some embodiments is able to provide an accurate indication of the health state of the gradient coil based on new parameter readings collected from the gradient coil. Thus, in some embodiments, the controller may generate a health state of the gradient coil by passing / feeding the parameter readings collected from the gradient coil to the neural network. Additionally, in some embodiments, the controller and / or the neural network is able to predict a time period in which the gradient coil is expected to fail.

[0036] See now Figure 1, which shows the main components of an MRI system 10 incorporating an exemplary embodiment of the present invention. Thus, the operation of the system 10 is controlled by an operator console 12, which includes a keyboard or other input device 14, a control panel 16, and a display screen 18. The console 12 can communicate with a separate computer system 22 via a link 20, which enables the operator to control the generation and display of images on the display screen 18. The computer system 22 may include multiple modules that communicate with each other via a backplane 24. In an embodiment, these modules include an image processor module 26, a CPU module 28, and a memory module 30 (which may include a frame buffer for storing an array of image data). The computer system 22 may communicate with a separate system control or control unit 32 via a high-speed serial link 34. The input device 14 may include a mouse, a joystick, a keyboard, a trackball, a touch-activated screen, a light stick, a voice control, or any similar or equivalent input device, and may be used for interactive geometry requirements. The computer system 22 and the MRI system control 32 together form an "MRI controller" 36.

[0037] In an embodiment, the MRI system control 32 includes a set of modules connected together by a backplane 38. These modules include a CPU module 40 and a pulse generator module 42, which is connected to the operator console 12 via a serial link 44. Through the link 44, the system control 32 receives commands from the operator indicating the scan sequence to be performed. The pulse generator module 42 operates the system components to perform the required scan sequence and generates data indicating the time, intensity and shape of the generated RF pulses and the time and length of the data acquisition window. The pulse generator module 42 is connected to a set of gradient amplifiers 46 to indicate the time and shape of the gradient pulses generated during the scan. The pulse generator module 42 can also receive patient data from a physiological acquisition controller 48, which receives signals from a plurality of different sensors connected to the patient, such as ECG signals from electrodes attached to the patient. And finally, the pulse generator module 42 is connected to a scan room interface circuit 50, which receives signals associated with the condition of the patient and the magnet system from various sensors. The patient positioning system 52 also receives commands to move the patient to the desired position for scanning through the scan room interface circuit 50.

[0038] The pulse generator module 42 operates the gradient amplifiers 46 to achieve the desired timing and shape of the gradient pulses generated during scanning. The gradient waveforms generated by the pulse generator module 42 are applied to the gradient amplifier system 46 having Gx, Gy and Gz amplifiers. Each gradient amplifier excites a corresponding physical gradient coil in a gradient coil assembly generally designated 54 to generate a magnetic field gradient for spatially encoding the acquired signals. The gradient coil assembly 54 forms part of a magnet assembly 56, which also includes a polarizing magnet 58 (which, in operation, provides a uniform longitudinal magnetic field B0 throughout a target space 60 surrounded by the magnet assembly 56) and an integral (transmit and receive) RF coil 62 (which, in operation, provides a transverse magnetic field B1 that is substantially perpendicular to B0 throughout the target space 60).

[0039] The resulting signals emitted by the excited nuclei in the patient's body can be sensed by the same RF coil 62 and coupled to the preamplifier 66 through the transmit / receive switch 64. The amplifier MR signals are demodulated, filtered, and digitized in the receiver portion of the transceiver 68. The transmit / receive switch 64 is controlled by a signal from the pulse generator module 42 to electrically connect the RF amplifier 70 to the RF coil 62 during the transmit mode and to connect the preamplifier 66 to the RF coil 62 during the receive mode. The transmit / receive switch 64 can also enable a separate RF coil (e.g., a surface coil) to be used in the transmit or receive mode.

[0040] The MR signals picked up by the RF coil 62 are digitized by the transceiver module 68 and transmitted to a memory module 72 in the system control 32. The scan is complete when an array of raw K-space data is acquired in the memory module 72. For each image to be reconstructed, the raw K-space data is rearranged into a separate K-space data array, and each of these separate K-space data arrays is input to an array processor 76, which operates to Fourier transform the data into an array of image data. The image data is transmitted to the computer system 22 via the serial link 34 and is stored in the memory 30. In response to commands received from the operator console 12, the image data may be archived in a long-term storage device, or may be further processed by the image processor 26, transmitted to the operator console 12, and presented on the display 18.

[0041] like Figure 2 , which shows a schematic side elevation view of a magnet assembly 56 according to an embodiment of the present invention. The magnet assembly 56 is cylindrical in shape with a center / central axis 78. The magnet assembly 56 includes a cryostat 80 and one or more radially aligned longitudinally spaced superconducting coils 82 that form the polarized magnet 58 ( Figure 1 ). The superconducting coils 82 are capable of carrying large currents and are designed to generate a B0 magnetic field within the patient / target space 60. As will be appreciated, the magnet assembly 56 may also include an end shield and vacuum vessel (not shown) surrounding the cryostat 80 to help isolate the cryostat 80 from heat generated by the rest of the MRI system 10 ( Figure 1 ). The magnet assembly 56 may further include other components, such as covers, supports, suspension members, end caps, brackets, etc. (not shown). Figure 1 and Figure 2 The illustrated embodiment of the magnet assembly 56 utilizes a cylindrical topology, but it should be understood that topologies other than cylindrical may also be used. For example, a flat geometry in an open MRI system may also utilize the embodiments of the present invention described below. Figure 2 As further shown, a patient / imaging subject 84 is inserted into the magnet assembly 56 .

[0042] Figures 3 to 6 Various gradient coils 86, 88, and 90 ( Figure 3 As will be appreciated, in embodiments, the gradient coil assembly 54 may include an x-gradient coil 86 operable to generate / apply a gradient magnetic field along / corresponding to an x-axis 92 (in Figure 4 ), y gradient coils 88 (in Figure 5 ) and a z gradient coil 90 (in Figure 6 ), the z-axis can be connected to the magnet assembly 56 ( Figure 1 and Figure 2 )'s central axis 78( Figure 2 As will be appreciated, in addition to applying magnetic gradients oriented along axes 91, 92, and 94, gradient coils 86, 88, and 90 may be used to apply magnetic gradients along / corresponding to any direction within the space defined by the three axes 91, 92, and 94.

[0043] Go to Figure 7 , which shows a method for monitoring gradient coils 86, 88 and / or 90 ( Figures 4 to 6 ) of the health status of the system 96. In an embodiment, it can be combined with the MRI system 10 ( Figure 1 and Figure 2 ) includes one or more sensors 98, 100, 102 (also shown in Figure 1 and Figures 4 to 6), a memory device 104 (optional), and a controller 106. The one or more sensors 98, 100, 102 are operable to obtain one or more parameter readings 107 of the gradient coils 86, 88, and / or 90, and the memory device 104 can store a neural network 108 ( Fig.12 ) and / or other types of models. Figure 1 and Figures 4 to 7 Three (3) sensors 98, 100, 102 are depicted in FIG. 1 , however it should be understood that the present invention is not limited to three (3) sensors and embodiments may include fewer and / or more than three (3) sensors.

[0044] The controller 106 is in electronic communication with the one or more sensors 98, 100, 102, the memory device 104, and is operable / configurable / adjustable to generate a health status 110 ( Fig.14 ), which health status provides an indication of structural degradation of the gradient coils 86, 88 and / or 90. As will be appreciated, in embodiments, the memory device 104 may be associated with the memory devices 30 and / or 72 ( Figure 1 ) may be separate from or include the memory device 30 and / or 72. Similarly, the controller 106 may be separate from the controller 36 ( Figure 1 ) separately from or including the controller 36.

[0045] As in Figures 4 to 6 As best shown in , the one or more sensors 98, 100, 102 may be disposed proximate, i.e., in proximity to, in contact with, and / or within the gradient coils 86, 88, and / or 90. In an embodiment, the parameter readings 107 obtained may include acoustic measurements and / or back EMF measurements.

[0046] For example, in certain embodiments, one or more of the sensors 98, 100, 102 may be microphones, such as condenser microphones or optical microphones, that collect acoustic measurements generated by the gradient coils 86, 88, and / or 90 that may have frequency, amplitude, or other sound-based metrics. As will be appreciated, the acoustic characteristics of the gradient coils 86, 88, and / or 90 vary as the level / amount of structural degradation of the coils 86, 88, and / or 90 varies. Thus, by mapping the acoustic characteristics sampled from the gradient coils 86, 88, and / or 90 to known levels of structural degradation, embodiments of the present invention create a model, such as a neural network, that can be used to discover structural degradation of other gradient coils based on their acoustic characteristics.

[0047] See now Figure 8 and Fig. 9As described above, in an embodiment, the one or more sensors 98, 100, and 102 may obtain back EMF measurements of the gradient coils 86, 88, and / or 90. Figure 8 and Fig. 9 As shown, in the embodiment, the gradient amplifier 46 ( Figure 1 ) can be applied to the gradient coils 86, 88 and / or 90 ( Figures 4 to 6 ) Apply and remove the excitation current 200( Figure 8 ), where the back EMF measurement has a voltage 202 (induced by the movement of the gradient coils 86, 88 and / or 90 within the B0 magnetic field) Fig. 9 ), also referred to herein as a "back EMF voltage". In an embodiment, the excitation current 200 can be an AC current having a back EMF voltage 202 generated by the Lorentz force due to deformation of the gradient coils 86, 88 and / or 90 (i.e., movement of the coils 86, 88 and / or 90 through the B0 magnetic field lines), wherein the coils 86, 88 and / or 90 eventually return to a state of equilibrium, i.e., without movement through the B0 magnetic field lines. As will be appreciated, the greater the structural degradation of the gradient coils 86, 88 and / or 90, the greater the displacement / movement of the coils 86, 88 and / or 90 through the B0 magnetic field lines, and therefore, the greater the back EMF voltage 202 induced in the coils 86, 88 and / or 90. For example, Figure 8 and Fig. 9 As shown, the gradient amplifier 46 can apply an excitation current 200 for a period of time, for example, t0 to t1, which in turn causes the back EMF voltage 202 to be induced in a decaying manner after the excitation current 200 is removed at t1, that is, the oscillations of the sinusoidal wave of the voltage 202 decay / contract over time as the coils 86, 88 and / or 90 return to a state of equilibrium.

[0048] like Fig.10 and Fig.11 As shown, the peak voltage ("Vpp") of the back EMF voltage 202 of the gradient coils 86, 88, and / or 90 may indicate any structural degradation experienced. Fig.10 A graph of the peak voltage of the gradient coils 86, 88, and / or 90 measured at several frequencies over the course of one or more days is shown, with the measurements for a particular day forming a curve. As will be appreciated, the shape of the curves (each curve formed by one or more peak voltage measurements at multiple frequencies on the same day) changes over time as the gradient coils 86, 88, and / or 90 experience structural degradation.

[0049] therefore, Fig.11 Curve 250 depicts a gradient coil that has experienced little or no structural degradation, while curve 252 is a gradient coil that has experienced structural degradation to the point of failure. Fig.11As shown, the back EMF voltage 252 of a gradient coil 86, 88, and / or 90 that has experienced a significant amount of structural degradation (i.e., a gradient coil that has failed or is about to fail) may exhibit statistical deviations at one or more excitation current frequencies from the back EMF voltage 250 of a similar gradient coil 86, 88, and / or 90 that has experienced little or no structural degradation. It should be appreciated that embodiments of the present invention may determine whether a particular gradient coil 86, 88, and / or 90 has failed based, at least in part, on detected statistical deviations from historical standards and / or baseline data (for similar gradient coils) in the back EMF voltage of the gradient coil 86, 88, and / or 90. For different gradient coils (e.g., different designs, different manufacturers, etc.), the range of excitation current frequencies and the range of back EMF voltages may vary and may be identified without undue experimentation.

[0050] Thus, in an embodiment, one or more sensors 98, 100, 102 may include a voltmeter that measures the back EMF voltage 202 in the gradient coils 86, 88, and / or 90. As will be appreciated, in an embodiment, the voltmeter 98, 100, 102 may be disposed in one or more of the x, y, or z gradient amplifiers 46, such as Figure 1 Thus, in some embodiments, the sensors 98 , 100 , 102 may be voltmeters that measure the back EMF voltage of the gradient coils 86 , 88 , and / or 90 from the perspective of the gradient amplifier 46 .

[0051] Although the previous paragraphs have referred to one or more parameter readings 107 ( Figure 7 and Fig.17 ) are discussed as acoustic and / or back EMF measurements, but it should be understood that in embodiments, the parameter readings 107 obtained may include: impedance measurements; inductance measurements; resistance measurements; strain measurements; temperature measurements; acceleration measurements, such as physical vibrations of gradient coils 86, 88, and / or 90; B0 drift measurements; junction box torque measurements, such as junction box 134 ( Figure 2 ) on the gradient coils 86, 88 and / or 90, which connect the cables that provide electrical power, such as the excitation current 200 ( Figure 8 ); and / or other measurements suitable for determining structural degradation of the gradient coils 86, 88, and / or 90. Thus, in such embodiments, the one or more sensors 98, 100, 102 may include resistance sensors, strain gauges, temperature probes, accelerometers, current meters, and / or other types of sensors suitable for measuring the aforementioned measurements.

[0052] Now go to Fig.12In some embodiments, the neural network 108 is used to monitor the health of the gradient coils. The neural network 108 may include one or more nodes / neurons 112, which in embodiments may be arranged in one or more layers 114, 116, 118, 120, 122, 124, 126. As used herein with respect to a neural network, the term "layer" refers to a collection of simulated neurons having inputs and / or outputs that are connected in a similar manner to other collections of simulated neurons. Thus, as Fig.12 As shown, neurons 112 may be connected to each other via one or more connections 128 so that data, such as parameter readings 107 (eg, obtained from one or more sensors 98, 100, 102) are transmitted to the user. Figure 7 and Fig.17 ) may propagate from the input layer 114 through one or more intermediate layers 116 , 118 , 120 , 122 , 124 to the output layer 126 .

[0053] like Fig.13 As shown, the connections 128 of a single neuron 112 may include one or more input connections 130 and one or more output connections 132. Each input connection 130 of a neuron 112 may be an output connection of a preceding neuron, and an output connection 132 of a neuron 112 may be an input connection of one or more subsequent neurons. Fig.13 Neuron 112 is depicted as having a single output connection 132, but it should be understood that a neuron may have multiple output connections that transmit / pass the same value. In an embodiment, neuron 112 may be a data construct, such as a structure, an instantiated class object, a matrix, etc., and input connections 130 may be received by neuron 112 as weighted numerical values, such as floating point or integer values. For example, Fig.13 As shown, the input connections X1, X2 and X3 can be weighted, summed, and sent / emitted / delivered as the output connection Y via weights W1, W2 and W3, respectively. As will be appreciated, the processing of a single neuron 112 can generally be represented by the following formula:

[0054]

[0055] Where n is the total number of input connections 130 to the neuron 112. In an embodiment, the value of Y may be based at least in part on whether the sum of WiXi exceeds a threshold. For example, if the sum of the weighted inputs does not exceed a desired threshold, Y may have a value of zero (0).

[0056] As will be further understood, the input layer 114 ( Fig.12 ) may be mapped to input connections 130 of neurons 112 from one or more sensors 98, 100, 102 ( Figure 1 and Figures 4 to 7) parameter readings obtained 107( Figure 7 and Fig.17 ), and the output layer 126 ( Fig.12 ) can be mapped to the output connection 132 of the neuron 112 in the health state 110 ( Fig.14 ). As used herein, "mapping" input connection 130 to parameter reading 107 refers to the manner in which parameter reading 107 affects / indicates the value of input connection 130. Similarly, as used herein, "mapping" output connection 132 to health state 110 refers to the manner in which the value of output connection 132 affects health state 110. Thus, in an embodiment, parameter reading 107, or a value derived therefrom, is passed / fed to input layer 114 of neural network 112 and propagates through layers 114, 116, 118, 120, 122, 124, 126 such that the mapped output connection of output layer 126 generates / corresponds to health state 110.

[0057] For example, Fig.14 As shown, the health status 110 may be a numerical score having a maximum value and a minimum value, such as a maximum value of one hundred (100) which may indicate that the gradient coils 86, 88, and / or 90 have no detectable structural degradation, and a minimum value of zero (0) which may indicate that the gradient coils 86, 88, and / or 90 have maximum detectable structural degradation. Thus, the health status 110 may be displayed on the display 18 (also shown at Figure 1 The health status 110 value is represented as a number (represented by box 136) and / or via a scale 138 having a top 140 and a bottom 142 corresponding to the maximum score value and the minimum score value, respectively, and a sliding bar 144 corresponding to the health status 110 value.

[0058] Go to Fig.15 , which shows a method of transmitting data via system 96 ( Figure 7 )Monitoring health status 110( Fig.14 ) method 146. The method 146 includes, at step 148, via one or more sensors 98, 100, 102 ( Figure 1 and Figures 4 to 7 ) to obtain gradient coils 86, 88 and / or 90 ( Figures 4 to 6 ) of one or more parameter readings 107 ( Figure 7 and Fig.17 ); and at step 150, based at least in part on one or more parameter readings 107, via neural network 108 ( Fig.12 ) or other models generate health status 110 ( Fig.14 ).

[0059] The method 146 may also include determining, at step 151, whether the generated health status 110 has exceeded a threshold 154 ( Fig.14 ), such as a lower threshold; and optionally, at step 152, when the health state 110 ( Fig.14 ) exceeds the lower threshold 154. For example, in an embodiment, the alarm can be sounded through the speaker 156 ( Fig.14 ) and / or visual prompts 158 ( Fig.14 ) The audio sound played. Although threshold 154 has been described above as a lower threshold, it should be understood that in an embodiment, threshold 154 can be a higher threshold.

[0060] In an embodiment, method 146 may also include: at step 160, when health status 110 ( Fig.14 ) exceeds the lower threshold of 154( Fig.14 ), via the controller 106 ( Figure 7 ) prevents excitation / energization / operation of gradient coils 86 , 88 and / or 90 .

[0061] Embodiments of the present invention may also provide for training neural network 108 ( Fig.12 ). For example, Fig.16 A method 300 of training a neural network 108 via supervised learning is shown. As used herein, the term "supervised learning" refers to the process of training the weights of a neural network 108 using a data set having inputs and known outputs (e.g., a labeled training data set). For example, in an embodiment, at step 302, a labeled training data set may be provided to the neural network 108, the labeled training data set including the known health states 110 ( Fig.14 ) of one or more pairs of parameter readings 107( Figure 7 and Fig.17 ), such as a set of recorded / acquired / sampled acoustic frequencies generated by the gradient coils 86, 88, and / or 90 paired with a set of corresponding known / expected health scores for each recorded / acquired / sampled acoustic frequency. In an embodiment, at step 304, a health state may be generated via the neural network 108 for each pairing in the training data set, and at step 306, compared to the known health state of the same pairing. If the generated health state matches or does not match the known health state of the pairing, respectively, then at step 308, the weights of the neural network 108 may be enhanced, or at step 310, the weights of the neural network may be adjusted. As indicated by decision block 312, steps 304, 306, 308, and 310 may be repeated multiple times for all pairs in the training data set until the accuracy of the neural network 108 reaches a sufficient level, such as, for example, a health state generated by the neural network 108 for a given parameter reading 107 in a pair of training data sets matches the known health state of the same pairing ninety-nine percent (99%) of the time.

[0062] Although the above paragraphs discuss training the neural network 108 via a supervised approach, as will be appreciated, other methods of training the neural network 108 may also be employed, such as unsupervised learning. As used herein, the term "unsupervised learning" refers to the process of training the weights of the neural network 108 without known outputs. For example, in such embodiments, the neural network 108 may be configured to train weights so as to maximize a cost function.

[0063] Go to Fig.17 , the neural network 108 may be trained on a data set obtained from a plurality of gradient coils generally represented by 162, 164, 166 ( Fig.12 ), the plurality of gradient coils may be disposed in different MRI systems generally represented by 168, 170, 172, located at different operating facilities such as hospitals and / or laboratories around the world connected via a network 174 such as the Internet to a server 176 located at a separate site 178 such as an MRI research center and / or laboratory. As will be appreciated, each of the MRI systems 168, 170, 172 may include system 96 ( Figure 7 ) so that the training data set may include parameter readings 107 acquired from the gradient coils 162, 164, 166 via sensors as described above ( Figure 7 and Fig.17 ), and a corresponding list of failure dates / times for the same gradient coils 162, 164, 166 and / or pulse sequences 180. The data set may then be sent to, stored in, or otherwise accessible by a server 176. The data set may then be used to train an instance of a neural network 108 on the server 176 to determine / discover acquired parameter readings 107 that indicate a gradient coil failure ( Figure 7 and Fig.17 ) within a relationship / symptom. As will be appreciated, the weights of the instantiated neural network trained on the server 176 may then be sent back to the system 96 for each of the MRI systems 168, 170, 172. In other words, historical data, such as parameter readings 107 ( Figure 7 and Fig.17 ) may be trended via the neural network 108 to detect / identify impending failures within the same or similar gradient coils 162 , 164 , 166 .

[0064] For example, parameter readings 107 may be obtained from the gradient coils 162, 164 of one or more of the MRI systems 168 and 170 at or near the time the MRI systems 168 and 170 come online (ie, begin operational use). Figure 7 and Fig.17 ), and stored as historical data in the server 176. The neural network 108 can then be trained to recognize a typical range of a given parameter reading 107 that corresponds to the beginning of the gradient coil useful life cycle (i.e., the time from first use of the gradient coil to failure). Over time, additional parameter readings 107 can be obtained from the gradient coils 162, 164 in the MRI systems 168 and 170 and stored in the server 176 so that the neural network 108 can be trained to recognize how the parameter readings 107 change throughout the useful life cycle of the gradient coil, up to and / or including failure of the gradient coil.

[0065] As will be appreciated, with newer / younger MRI systems such as MRI system 172, the neural network 108 has had the opportunity to take parameter readings 107 (107) over the full useful life of one or more gradient coils 162, 164 of the older MRI systems 168 and 170. Figure 7 and 17 ) and then come online, the health status of the gradient coils 166 of the new / younger MRI system 172 may be repeatedly generated at various points in its own life cycle via an instance of the neural network 108 operating locally on the new system 172, using the training results / weights obtained / learned from the life cycle of the older gradient coils 162 and 164. For example, in such embodiments, the health status 110 may be a score ranging from zero (0) to ten (10), where the neural network 108 is a parameter reading 107 ( Figure 7 and Fig.17 ) generates / assigns ten (10) to a gradient coil 166 that meets historical standards for a particular point in the useful life cycle of the gradient coil 166, as learned / learned via the neural network 108 through training using historical data sets / parameter readings 107 acquired from gradient coils 162 and 164 of older MRI systems 168 and 170.

[0066] As will be appreciated, in such embodiments, the neural network 108 may generate progressively lower scores corresponding to the amount by which the parameter readings 107 of the gradient coils 166 in the newer / younger MRI system 172 deviate from historical norms, with zero (0) being the maximum amount of deviation that can be detected. In other words, in such embodiments, the health status / score 110 ( Fig.10) of ten (10) indicates that the gradient coil 166 has aged, i.e., has caused structural degradation, as expected based on historical standards, wherein a health status / score of zero (0) indicates that the gradient coil 166 has aged faster than expected based on historical standards. In such embodiments, the lower threshold 154 ( Fig.14 ) may be five (5), which lower threshold may indicate that the gradient coil should be replaced and / or corrective action should be taken to mitigate / reduce the risk of failure, such as reducing the temperature acceleration of the gradient coil during future operating cycles and / or using software interlocks to prevent excitation frequencies that may increase structural degradation.

[0067] Additionally, in an embodiment, the neural network 108 may provide correlations between gradient coil acceleration (ie, physical vibrations and / or acoustics) and gradient coil failure; and similarly, correlations between inductance-resistance ("LR") and back EMF and gradient coil failure.

[0068] As will be appreciated, while the above-described training scenario involves parameter readings 107 ( Figure 7 and Fig.17 ), however, it should be understood that in an embodiment, the data set used to train the weights of the neural network 108 can be obtained from a single gradient coil 86, 88 and / or 90 of a single MRI system 168.

[0069] See again Fig.15 In an embodiment, the method 146 may further include, at step 182, predicting the health state 110 ( Fig.14 ) will exceed the lower threshold of 154 ( Fig.14 ) time period 184( Fig.14 ). For example, the neural network 108 may be trained as described above so that the neural network 108 is able to generate a parameter based on the acquired parameter readings 107 ( Figure 7 and Fig.17 ) accurately predicts a future time / date when the gradient coils 86, 88, and / or 90 will have sustained structural degradation sufficient to warrant repair and / or replacement of the gradient coils 86, 88, and / or 90. The time period 184 may be displayed on the monitor 18 in a variety of formats (e.g., Fig.14 ), the various formats such as weeks, years, days, hours, minutes, seconds, standard date formats and / or any other format capable of conveying the amount of time remaining before the health state 110 exceeds the lower threshold 154.

[0070] Finally, it should also be understood that the system 10 and / or 96 may include the necessary electronics, software, memory, storage, databases, firmware, logic / state machines, microprocessors, communication links, displays or other visual or audio user interfaces, printing devices, and any other input / output interfaces for performing the functions described herein and / or achieving the results described herein. For example, as previously described, the system 10 and / or 96 may include at least one processor and system memory / data storage structures, which may include random access memory (RAM) and read-only memory (ROM). The at least one processor of the system 10 and / or 96 may include one or more conventional microprocessors and one or more auxiliary coprocessors, such as math coprocessors, etc. The data storage structures discussed herein may include appropriate combinations of magnetic, optical, and / or semiconductor memory, and may include, for example, RAM, ROM, flash drives, optical disks such as compact disks, and / or hard disks or drives.

[0071] Additionally, a software application that adapts the controller to perform the methods disclosed herein may be read into the main memory of at least one processor from a computer-readable medium, such as a medium that provides or participates in providing instructions to at least one processor of the system 10 and / or 96 (or any other processor of the devices described herein) for execution. Such media may take many forms, including but not limited to non-volatile media and volatile media.

[0072] Although in an embodiment, execution of a sequence of instructions in a software application causes at least one processor to perform the methods / processes described herein, hard-wired circuitry may be used in place of or in combination with software instructions to implement the methods / processes of the present invention. Accordingly, embodiments of the present invention are not limited to any specific combination of hardware and / or software.

[0073] It should be further understood that the above description is intended to be illustrative rather than restrictive. For example, the above embodiments (and / or aspects thereof) can be used in combination with each other. In addition, without departing from the scope of the present invention, many modifications can be made to adapt specific situations or materials to the teachings of the present invention.

[0074] Therefore, by providing automatic monitoring of the health of the gradient coils, some embodiments of the present invention can reduce the risk of gradient coil failure during an MRI scan without the need to manually check the gradient coils. As will be appreciated, such embodiments can improve the patient throughput of an MRI system by avoiding the need to rescan a patient due to a gradient coil failure, and reducing and / or avoiding downtime of the MRI system caused during manual inspections of the gradient coils.

[0075] In addition, by predicting the time period when the gradient coil is expected to fail, some embodiments of the present invention can provide higher patient throughput than conventional MRI systems by facilitating improved coordination between scheduling patient scans and MRI system downtime due to gradient coil maintenance / replacement. In other words, rather than reactively detecting a failed gradient coil, some embodiments of the present invention provide proactive maintenance of the gradient coil. In such embodiments, proactive maintenance of the gradient coil can improve patient safety / comfort by reducing the risk that the patient will be exposed to excessive gradient coil noise / vibration.

[0076] Furthermore, by using the MRI system's gradient amplifiers to obtain back EMF parameter readings, some embodiments of the present invention provide a system for monitoring the health of the gradient coils that utilizes existing sensors / devices currently present in many MRI systems. Thus, such embodiments of the present invention provide improved monitoring of the health of the gradient coils without incurring the substantial expense typically associated with developing and / or installing new sensors in existing MRI systems already in use at various locations.

[0077] Still further, by storing and analyzing historical data sets / parameter readings of gradient coils in multiple MRI systems in a server, some embodiments of the present invention may provide an improved understanding of future gradient coil failures. For example, in such embodiments, analysis of historical data of the acceleration history of failed gradient coils by a neural network may help identify the root cause of the failure more quickly.

[0078] In addition to any modifications previously indicated, numerous other variations and alternative arrangements may be devised by those skilled in the art without departing from the spirit and scope of the invention, and the appended claims are intended to cover such modifications and arrangements. Therefore, although the present invention has been particularly and in detail described above in connection with what are presently considered to be the most practical and preferred aspects, it will be apparent to those skilled in the art that many modifications, including but not limited to form, function, mode of operation and / or use, may be made without departing from the principles and concepts set forth herein.

[0079] Finally, the examples and embodiments used herein are intended to be illustrative only and should not be construed as limiting in any way.

Claims

1. A system for monitoring the health status of a gradient coil disposed in a magnetic resonance imaging system, the system comprising: one or more sensors operative to obtain parameter readings of the gradient coil, wherein the parameter readings include at least one of acoustic measurements and back EMF measurements; and a controller in electronic communication with the one or more sensors and operative to: feeding a training data set to a neural network, wherein the training data set includes a plurality of pairs, each pair including the parameter reading and a known health state of the gradient coil; The neural network is trained in a supervised manner based on the training data set such that, for one or more of the pairings, the neural network generates a health state that substantially matches the known health state. 2 . The system of claim 1 , wherein the controller is further operative to trigger an alarm when the health status exceeds a threshold. 3 . The system of claim 1 , wherein the controller further operates to prevent firing of the gradient coil when the health status exceeds a threshold. 4 . The system of claim 1 , wherein the one or more sensors include a microphone disposed proximate to the gradient coil and operated to obtain the acoustic measurements.

5. The system of claim 1, wherein the one or more sensors include a voltmeter operative to obtain the back EMF measurement.

6. The system of claim 5, wherein the voltmeter is disposed within a gradient amplifier driving the gradient coil.

7. The system of claim 6, wherein the controller operates to obtain the back EMF measurement via the voltmeter after the gradient amplifier removes the excitation current from the gradient coil.

8. The system of claim 1, wherein the parameter readings further include at least one of impedance measurements, inductance measurements, resistance measurements, strain measurements, temperature measurements, acceleration measurements, B0 drift measurements, and junction box torque measurements.

9. The system of claim 1, wherein the controller further operates to predict a time period during which the health state will exceed a threshold.

10. The system of claim 1, wherein the neural network is trained on a data set corresponding to one or more pulse sequences of one or more magnetic resonance imaging systems.

11. A method for monitoring the health of a gradient coil in a magnetic resonance imaging system, the method comprising: obtaining parameter readings of the gradient coil via one or more sensors, wherein the parameter readings include at least one of acoustic measurements and back electromotive force measurements; feeding a training data set to a neural network, wherein the training data set includes a plurality of pairs, each pair including the parameter reading and a known health state of the gradient coil; training the neural network in a supervised manner based on the training data set such that, for one or more of the pairings, the neural network generates a health state that substantially matches the known health state; as well as One or more weights of the neural network are output after the neural network is trained.

12. The method according to claim 11, further comprising: When the health status exceeds a threshold, an alarm is triggered via the controller.

13. The method according to claim 11, further comprising: When the health status exceeds a threshold, excitation of the gradient coil is prevented via a controller.

Images