US20080100292A1 - System and method for fast mr coil sensitivity mapping - Google Patents
System and method for fast mr coil sensitivity mapping Download PDFInfo
- Publication number
- US20080100292A1 US20080100292A1 US11/554,516 US55451606A US2008100292A1 US 20080100292 A1 US20080100292 A1 US 20080100292A1 US 55451606 A US55451606 A US 55451606A US 2008100292 A1 US2008100292 A1 US 2008100292A1
- Authority
- US
- United States
- Prior art keywords
- coil
- data
- neural network
- sensitivity
- mapping
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R33/00—Arrangements or instruments for measuring magnetic variables
- G01R33/02—Measuring direction or magnitude of magnetic fields or magnetic flux
- G01R33/10—Plotting field distribution ; Measuring field distribution
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R33/00—Arrangements or instruments for measuring magnetic variables
- G01R33/20—Arrangements or instruments for measuring magnetic variables involving magnetic resonance
- G01R33/24—Arrangements or instruments for measuring magnetic variables involving magnetic resonance for measuring direction or magnitude of magnetic fields or magnetic flux
- G01R33/246—Spatial mapping of the RF magnetic field B1
Definitions
- the present invention relates generally to magnetic resonance imaging, and more specifically, to a system and method for mapping the sensitivity of radio frequency (RF) coils of a magnetic resonance (MR) system.
- RF radio frequency
- MR magnetic resonance
- the invention finds particular utility in parallel imaging applications by providing quickly determined coil sensitivity maps, requiring fewer flip angles/scans to determine.
- polarizing field B 0 When a substance such as human tissue is subjected to a uniform magnetic field (polarizing field B 0 ), the individual magnetic moments of the spins in the tissue attempt to align with this polarizing field, but precess about it in random order at their characteristic Larmor frequency. If the substance, or tissue, is subjected to a magnetic field (excitation field B 1 ) which is in the x-y plane and which is near the Larmor frequency, the net aligned moment, or “longitudinal magnetization”, M z , may be rotated, or “tipped”, into the x-y plane to produce a net transverse magnetic moment M t . A signal is emitted by the excited spins after the excitation signal B 1 is terminated and this signal may be received and processed to form an image.
- excitation field B 1 which is in the x-y plane and which is near the Larmor frequency
- magnetic field gradients (G x , G y , and G z ) are employed.
- the region to be imaged is scanned by a sequence of measurement cycles in which these gradients vary according to the particular localization method being used.
- the resulting set of received NMR signals are digitized and processed to reconstruct the image using one of many well known reconstruction techniques.
- Methods for mapping the sensitivity of RF coils typically require that multiple data acquisitions take place for each coil of an RF coil assembly, where each data acquisition is taken at a different flip angle. This typically results in the use of 2-8 different flip angles or transmit powers for each coil of an N-channel coil array. Therefore, a total of 2N-8N different acquisitions are often used to generate one sensitivity mapping.
- the signals acquired in these acquisitions are proportional to the sine of the flip angle. Therefore, a pixel by pixel fit of the intensity of the signal versus the flip angle (from the known RF pulse height or width) can produce a map of the coil sensitivity.
- This method of coil sensitivity mapping is well known in the art. However, as the number of calibration or sensitivity-measuring scans required to produce this pixel by pixel fit is increased, overall scan time and patient throughput decrease.
- coil sensitivity mapping is particularly advantageous is parallel imaging.
- a homogeneous coil such as a whole-body RF coil
- an array of surface RF coils is used to receive the MR signals, in order to increase the step size between phase-encoding lines, or equivalently to reduce the size of the field of view and the amount of data collected.
- Scan time reduction is achieved by under-sampling k-space and recording images simultaneously from the multiple imaging or receive coils. Under-sampling generally reduces the data acquisition time by increasing the distance of sampling positions in k-space.
- Parallel imaging techniques not only expedite data acquisition, but also reduce aliasing or wrapping that occurs in the phase-encoding direction when an imaging object extends outside a field-of-view (FOV).
- parallel imaging techniques remove or reduce the aliasing by using surface coil B 1 fields (sensitivities), to define or determine an un-aliased spin distribution.
- Information regarding the surface coil B 1 fields or sensitivities is typically acquired with an external calibration or a self-calibration technique.
- the coil sensitivity data is used to weight the imaging data such that coil sensitivity is reflected in the reconstructed image, and, as a result, the coil sensitivity data reduces aliasing in the reconstructed image that can occur as a result of under-sampling.
- an array of RF coils can also be used to both transmit and receive the MR signals (as opposed to using a homogenous RF coil for transmitting and many surface coils for receiving the signal).
- Such an approach has been shown to be useful for RF shimming, pulse designs for small FOV imaging, or reducing specific absorption rate for certain other MR applications.
- methods of pulse design typically rely on knowledge of the particular excitation and reception pattern of each of the small RF coils employed.
- the present invention provides for rapid coil sensitivity mapping for MR systems using computer-developed normalization or weighting functions.
- Neural networks or similar computer intelligence are exemplary techniques employed to generate the weighting functions, further allowing computers to construct coil sensitivity profiles for each coil of an RF coil assembly with minimal input data.
- an MRI system includes a plurality of gradient coils positioned about a bore of a magnet to impress a polarizing magnetic field, an RF coil assembly, a system control, and an image processing unit.
- the RF coil assembly is connected to emit RF pulse sequences and is arranged to receive MR data from a subject within the bore.
- the system control is configured to develop a coil sensitivity mapping function from sample sets of MR data and generate a given coil sensitivity map.
- the image processing unit is configured to apply the given coil sensitivity map in an MR imaging procedure.
- the invention is embodied in a method for sensitivity mapping of RF coils.
- the method includes training a neural network to determine coil sensitivity profiles, acquiring MR mapping data at a given transmit power and with a given RF pulse characteristic for each of a plurality of coils of an MR system, and inputting the MR mapping data to the neural network to obtain a coil sensitivity profile for each of the plurality of coils.
- a coil sensitivity map generator includes a processing unit programmed to operate a neural network to produce B1 maps, a training data interface connected to receive sample MR data and input the sample MR data to train the neural network, and a data acquisition unit configured to acquire MR mapping signals from each coil of an RF coil assembly and transmit the signals to the neural network.
- FIG. 1 is a schematic block diagram of an MR imaging system incorporating the present invention.
- FIG. 2 is a flowchart representing an exemplary sensitivity mapping technique.
- FIG. 3 is a flowchart representing a training technique in accordance with the present invention.
- FIG. 4 is a flowchart representing a sensitivity mapping technique, used after the system is trained, in accordance with the present invention.
- FIG. 1 the major components of an example magnetic resonance imaging (MRI) system incorporating the present invention are shown.
- the operation of the system may be controlled from an operator console 12 which includes a keyboard or other input device 13 , a control panel 14 , and a display screen 16 .
- the input device 13 can include a mouse, keyboard, track ball, touch activated screen, light wand, or any similar or equivalent input device, and may be used for interactive geometry prescription.
- the console 12 communicates through a link 18 with a separate computer system 20 that enables an operator to control the production and display of images on the display screen 16 .
- the computer system 20 includes a number of modules which communicate with each other through a backplane 20 a .
- the computer system 20 may also be connected to permanent or back-up memory storage, a network, or may communicate with a separate system control 32 through a high speed serial link 34 .
- the system control 32 includes a set of modules connected together by a backplane 32 a and connected to the operator console 12 through a serial link 40 . It is through link 40 that the system control 32 receives commands from the operator to indicate the scan sequence that is to be performed.
- the pulse sequence transmit module 46 commands the scanner components to carry out the desired scan sequence, by sending instructions, commands, and/or requests describing the timing, strength and shape of the RF pulses and pulse sequences to be produced, to correspond to the timing and length of the data acquisition window.
- the system control 32 also connects to a set of gradient amplifiers 42 , to indicate the timing and shape of the gradient pulses that are produced during the scan.
- the system control 32 may also receive patient data from a scan room interface 44 , which may relate data from a user or from a number of different sensors connected to the patient, such as ECG signals from electrodes attached to the patient.
- the gradient waveform instructions produced by system control 32 are sent to the gradient amplifier system 42 having Gx, Gy, and Gz amplifiers.
- Amplifiers 42 may be external of scanner 48 , or may be integrated therein.
- Each gradient amplifier excites a corresponding physical gradient coil in a gradient coil assembly generally designated 50 to produce the magnetic field gradients used for spatially encoding acquired signals.
- the gradient coil assembly 50 forms part of a magnet assembly 52 which includes a polarizing magnet 54 and an RF coil assembly 56 , 58 .
- RF coil assembly may include a whole-body RF transmit coil 56 , surface or parallel imaging coils 58 , or both.
- the coils 56 , 58 of the RF coil assembly may be configured for both transmitting and receiving, or for transmit-only or receive-only.
- a pulse generator (not shown) integrated into the scanner equipment 48 produces RF pulses in accordance with the instructions of the pulse sequence transmit module 46 which are amplified and coupled to the RF coil 56 for transmission.
- RF transmit coil 56 may be replaced or augmented with surface and/or parallel transmit coils, such as coil 58 .
- the resulting signals emitted by the excited nuclei in the patient may be sensed by separate receive coils, such as parallel coils or surface coils 58 , and are then sent over a data link 60 .
- the MR signals are demodulated, filtered, and digitized in the data processing section 62 of the system control 32 .
- a scan is complete when an array of raw k-space data has been acquired in the memory module 66 .
- This raw k-space data is rearranged into separate k-space data arrays for each image to be reconstructed, and each of these is input to an array processor 64 which operates to Fourier transform the data into an array of image data.
- This image data is conveyed through the serial link 34 to the computer system 20 where it is stored in memory 26 .
- this image data may be archived in long term storage or may be further processed by the image processor 22 and conveyed to the operator console 12 and presented on the display 16 .
- FIG. 2 One particular way to determine coil sensitivity maps is illustrated in the flowchart of FIG. 2 . It should be noted, however, that the techniques described herein and depicted in FIG. 2 are merely exemplary, and that other known coil sensitivity mapping techniques may equivalently be substituted for or combined with the following examples.
- the technique of FIG. 2 begins with MR signals being acquired 80 for one coil using an RF pulse at a first flip angle ⁇ 1 . Using the same coil, an additional set of MR data is acquired 82 using an RF pulse at the same transit gain, but with a second flip angle ⁇ 2 which is 2* ⁇ 1 . The flip angle may be doubled by altering a pulse width or amplitude of the RF excitation pulse.
- the second applied flip angle ⁇ 2 may be half of the first flip angle, or some other proportion which lends to a determination of actual flip angle, as described below.
- This process of MR data acquisition is repeated for each of the N coils of a multi-channel array, such as a surface coil array.
- each subsequent coil is individually controlled to acquire a first set of MR data 84 at the same transmit gain with flip angle ⁇ 1 , and a second set of MR data is acquired 86 at the same transmit gain but with a flip angle 2* ⁇ 1 . From the two sets of MR data acquired for all coils of the RF coil assembly, two images may be reconstructed 88 .
- a pixel by pixel fit of image intensity versus transmit power can generate a spatial map of flip angles produced by each coil at each location in space.
- the number of flip angles i used to acquire sets of MR data with each coil may depend upon the degree of B1 inhomogeneity or the desired precision in coil sensitivity determination.
- a set of i images can be reconstructed for each coil, corresponding to the i flip angles applied, ⁇ 1 , ⁇ 2 , . . . ⁇ i .
- a table of image intensities will exist, I 1 , I 2 , . . .
- the present invention utilizes neural networks, or other forms of computer intelligence, to ultimately produce coil sensitivity maps faster than the techniques described above with respect to FIG. 2 .
- Neural networks operate by managing the coefficients of a weighting function that converts a set of inputs into a set of outputs.
- Neural networks “learn” (or are trained) by adjusting the values of the weighting function in accordance with complete data sets of sample inputs and target outputs.
- the weighting function is often represented as a hidden logic level between the inputs and outputs of the system. Usually, each node of a hidden level reads all input values, multiplies by various weights or coefficients, and produces one or more output values.
- the initial stage during which a neural network tunes its weighting function to consistently fit sets of sample inputs and target outputs is commonly referred to as the training stage. Once trained, neural networks can be used to rapidly produce outputs based on a limited subset of real input data.
- One method for updating or adjusting the weights of the hidden level, or weighting function is known as back-propagation.
- An arbitrary set of weights is given to each node of the hidden level, and a weight adjustment increment is set.
- the arbitrary weighting function produces a set of outputs. These outputs are compared to a set of target outputs to determine how the weights must be adjusted or tuned to produce a better fit of the target outputs. After multiple sets of sample inputs and target outputs have been fed to the neural network, the weights will eventually become tuned to consistently produce an accurate fit or prediction of the target outputs.
- o(x) is a vector of one or more outputs of the given hidden node, determined as weight vector w multiplied by input vector x.
- each given hidden node will determine the error E (as a function of the weight vector w) between its computed output o d and the target output t d for each training sample set of inputs and outputs d of a dataset of training examples D by:
- An exemplary back-propagation weight-adjustment function, based on the determined error E between computed and target outputs can be described as:
- ⁇ w ji is the change in the weighting coefficient associated with the ith input to hidden unit j
- ⁇ is the chosen weight adjustment increment (such as 0.001, for example)
- t j is the target output for unit j
- o j is the computed output of unit j
- x ji is the ith training input value to hidden unit j.
- Such a function is generally described as being a gradient descent-based algorithm (i.e.
- Technique 98 begins with the acquisition of MR mapping data for each coil of an RF coil assembly 100 .
- MR mapping data is acquired multiple times for each coil, at varying flip angles. This may be performed in a manner similar to that shown in FIG. 2 .
- MR mapping data may be acquired with all coils simultaneously 102 at a different flip angle from those used in the initial acquisition. In a preferred embodiment, this flip angle may be twice or half the first flip angle used for each coil.
- coil sensitivity profiles are mapped for each coil 104 , as known in the art.
- the MR mapping data acquired from each coil at the first flip angle and the MR mapping data acquired from all coils simultaneously are then fed to a neural network as a set of sample input values (x ji ) 106 .
- the calculated sensitivity maps are fed to the network as the corresponding set of target outputs (t j ) 108 .
- the neural network is instructed to tune its weight function for the hidden level (or may do so automatically) 110 in accordance with the desired weight adjustment algorithm and associated features. If the network determines 112 that the weight function has been optimized 114 , the training process ends 116 . Otherwise, if the weight function has not yet been optimized 118 , then the training process repeats 120 . The number of repetitions will typically be proportionate to the dimensionality of the input matrix and target output matrix.
- MR mapping data (e.g. low resolution data) is acquired 124 from each coil of an RF coil assembly with an RF pulse at a first flip angle, ⁇ 1 , and having a given transmit gain and a given RF wave characteristic, such as a particular amplitude or pulse width.
- MR mapping data may then be acquired 126 from all coils simultaneously at a second flip angle, ⁇ 2 .
- the second flip angle is equal to double or half the first flip angle.
- any flip angle not equal to the first flip angle may be used for the second acquisition 126 .
- the second acquisition 126 may be optional.
- the MR mapping data acquired from the first set of acquisitions 124 and the second acquisition 126 are then fed to the trained neural network as input values 128 .
- the neural network is caused to produce an output via its tuned weight function, which is received by a system control or other processing unit as a coil sensitivity map or set of sensitivity profiles 130 .
- the sensitivity map may be used for image data normalization or tailoring specific coil currents. This process may take place in a significantly shorter time due to the reduced number of mapping data acquisitions and to the rapidity of neural network calculations.
- An imaging process may take place in which higher resolution MR signals are acquired in a conventional manner 132 .
- the sensitivity map or normalization computed by the neural network may be applied to the MR signals 134 to compensate for the non-uniformities of the B1 field.
- a parallel imaging reconstruction may be performed based on the sensitivity map, or a pulse sequence (e.g. tailored currents for each coil to produce specific excitation pulses) may be designed for specific or narrow field of view imaging or other uses.
- a data dimensionality reduction procedure such as a principal component analysis
- a principal component analysis is a technique for reducing the expression of given datasets (such as images or sets of MR signals) to their variations from a number of basis sets.
- given datasets such as images or sets of MR signals
- a number of datasets can be analyzed to factor out commonalities and major variations, which are expressed as a basis set.
- each given dataset is reduced to a vector or matrix of its projection onto the basis set. The result is a significantly reduced way to express datasets which have similarities.
- a set of sample input data and target data may be reduced in dimensionality via principal component analysis before being used to train a neural network.
- the neural network may use fewer hidden nodes and fewer calculations to achieve a properly tuned weighting function, as described in FIG. 3 . Consequently, in such embodiments the actual MR mapping data acquired in the technique of FIG. 4 will also be reduced via principal component analysis.
- the neural network produces an output, the resulting vector or matrix can be expanded across a basis set to produce a full coil sensitivity map or normalization. Imaging processes can therefore be performed in a standard fashion.
- a MR system includes a plurality of gradient coils, an RF coil assembly, a system control, and an image processing unit.
- the gradient coils are positioned about a bore of a magnet to impress a polarizing magnetic field therein.
- the RF coil assembly emits RF pulse sequences and receives resulting MR data from a subject within the bore.
- the system control is configured to develop a coil sensitivity mapping function from sample sets of MR data and determine a given coil sensitivity map which is applied by the imaging processing unit in an MR imaging procedure.
- the present invention is also embodied in a method for sensitivity mapping of RF coils.
- the method includes training a neural network to determine sensitivity profiles, acquiring MR mapping data at a given transmit power and with a given RF pulse characteristic for each of a plurality of coils of an MR system, and inputting the MR mapping data to the neural network to obtain a coil sensitivity profile.
- a coil sensitivity generator in accordance with another embodiment of the invention, includes a processing unit that operates a neural network to produce B1 maps.
- a training data interface is connected to receive sample MR data and input the data to train the neural network.
- MR mapping signals are acquired by a data acquisition unit from each coil of an RF coil assembly and transmitted to the neural network.
Abstract
Description
- The present invention relates generally to magnetic resonance imaging, and more specifically, to a system and method for mapping the sensitivity of radio frequency (RF) coils of a magnetic resonance (MR) system. The invention finds particular utility in parallel imaging applications by providing quickly determined coil sensitivity maps, requiring fewer flip angles/scans to determine.
- When a substance such as human tissue is subjected to a uniform magnetic field (polarizing field B0), the individual magnetic moments of the spins in the tissue attempt to align with this polarizing field, but precess about it in random order at their characteristic Larmor frequency. If the substance, or tissue, is subjected to a magnetic field (excitation field B1) which is in the x-y plane and which is near the Larmor frequency, the net aligned moment, or “longitudinal magnetization”, Mz, may be rotated, or “tipped”, into the x-y plane to produce a net transverse magnetic moment Mt. A signal is emitted by the excited spins after the excitation signal B1 is terminated and this signal may be received and processed to form an image.
- When utilizing these signals to produce images, magnetic field gradients (Gx, Gy, and Gz) are employed. Typically, the region to be imaged is scanned by a sequence of measurement cycles in which these gradients vary according to the particular localization method being used. The resulting set of received NMR signals are digitized and processed to reconstruct the image using one of many well known reconstruction techniques.
- Often, it is desirable in MR imaging procedures to map the strength and variations in a B1 magnetic field across the desired field of view, to decrease the potential effects on image reconstruction, or to allow for tailored excitation profiles across the field of view by manipulating individual currents in surface coils. Normalization or compensation functions can be generated from B1 maps and applied to the received NMR signals to account for magnetic field in-homogeneities. The same B1 maps can be used to extract information for driving the currents in individual RF coils in the transmit mode to achieve tailored excitations across the field of view. These mappings are usually performed for each scan subject, as subject positioning within the coil and subject-specific local physical properties (such as dielectric constant) can influence the coil B1 homogeneity.
- Methods for mapping the sensitivity of RF coils, or alternatively the uniformity of the B1 field, typically require that multiple data acquisitions take place for each coil of an RF coil assembly, where each data acquisition is taken at a different flip angle. This typically results in the use of 2-8 different flip angles or transmit powers for each coil of an N-channel coil array. Therefore, a total of 2N-8N different acquisitions are often used to generate one sensitivity mapping. The signals acquired in these acquisitions are proportional to the sine of the flip angle. Therefore, a pixel by pixel fit of the intensity of the signal versus the flip angle (from the known RF pulse height or width) can produce a map of the coil sensitivity. This method of coil sensitivity mapping is well known in the art. However, as the number of calibration or sensitivity-measuring scans required to produce this pixel by pixel fit is increased, overall scan time and patient throughput decrease.
- One particular application where coil sensitivity mapping is particularly advantageous is parallel imaging. Usually the system of spins is excited using a homogeneous coil (such as a whole-body RF coil), and then an array of surface RF coils is used to receive the MR signals, in order to increase the step size between phase-encoding lines, or equivalently to reduce the size of the field of view and the amount of data collected. Scan time reduction is achieved by under-sampling k-space and recording images simultaneously from the multiple imaging or receive coils. Under-sampling generally reduces the data acquisition time by increasing the distance of sampling positions in k-space.
- Parallel imaging techniques not only expedite data acquisition, but also reduce aliasing or wrapping that occurs in the phase-encoding direction when an imaging object extends outside a field-of-view (FOV). In particular, parallel imaging techniques remove or reduce the aliasing by using surface coil B1 fields (sensitivities), to define or determine an un-aliased spin distribution. Information regarding the surface coil B1 fields or sensitivities is typically acquired with an external calibration or a self-calibration technique. Generally, the coil sensitivity data is used to weight the imaging data such that coil sensitivity is reflected in the reconstructed image, and, as a result, the coil sensitivity data reduces aliasing in the reconstructed image that can occur as a result of under-sampling.
- In a manner similar to that previously described, an array of RF coils can also be used to both transmit and receive the MR signals (as opposed to using a homogenous RF coil for transmitting and many surface coils for receiving the signal). Such an approach has been shown to be useful for RF shimming, pulse designs for small FOV imaging, or reducing specific absorption rate for certain other MR applications. For these applications, methods of pulse design typically rely on knowledge of the particular excitation and reception pattern of each of the small RF coils employed.
- In all these applications, the time spent determining coil sensitivity should be as little as possible. It would therefore be desirable to have a system and method capable of quickly producing coil sensitivity maps while requiring a reduced number of mapping acquisitions.
- The present invention provides for rapid coil sensitivity mapping for MR systems using computer-developed normalization or weighting functions. Neural networks or similar computer intelligence are exemplary techniques employed to generate the weighting functions, further allowing computers to construct coil sensitivity profiles for each coil of an RF coil assembly with minimal input data.
- Therefore, in accordance with one aspect of the invention, an MRI system includes a plurality of gradient coils positioned about a bore of a magnet to impress a polarizing magnetic field, an RF coil assembly, a system control, and an image processing unit. The RF coil assembly is connected to emit RF pulse sequences and is arranged to receive MR data from a subject within the bore. The system control is configured to develop a coil sensitivity mapping function from sample sets of MR data and generate a given coil sensitivity map. The image processing unit is configured to apply the given coil sensitivity map in an MR imaging procedure.
- In accordance with another aspect of the invention, the invention is embodied in a method for sensitivity mapping of RF coils. The method includes training a neural network to determine coil sensitivity profiles, acquiring MR mapping data at a given transmit power and with a given RF pulse characteristic for each of a plurality of coils of an MR system, and inputting the MR mapping data to the neural network to obtain a coil sensitivity profile for each of the plurality of coils.
- In accordance with a further aspect of the invention, a coil sensitivity map generator is disclosed. The generator includes a processing unit programmed to operate a neural network to produce B1 maps, a training data interface connected to receive sample MR data and input the sample MR data to train the neural network, and a data acquisition unit configured to acquire MR mapping signals from each coil of an RF coil assembly and transmit the signals to the neural network.
- Various other features and advantages of the present invention will be made apparent from the following detailed description and the drawings.
- The drawings illustrate one preferred embodiment presently contemplated for carrying out the invention.
- In the drawings:
-
FIG. 1 is a schematic block diagram of an MR imaging system incorporating the present invention. -
FIG. 2 is a flowchart representing an exemplary sensitivity mapping technique. -
FIG. 3 is a flowchart representing a training technique in accordance with the present invention. -
FIG. 4 is a flowchart representing a sensitivity mapping technique, used after the system is trained, in accordance with the present invention. - Referring to
FIG. 1 , the major components of an example magnetic resonance imaging (MRI) system incorporating the present invention are shown. The operation of the system may be controlled from anoperator console 12 which includes a keyboard orother input device 13, acontrol panel 14, and adisplay screen 16. Theinput device 13 can include a mouse, keyboard, track ball, touch activated screen, light wand, or any similar or equivalent input device, and may be used for interactive geometry prescription. Theconsole 12 communicates through alink 18 with aseparate computer system 20 that enables an operator to control the production and display of images on thedisplay screen 16. Thecomputer system 20 includes a number of modules which communicate with each other through a backplane 20 a. These include animage processor module 22, aCPU module 24 and amemory module 26, which may include a frame buffer for storing image data arrays. Thecomputer system 20 may also be connected to permanent or back-up memory storage, a network, or may communicate with aseparate system control 32 through a highspeed serial link 34. - The
system control 32 includes a set of modules connected together by abackplane 32 a and connected to theoperator console 12 through aserial link 40. It is throughlink 40 that thesystem control 32 receives commands from the operator to indicate the scan sequence that is to be performed. The pulsesequence transmit module 46 commands the scanner components to carry out the desired scan sequence, by sending instructions, commands, and/or requests describing the timing, strength and shape of the RF pulses and pulse sequences to be produced, to correspond to the timing and length of the data acquisition window. Thesystem control 32 also connects to a set ofgradient amplifiers 42, to indicate the timing and shape of the gradient pulses that are produced during the scan. Thesystem control 32 may also receive patient data from ascan room interface 44, which may relate data from a user or from a number of different sensors connected to the patient, such as ECG signals from electrodes attached to the patient. - The gradient waveform instructions produced by
system control 32 are sent to thegradient amplifier system 42 having Gx, Gy, and Gz amplifiers.Amplifiers 42 may be external ofscanner 48, or may be integrated therein. Each gradient amplifier excites a corresponding physical gradient coil in a gradient coil assembly generally designated 50 to produce the magnetic field gradients used for spatially encoding acquired signals. Thegradient coil assembly 50 forms part of amagnet assembly 52 which includes apolarizing magnet 54 and anRF coil assembly coil 56, surface or parallel imaging coils 58, or both. Thecoils scanner equipment 48 produces RF pulses in accordance with the instructions of the pulse sequence transmitmodule 46 which are amplified and coupled to theRF coil 56 for transmission. Alternatively, RF transmitcoil 56 may be replaced or augmented with surface and/or parallel transmit coils, such ascoil 58. Similarly, the resulting signals emitted by the excited nuclei in the patient may be sensed by separate receive coils, such as parallel coils or surface coils 58, and are then sent over adata link 60. The MR signals are demodulated, filtered, and digitized in thedata processing section 62 of thesystem control 32. - A scan is complete when an array of raw k-space data has been acquired in the
memory module 66. This raw k-space data is rearranged into separate k-space data arrays for each image to be reconstructed, and each of these is input to anarray processor 64 which operates to Fourier transform the data into an array of image data. This image data is conveyed through theserial link 34 to thecomputer system 20 where it is stored inmemory 26. In response to commands received from theoperator console 12, this image data may be archived in long term storage or may be further processed by theimage processor 22 and conveyed to theoperator console 12 and presented on thedisplay 16. - In applications such as parallel imaging, RF shimming, reducing specific absorption rates, or narrow FOV imaging, it will be desirable to map the sensitivity and uniformity (or non-uniformity) of parallel or surface coils 58. This sensitivity map may then be used by
system control 32 orcomputer system 20 in subsequent data acquisition, processing or image reconstruction to normalize the signal outputs ofcoils 58, or otherwise account for field inhomogeneities, or to generate individualized or tailored excitation profiles which have particular characteristics, such as for narrow FOV imaging or controlling specific absorption rates. - One particular way to determine coil sensitivity maps is illustrated in the flowchart of
FIG. 2 . It should be noted, however, that the techniques described herein and depicted inFIG. 2 are merely exemplary, and that other known coil sensitivity mapping techniques may equivalently be substituted for or combined with the following examples. The technique ofFIG. 2 begins with MR signals being acquired 80 for one coil using an RF pulse at a first flip angle α1. Using the same coil, an additional set of MR data is acquired 82 using an RF pulse at the same transit gain, but with a second flip angle α2 which is 2*α1. The flip angle may be doubled by altering a pulse width or amplitude of the RF excitation pulse. Alternatively, the second applied flip angle α2 may be half of the first flip angle, or some other proportion which lends to a determination of actual flip angle, as described below. This process of MR data acquisition is repeated for each of the N coils of a multi-channel array, such as a surface coil array. Thus, each subsequent coil is individually controlled to acquire a first set ofMR data 84 at the same transmit gain with flip angle α1, and a second set of MR data is acquired 86 at the same transmit gain but with aflip angle 2*α1. From the two sets of MR data acquired for all coils of the RF coil assembly, two images may be reconstructed 88. - Since the intensity of the pixels of a reconstructed image is generally proportional to sin(α), the ratio of the intensity of the two images reconstructed from one coil will be proportional to cos(α), as the relation between the two applied flip angles, α2=2*α1, is known. Therefore, this ratio of intensities may be expressed 90 for each pixel of the two images for each coil as I2/2* I1, where I2 is the set of pixel intensities for the second image acquired with α2 and I1 is the set of pixel intensities for the first image acquired with α1. This ratio is independent of spin density and transversal relaxation time. A measure of the actual flip angle generated by each coil can be evaluated 92 as α=arccos(I2/2*I1) for each pixel intensity of the two images of each coil. It is this pixel by pixel evaluation which produces a sensitivity map as a spatial distribution or matrix of apparent/actual flip angles.
- In the more general case (i.e. when acquisitions with more than two flip angles are used or when the flip angles are not related by a factor of 2), a pixel by pixel fit of image intensity versus transmit power can generate a spatial map of flip angles produced by each coil at each location in space. The number of flip angles i used to acquire sets of MR data with each coil may depend upon the degree of B1 inhomogeneity or the desired precision in coil sensitivity determination. Thus, a set of i images can be reconstructed for each coil, corresponding to the i flip angles applied, α1, α2, . . . αi. And, for each coil, a table of image intensities will exist, I1, I2, . . . Ii. Through fitting processes, a maximum image intensity parameter and an actual flip angle parameter can be fit to the data sets of applied flip angles and actual image intensities. Once a fit of these data points has produced an intensity parameter and a flip angle parameter, coil sensitivity maps can be determined as distributions of the flip angle parameters across the desired field of view.
- The present invention utilizes neural networks, or other forms of computer intelligence, to ultimately produce coil sensitivity maps faster than the techniques described above with respect to
FIG. 2 . Neural networks operate by managing the coefficients of a weighting function that converts a set of inputs into a set of outputs. Neural networks “learn” (or are trained) by adjusting the values of the weighting function in accordance with complete data sets of sample inputs and target outputs. The weighting function is often represented as a hidden logic level between the inputs and outputs of the system. Usually, each node of a hidden level reads all input values, multiplies by various weights or coefficients, and produces one or more output values. The initial stage during which a neural network tunes its weighting function to consistently fit sets of sample inputs and target outputs is commonly referred to as the training stage. Once trained, neural networks can be used to rapidly produce outputs based on a limited subset of real input data. - One method for updating or adjusting the weights of the hidden level, or weighting function, is known as back-propagation. An arbitrary set of weights is given to each node of the hidden level, and a weight adjustment increment is set. When a sample set of inputs is presented, the arbitrary weighting function produces a set of outputs. These outputs are compared to a set of target outputs to determine how the weights must be adjusted or tuned to produce a better fit of the target outputs. After multiple sets of sample inputs and target outputs have been fed to the neural network, the weights will eventually become tuned to consistently produce an accurate fit or prediction of the target outputs.
- An exemplary weight adjustment function for each given node of a hidden level is given by:
-
- where o(x) is a vector of one or more outputs of the given hidden node, determined as weight vector w multiplied by input vector x. During training, each given hidden node will determine the error E (as a function of the weight vector w) between its computed output od and the target output td for each training sample set of inputs and outputs d of a dataset of training examples D by:
-
- An exemplary back-propagation weight-adjustment function, based on the determined error E between computed and target outputs can be described as:
-
- where Δwji is the change in the weighting coefficient associated with the ith input to hidden unit j, η is the chosen weight adjustment increment (such as 0.001, for example), tj is the target output for unit j, oj is the computed output of unit j, and xji is the ith training input value to hidden unit j. Such a function is generally described as being a gradient descent-based algorithm (i.e. it “descends” through a hypothesis space of potential weight functions), and may be further manipulated to take into account such things as momentum (varying η based on the degree of error E), thresholded output values, dynamic hidden layers, stopping algorithms, and local-minima compensation algorithms (to overcome false indications of optimization based upon local minima of hypothesis space). In sum, one of ordinary skill in the art will appreciate that the weight-adjustment function and other features of a neural network used in accordance with the present invention may take many forms and variations.
- Referring now to
FIG. 3 , atraining technique 98 of the present invention, preferably incorporating the use of a neural network, is shown in flowchart form.Technique 98 begins with the acquisition of MR mapping data for each coil of anRF coil assembly 100. MR mapping data is acquired multiple times for each coil, at varying flip angles. This may be performed in a manner similar to that shown inFIG. 2 . In addition, MR mapping data may be acquired with all coils simultaneously 102 at a different flip angle from those used in the initial acquisition. In a preferred embodiment, this flip angle may be twice or half the first flip angle used for each coil. Once all mapping data has been acquired, coil sensitivity profiles are mapped for eachcoil 104, as known in the art. - The MR mapping data acquired from each coil at the first flip angle and the MR mapping data acquired from all coils simultaneously are then fed to a neural network as a set of sample input values (xji) 106. The calculated sensitivity maps are fed to the network as the corresponding set of target outputs (tj) 108. The neural network is instructed to tune its weight function for the hidden level (or may do so automatically) 110 in accordance with the desired weight adjustment algorithm and associated features. If the network determines 112 that the weight function has been optimized 114, the training process ends 116. Otherwise, if the weight function has not yet been optimized 118, then the training process repeats 120. The number of repetitions will typically be proportionate to the dimensionality of the input matrix and target output matrix.
- Referring now to
FIG. 4 , atechnique 122 for using a fully-trained neural network (as achieved by the process ofFIG. 3 ) to produce coil sensitivity maps for use in image reconstruction is shown. First, MR mapping data (e.g. low resolution data), is acquired 124 from each coil of an RF coil assembly with an RF pulse at a first flip angle, α1, and having a given transmit gain and a given RF wave characteristic, such as a particular amplitude or pulse width. MR mapping data may then be acquired 126 from all coils simultaneously at a second flip angle, α2. Preferably, the second flip angle is equal to double or half the first flip angle. This may be done by modifying the given RF wave amplitude or pulse width from the first set of acquisitions. However, any flip angle not equal to the first flip angle may be used for thesecond acquisition 126. Further, depending upon the degree of B1 inhomogeneity, thesecond acquisition 126 may be optional. - The MR mapping data acquired from the first set of acquisitions 124 and the
second acquisition 126 are then fed to the trained neural network as input values 128. The neural network is caused to produce an output via its tuned weight function, which is received by a system control or other processing unit as a coil sensitivity map or set of sensitivity profiles 130. The sensitivity map may be used for image data normalization or tailoring specific coil currents. This process may take place in a significantly shorter time due to the reduced number of mapping data acquisitions and to the rapidity of neural network calculations. - An imaging process may take place in which higher resolution MR signals are acquired in a
conventional manner 132. In a data processing or image reconstruction stage, the sensitivity map or normalization computed by the neural network may be applied to the MR signals 134 to compensate for the non-uniformities of the B1 field. Alternatively, a parallel imaging reconstruction may be performed based on the sensitivity map, or a pulse sequence (e.g. tailored currents for each coil to produce specific excitation pulses) may be designed for specific or narrow field of view imaging or other uses. - Alternatively, to speed up both the training and calculating functions of the neural network, a data dimensionality reduction procedure, such as a principal component analysis, may be used to reduce the dimensionality of the input and target output samples of the neural network, as well as the higher-resolution MR imaging inputs. As known in the art, a principal component analysis is a technique for reducing the expression of given datasets (such as images or sets of MR signals) to their variations from a number of basis sets. In other words, a number of datasets can be analyzed to factor out commonalities and major variations, which are expressed as a basis set. Then, each given dataset is reduced to a vector or matrix of its projection onto the basis set. The result is a significantly reduced way to express datasets which have similarities.
- Therefore, a set of sample input data and target data may be reduced in dimensionality via principal component analysis before being used to train a neural network. Thus, the neural network may use fewer hidden nodes and fewer calculations to achieve a properly tuned weighting function, as described in
FIG. 3 . Consequently, in such embodiments the actual MR mapping data acquired in the technique ofFIG. 4 will also be reduced via principal component analysis. When the neural network produces an output, the resulting vector or matrix can be expanded across a basis set to produce a full coil sensitivity map or normalization. Imaging processes can therefore be performed in a standard fashion. - Accordingly, in one embodiment of the present invention, a MR system includes a plurality of gradient coils, an RF coil assembly, a system control, and an image processing unit. The gradient coils are positioned about a bore of a magnet to impress a polarizing magnetic field therein. The RF coil assembly emits RF pulse sequences and receives resulting MR data from a subject within the bore. The system control is configured to develop a coil sensitivity mapping function from sample sets of MR data and determine a given coil sensitivity map which is applied by the imaging processing unit in an MR imaging procedure.
- The present invention is also embodied in a method for sensitivity mapping of RF coils. The method includes training a neural network to determine sensitivity profiles, acquiring MR mapping data at a given transmit power and with a given RF pulse characteristic for each of a plurality of coils of an MR system, and inputting the MR mapping data to the neural network to obtain a coil sensitivity profile.
- In accordance with another embodiment of the invention, a coil sensitivity generator is disclosed. The generator includes a processing unit that operates a neural network to produce B1 maps. A training data interface is connected to receive sample MR data and input the data to train the neural network. MR mapping signals are acquired by a data acquisition unit from each coil of an RF coil assembly and transmitted to the neural network.
- The present invention has been described in terms of the preferred embodiment, and it is recognized that equivalents, alternatives, and modifications, aside from those expressly stated, are possible and within the scope of the appending claims.
Claims (26)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US11/554,516 US7375523B1 (en) | 2006-10-30 | 2006-10-30 | System and method for fast MR coil sensitivity mapping |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US11/554,516 US7375523B1 (en) | 2006-10-30 | 2006-10-30 | System and method for fast MR coil sensitivity mapping |
Publications (2)
Publication Number | Publication Date |
---|---|
US20080100292A1 true US20080100292A1 (en) | 2008-05-01 |
US7375523B1 US7375523B1 (en) | 2008-05-20 |
Family
ID=39329359
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US11/554,516 Active US7375523B1 (en) | 2006-10-30 | 2006-10-30 | System and method for fast MR coil sensitivity mapping |
Country Status (1)
Country | Link |
---|---|
US (1) | US7375523B1 (en) |
Cited By (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080150528A1 (en) * | 2006-12-21 | 2008-06-26 | The Board Of Trustees Of The Leland Stanford Junior University | Rf field mapping for magnetic resonance imaging |
US20080231274A1 (en) * | 2007-03-20 | 2008-09-25 | Joerg Ulrich Fontius | Method to control a magnetic resonance system |
US20100066361A1 (en) * | 2008-04-10 | 2010-03-18 | Kawin Setsompop | Method for fast magnetic resonance radiofrequency coil transmission profile mapping |
US20100121179A1 (en) * | 2008-11-13 | 2010-05-13 | Pacesetter, Inc. | Systems and Methods for Reducing RF Power or Adjusting Flip Angles During an MRI for Patients with Implantable Medical Devices |
US20100315084A1 (en) * | 2009-06-15 | 2010-12-16 | Sacolick Laura I | System, method, and apparatus for magnetic resonance rf-field measurement |
WO2011039692A2 (en) | 2009-10-02 | 2011-04-07 | Koninklijke Philips Electronics N.V. | Mr imaging using multi-channel rf excitation |
US20130289912A1 (en) * | 2012-03-30 | 2013-10-31 | Siemens Aktiengesellschaft | Eigen-vector approach for coil sensitivity maps estimation |
US20140232394A1 (en) * | 2013-02-20 | 2014-08-21 | Industry-Academic Cooperation Foundation, Yonsei University | Method and apparatus for obtaining main magnetic field information and radio pulse related information in a magnetic resonance imaging system with different flip angles |
US20140253121A1 (en) * | 2011-11-08 | 2014-09-11 | Hitachi Medical Corporation | Magnetic resonance imaging apparatus and irradiation magnetic field distribution measurement method |
DE102013218224B3 (en) * | 2013-09-11 | 2015-01-29 | Siemens Aktiengesellschaft | Determination of B1 cards |
WO2015159664A1 (en) * | 2014-04-16 | 2015-10-22 | 株式会社 日立メディコ | Nuclear magnetic resonance imaging device and rf shimming method |
US20160327606A1 (en) * | 2014-01-03 | 2016-11-10 | Koninklijke Philips N.V. | Calculation of the probability of gradient coil amplifer failure using environment data |
CN108335339A (en) * | 2018-04-08 | 2018-07-27 | 朱高杰 | A kind of magnetic resonance reconstruction method based on deep learning and convex set projection |
CN109791686A (en) * | 2016-07-15 | 2019-05-21 | 弗劳恩霍夫应用研究促进协会 | System for generating image, especially magnetic resonance system |
CN109863512A (en) * | 2016-09-01 | 2019-06-07 | 通用医疗公司 | System and method for carrying out auto-changing by manifold approximation |
US10527699B1 (en) * | 2018-08-01 | 2020-01-07 | The Board Of Trustees Of The Leland Stanford Junior University | Unsupervised deep learning for multi-channel MRI model estimation |
CN111513716A (en) * | 2019-02-05 | 2020-08-11 | 通用电气精准医疗有限责任公司 | Method and system for magnetic resonance image reconstruction using extended sensitivity model and deep neural network |
JP2020121032A (en) * | 2019-01-31 | 2020-08-13 | キヤノンメディカルシステムズ株式会社 | Medical image processing apparatus, magnetic resonance imaging apparatus, medical image processing method, and program |
CN111656392A (en) * | 2018-02-15 | 2020-09-11 | 通用电气公司 | System and method for synthesizing magnetic resonance images |
CN112240995A (en) * | 2019-07-17 | 2021-01-19 | 通用电气精准医疗有限责任公司 | For predicting B from magnetic resonance calibration images1+System and method for mapping maps |
US11231476B2 (en) * | 2019-04-26 | 2022-01-25 | Regents Of The University Of Minnesota | Accelerated magnetic resonance imaging acquisition using two-dimensional pulse segments as virtual receivers |
CN114599985A (en) * | 2019-09-13 | 2022-06-07 | 西达-赛奈医疗中心 | Deep learning system and method for large-scale dynamic magnetic resonance image reconstruction |
US11385309B1 (en) * | 2021-04-29 | 2022-07-12 | Shanghai United Imaging Healthcare Co., Ltd. | Systems and methods for actual gradient waveform estimation |
Families Citing this family (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2008026174A2 (en) * | 2006-08-30 | 2008-03-06 | Koninklijke Philips Electronics N.V. | Multi-channel magnetic resonance imaging and spectroscopy |
US8049497B2 (en) * | 2006-10-31 | 2011-11-01 | Koninklijke Philips Electronics N.V. | MRI RF encoding using multiple transmit coils |
JP5355866B2 (en) * | 2007-06-14 | 2013-11-27 | ジーイー・メディカル・システムズ・グローバル・テクノロジー・カンパニー・エルエルシー | Image generation method, position correction method, and magnetic resonance imaging apparatus |
DE102007047020B4 (en) * | 2007-10-01 | 2012-07-12 | Siemens Ag | Arrangement for transmitting magnetic resonance signals |
JP5471072B2 (en) * | 2009-06-26 | 2014-04-16 | 富士通株式会社 | Display test apparatus, display test program, and display test method |
US9146293B2 (en) | 2012-02-27 | 2015-09-29 | Ohio State Innovation Foundation | Methods and apparatus for accurate characterization of signal coil receiver sensitivity in magnetic resonance imaging (MRI) |
US9316707B2 (en) * | 2012-04-18 | 2016-04-19 | General Electric Company | System and method of receive sensitivity correction in MR imaging |
DE102014203057B4 (en) * | 2014-02-20 | 2018-12-27 | Siemens Healthcare Gmbh | Method for determining a radio-frequency pulse by means of vector representation of sample points |
US10614359B2 (en) * | 2014-04-19 | 2020-04-07 | Case Western Reserve University | Magnetic resonance fingerprinting (MRF) serial artificial neural network (ANN) sequence design |
KR101659578B1 (en) | 2015-09-01 | 2016-09-23 | 삼성전자주식회사 | Method and apparatus for processing magnetic resonance imaging |
WO2019070848A1 (en) | 2017-10-06 | 2019-04-11 | University Of Cincinnati | Systems and methods for estimating complex radiofrequency fields in a magnetic resonance imaging |
US11125845B2 (en) * | 2019-03-22 | 2021-09-21 | Canon Medical Systems Corporation | Apparatus and method for deep learning to mitigate artifacts arising in simultaneous multi slice (SMS) magnetic resonance imaging (MRI) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5910728A (en) * | 1996-11-12 | 1999-06-08 | Beth Israel Deaconess Medical Center | Simultaneous acquisition of spatial harmonics (SMASH): ultra-fast imaging with radiofrequency coil arrays |
US6728404B1 (en) * | 1991-09-12 | 2004-04-27 | Fuji Photo Film Co., Ltd. | Method for recognizing object images and learning method for neural networks |
US6949928B2 (en) * | 2001-01-19 | 2005-09-27 | Kabushiki Kaisha Toshiba | Parallel MR imaging using high-precision coil sensitivity map |
-
2006
- 2006-10-30 US US11/554,516 patent/US7375523B1/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6728404B1 (en) * | 1991-09-12 | 2004-04-27 | Fuji Photo Film Co., Ltd. | Method for recognizing object images and learning method for neural networks |
US5910728A (en) * | 1996-11-12 | 1999-06-08 | Beth Israel Deaconess Medical Center | Simultaneous acquisition of spatial harmonics (SMASH): ultra-fast imaging with radiofrequency coil arrays |
US6949928B2 (en) * | 2001-01-19 | 2005-09-27 | Kabushiki Kaisha Toshiba | Parallel MR imaging using high-precision coil sensitivity map |
Cited By (45)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7446526B2 (en) * | 2006-12-21 | 2008-11-04 | The Board Of Trustees Of The Leland Stanford Junior University | RF field mapping for magnetic resonance imaging |
US20080150528A1 (en) * | 2006-12-21 | 2008-06-26 | The Board Of Trustees Of The Leland Stanford Junior University | Rf field mapping for magnetic resonance imaging |
US7719281B2 (en) * | 2007-03-20 | 2010-05-18 | Siemens Aktiengesellschaft | Method to control a magnetic resonance system with individually controllable RF transmission channels |
US20080231274A1 (en) * | 2007-03-20 | 2008-09-25 | Joerg Ulrich Fontius | Method to control a magnetic resonance system |
US8076939B2 (en) * | 2008-04-10 | 2011-12-13 | The General Hospital Corporation | Method for fast magnetic resonance radiofrequency coil transmission profile mapping |
US20100066361A1 (en) * | 2008-04-10 | 2010-03-18 | Kawin Setsompop | Method for fast magnetic resonance radiofrequency coil transmission profile mapping |
US20100121179A1 (en) * | 2008-11-13 | 2010-05-13 | Pacesetter, Inc. | Systems and Methods for Reducing RF Power or Adjusting Flip Angles During an MRI for Patients with Implantable Medical Devices |
US20100315084A1 (en) * | 2009-06-15 | 2010-12-16 | Sacolick Laura I | System, method, and apparatus for magnetic resonance rf-field measurement |
US8198891B2 (en) | 2009-06-15 | 2012-06-12 | General Electric Company | System, method, and apparatus for magnetic resonance RF-field measurement |
WO2011039692A2 (en) | 2009-10-02 | 2011-04-07 | Koninklijke Philips Electronics N.V. | Mr imaging using multi-channel rf excitation |
WO2011039692A3 (en) * | 2009-10-02 | 2011-11-10 | Koninklijke Philips Electronics N.V. | Mr imaging using multi-channel rf excitation |
CN102576061A (en) * | 2009-10-02 | 2012-07-11 | 皇家飞利浦电子股份有限公司 | MR imaging using multi-channel RF excitation |
US8938281B2 (en) | 2009-10-02 | 2015-01-20 | Koninklijke Philips N.V. | MR imaging using multi-channel RF excitation |
US20140253121A1 (en) * | 2011-11-08 | 2014-09-11 | Hitachi Medical Corporation | Magnetic resonance imaging apparatus and irradiation magnetic field distribution measurement method |
US10162027B2 (en) * | 2011-11-08 | 2018-12-25 | Hitachi, Ltd. | Magnetic resonance imaging apparatus and irradiation magnetic field distribution measurement method |
US20130289912A1 (en) * | 2012-03-30 | 2013-10-31 | Siemens Aktiengesellschaft | Eigen-vector approach for coil sensitivity maps estimation |
US10914798B2 (en) * | 2012-03-30 | 2021-02-09 | Siemens Healthcare Gmbh | Eigen-vector approach for coil sensitivity maps estimation |
US20140088899A1 (en) * | 2012-03-30 | 2014-03-27 | Siemens Aktiengesellschaft | Eigen-vector approach for coil sensitivity maps estimation |
US20140232394A1 (en) * | 2013-02-20 | 2014-08-21 | Industry-Academic Cooperation Foundation, Yonsei University | Method and apparatus for obtaining main magnetic field information and radio pulse related information in a magnetic resonance imaging system with different flip angles |
US9500733B2 (en) * | 2013-02-20 | 2016-11-22 | Samsung Electronics Co., Ltd. | Method and apparatus for obtaining main magnetic field information and radio pulse related information in a magnetic resonance imaging system with different flip angles |
DE102013218224B3 (en) * | 2013-09-11 | 2015-01-29 | Siemens Aktiengesellschaft | Determination of B1 cards |
CN104414643A (en) * | 2013-09-11 | 2015-03-18 | 西门子公司 | Establishing B1-Maps |
US9714996B2 (en) | 2013-09-11 | 2017-07-25 | Siemens Aktiengesellschaft | Establishing B1-maps |
US20160327606A1 (en) * | 2014-01-03 | 2016-11-10 | Koninklijke Philips N.V. | Calculation of the probability of gradient coil amplifer failure using environment data |
US10444277B2 (en) * | 2014-01-03 | 2019-10-15 | Koninklijke Philips N.V. | Calculation of the probability of gradient coil amplifier failure using environment data |
JPWO2015159664A1 (en) * | 2014-04-16 | 2017-04-13 | 株式会社日立製作所 | Nuclear magnetic resonance imaging apparatus and RF shimming method |
US10156616B2 (en) | 2014-04-16 | 2018-12-18 | Hitachi, Ltd. | Nuclear magnetic resonance imaging apparatus and RF shimming method |
WO2015159664A1 (en) * | 2014-04-16 | 2015-10-22 | 株式会社 日立メディコ | Nuclear magnetic resonance imaging device and rf shimming method |
CN109791686B (en) * | 2016-07-15 | 2023-10-24 | 弗劳恩霍夫应用研究促进协会 | System for generating images, in particular magnetic resonance system |
CN109791686A (en) * | 2016-07-15 | 2019-05-21 | 弗劳恩霍夫应用研究促进协会 | System for generating image, especially magnetic resonance system |
US11055883B2 (en) * | 2016-07-15 | 2021-07-06 | Fraunhofer-Gesellschaft Zur Foerderung Der Angewandten Forschung E.V. | Magnetic resonance system and method for producing images |
CN109863512A (en) * | 2016-09-01 | 2019-06-07 | 通用医疗公司 | System and method for carrying out auto-changing by manifold approximation |
US11620772B2 (en) | 2016-09-01 | 2023-04-04 | The General Hospital Corporation | System and method for automated transform by manifold approximation |
CN111656392A (en) * | 2018-02-15 | 2020-09-11 | 通用电气公司 | System and method for synthesizing magnetic resonance images |
CN108335339A (en) * | 2018-04-08 | 2018-07-27 | 朱高杰 | A kind of magnetic resonance reconstruction method based on deep learning and convex set projection |
US10527699B1 (en) * | 2018-08-01 | 2020-01-07 | The Board Of Trustees Of The Leland Stanford Junior University | Unsupervised deep learning for multi-channel MRI model estimation |
JP2020121032A (en) * | 2019-01-31 | 2020-08-13 | キヤノンメディカルシステムズ株式会社 | Medical image processing apparatus, magnetic resonance imaging apparatus, medical image processing method, and program |
JP7213099B2 (en) | 2019-01-31 | 2023-01-26 | キヤノンメディカルシステムズ株式会社 | Medical image processing device, magnetic resonance imaging device, medical image processing method, and program |
CN111513716A (en) * | 2019-02-05 | 2020-08-11 | 通用电气精准医疗有限责任公司 | Method and system for magnetic resonance image reconstruction using extended sensitivity model and deep neural network |
US11231476B2 (en) * | 2019-04-26 | 2022-01-25 | Regents Of The University Of Minnesota | Accelerated magnetic resonance imaging acquisition using two-dimensional pulse segments as virtual receivers |
US11009578B2 (en) * | 2019-07-17 | 2021-05-18 | GE Precision Healthcare LLC | Systems and methods for predicting B1+ maps from magnetic resonance calibration images |
CN112240995A (en) * | 2019-07-17 | 2021-01-19 | 通用电气精准医疗有限责任公司 | For predicting B from magnetic resonance calibration images1+System and method for mapping maps |
CN114599985A (en) * | 2019-09-13 | 2022-06-07 | 西达-赛奈医疗中心 | Deep learning system and method for large-scale dynamic magnetic resonance image reconstruction |
US11385309B1 (en) * | 2021-04-29 | 2022-07-12 | Shanghai United Imaging Healthcare Co., Ltd. | Systems and methods for actual gradient waveform estimation |
US11703557B2 (en) | 2021-04-29 | 2023-07-18 | Shanghai United Imaging Healthcare Co., Ltd. | Systems and methods for actual gradient waveform estimation |
Also Published As
Publication number | Publication date |
---|---|
US7375523B1 (en) | 2008-05-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7375523B1 (en) | System and method for fast MR coil sensitivity mapping | |
US8082127B2 (en) | Method for determining the spatial distribution of magnetic resonance signals through multi-dimensional RF excitation pulses | |
US6369568B1 (en) | Fast spin echo phase correction for MRI system | |
US7633293B2 (en) | Radio frequency field localization for magnetic resonance | |
US7705594B2 (en) | System and method for designing improved RF pulse profiles | |
CN102159965B (en) | B1-mapping and b1l-shimming for mri | |
KR102236865B1 (en) | Establishing a magnetic resonance system actuation sequence | |
US7495437B2 (en) | Method for MRT imaging on the basis of conventional PPA reconstruction methods | |
US7492153B2 (en) | System and method of parallel imaging with calibration to a separate coil | |
US8938281B2 (en) | MR imaging using multi-channel RF excitation | |
US10488481B2 (en) | Systems and methods for multislice magetic resonance fingerprinting | |
EP1136836A2 (en) | T2 contrast in magnetic resonance imaging with gradient echoes | |
US8049497B2 (en) | MRI RF encoding using multiple transmit coils | |
US7038453B2 (en) | Method and magnetic resonance tomography apparatus for spatially resolved measurement of the B1 field distribution | |
US20160313421A1 (en) | Zero echo time mr imaging with water/fat separation | |
US20170261577A1 (en) | Zero echo time mr imaging | |
US7570141B2 (en) | Method of designing a shim coil to reduce field settling time | |
US5093620A (en) | Encoding for nmr phase contrast flow measurement | |
US6239597B1 (en) | Method and apparatus for rapid T2 weighted MR image acquisition | |
US20050192497A1 (en) | Magnetic resonance imaging method | |
US7952355B2 (en) | Apparatus and method for reconstructing an MR image | |
US11147519B2 (en) | Magnetic resonance imaging apparatus for generating parametric map from magnetic resonance signal data and method of operating the same | |
US20050256393A1 (en) | System and method for generating t1 and t2 maps | |
US7064544B1 (en) | Method and system of scaling MR spectroscopic data acquired with phased-array coils | |
US20230160982A1 (en) | Medical data processing apparatus, medical data processing method, and magnetic resonance imaging apparatus |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: GENERAL ELECTRIC COMPANY, NEW YORK Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:HANCU, ILEANA;REEL/FRAME:018455/0722 Effective date: 20061030 |
|
STCF | Information on status: patent grant |
Free format text: PATENTED CASE |
|
FEPP | Fee payment procedure |
Free format text: PAYOR NUMBER ASSIGNED (ORIGINAL EVENT CODE: ASPN); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY |
|
FPAY | Fee payment |
Year of fee payment: 4 |
|
FPAY | Fee payment |
Year of fee payment: 8 |
|
MAFP | Maintenance fee payment |
Free format text: PAYMENT OF MAINTENANCE FEE, 12TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1553); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY Year of fee payment: 12 |