Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Election/Restrictions
Applicant’s election without traverse of Group I, Claims 1-12 in the reply filed on 02/02/2026 is acknowledged. Claims 13-18 are drawn to Group II, and Claims 19-20 are drawn to Group III are withdrawn from further consideration pursuant to 37 CFR 1.142(b) as being drawn to a nonelected inventions, there being no allowable generic or linking claim. Election was made without traverse in the reply filed on 02/02/2026.
DETAILED ACTION
The following NON-FINAL Office Action is in response to application 18/468,756
filed on 09/18/2023. This communication is the first action on the merits.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 09/18/2023 has been
considered by the examiner.
Drawings
The drawings were received on 09/18/2023. These drawings are acceptable.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-12 are rejected under 35 U.S.C. 103 as being unpatentable over US 20190172230 A1, Mailhe et al (hereinafter Mailhe) in view of EP 4023152 A1, Sreenivasan et al (hereinafter Sreenivasan).
Regarding Claim 1, Mailhe discloses a method of establishing a magnetic resonance (MR) pulse sequence for a MR scanner (Mailhe, [0034] This disclosure describes a system and method using a deep neural network (DNN) for magnetic resonance (MR) image reconstruction), the method comprising:
receiving indications of an environment for MR scanning of a patient (Mailhe, [0139] the trained recurrent reconstruction engine 1104 and the trained policy estimation network 1310 with the learned policy receive the input feature vectors (i.e., input data for the particular patient));
establishing the MR pulse sequence (Mailhe, [0134] the MR scanner or other MR system scans the patient with the MR compressed sequence. Based on the configuration of the MR scanner, a pulse sequence is created. The pulse sequence is transmitted from coils into the patient) for the environment by reinforcement learned artificial intelligence (Mailhe, [0139] the image processor applies the deep reinforcement machine-learnt network to reconstruct. The actor-critic architecture, as trained, is applied. An initial reconstruction based on a model of the MR scanner generates an initial image. The initial image and any meta data used by the critic are input. For example, the initial image, power spectrum of the k-space data, sample pattern or mask from the scanning, and a calibrated noise level are input) based on input of the environment (Mailhe, [0113] The measurements, y, 1102 are the input k-space data, which is under sampled. For reinforcement, meta-data 1110 is also input. The meta-data is one or more measurements of covariance, such as the sampling pattern, noise level, and/or power spectrum. Other meta data may be used, such as any representation of the pulse sequence design (e.g., settings for contrast, a perceptual measure of an image (e.g., perceptual hash), repetition time, echo time, or other input to control operation of the MR scanner). Other meta data may include any calibration measurements, such as a level of cross-talk of coils); establishing comprising optimization by the reinforcement learned artificial
intelligence (Mailhe, [0141] The current image and meta data are used to determine a probability distribution of settings of a reconstruction parameter. The distribution provides the settings with greater and lesser probabilities of being rewarded to provide optimized reconstruction in a final MR image) of the MR pulse sequence in the environment for the digital twin of the patient (Mailhe, [0146] the feedback from act 1406 to act 1402 represents receiving another setting of the acceleration or other scan parameter and repeating the iterative reconstructing with the deep reinforcement machine-learnt network for the other setting. The same deep reinforcement machine-learnt network may be used for different scan settings and corresponding sample patterns. Different patients or the same patient at a different time may be scanned differently while the same trained reconstruction network is used to optimize the reconstruction of the MR image);
configuring the MR scanner with the MR pulse sequence (Mailhe, [0134] In act 1404, the MR scanner or other MR system scans the patient with the MR compressed sequence. Based on the configuration of the MR scanner, a pulse sequence is created);
imaging the patient by the MR scanner as configured with the MR pulse sequence (Mailhe, [0134] In act 1404, the MR scanner or other MR system scans the patient with the MR compressed sequence. Based on the configuration of the MR scanner, a pulse sequence is created. The pulse sequence is transmitted from coils into the patient. The resulting response of tissue is measured by receiving radio frequency signals at the same or different coils, [0137] In act 1406, an image processor iteratively reconstructs an MR image from the scan data. Any of various types of reconstruction may be used. The k-space data is Fourier transformed into scalar values representing different spatial locations, such as spatial locations representing a plane through the patient).
Mailhe does not disclose a digital twin of the Patient
However, Sreenivasan teaches a digital twin of the Patient (Sreenivasan, [Page 5] the patient profile generation module 30 may include an artificial intelligence (AI) module using a pre-trained AI model for determining the patient's behaviour and physical condition during the simulated scan procedure. The AI model may be pre-trained using the known ground truth samples that can be obtained by feedback from the patient as labels. The al6gorithm may be a combination of a machine learning approach for the estimation of the current stress level, such as support vector machine (SVM), convolutional neural network (CNN), etc., and a machine learning approach for predicting the development of the stress level during the next few minutes, such as recurrent neural network (RNN) or long short-term memory (LSTM));
Before the effective filing date of the claimed invention, It would have been obvious to one of ordinary skill in the art to combine Mailhe and Sreenivasan teaching because Mailhe teaches reconstructing MR images and optimizing reconstruction parameters using machine learning techniques based on acquired scan data. However, Mailhe does not disclose modeling the patient using a digital twin, while Sreenivasan teaches generating a patient specific model by simulating a scan environment and determining patient behavior and physiological condition using artificial intelligence, thereby providing a digital representation (digital twin) of the patient. A person of ordinary skill in the art would have integrated the patient modeling and simulation techniques of Sreenivasan into the system of Mailhe in order to improve the accuracy and effectiveness of MR imaging.
Regarding Claim 2, Mailhe in view of Sreenivasan discloses the method of claim 1 wherein receiving the indication comprises receiving state information for the environment (Mailhe, [0130] The method is performed by the systems of FIGS. 1, 9, and/or 15 or another system. The MR scanner receives scan settings based on user input of the user interface of the MR scanner and scans the patient. An image processor iteratively reconstructs the MR image using the machine-trained neural network, and a display displays the MR image), the state information used by an agent of the reinforcement learned artificial intelligence in establishing the MR pulse sequence (Mailhe, [0133] Any scan parameter for MR compressed sensing may be set. For example, the acceleration or other parameter changing the relative speed and image quality is set. The acceleration may be a setting for the type of sampling (e.g., relative number of measures by frequency), spatial density of sampling, or a number of measurements. Settings for contrast, the coil array being used, repetition time, echo time, or other setting establishing the pulse sequence of the MR compressed sensing may be set, [0140] The deep reinforcement machine-learnt network is trained to control the sequence of actions through the iterations based on the learned policy. The policy uses the current image and meta data to determine a change to be made in one or more settings of parameters of reconstruction).
Regarding Claim 3, Mailhe discloses the method of claim 1 wherein receiving the indication comprises receiving a body region of the patient (Mailhe, [0130] The MR scanner receives scan settings based on user input of the user interface of the MR scanner and scans the patient), an output task of the MR scanning (Mailhe, [0145] The output is to a display plane or buffer. Color mapping or other post reconstruction processing is used to generate the MR image), and specification of the MR scanner (Mailhe, [0134] In act 1404, the MR scanner or other MR system scans the patient with the MR compressed sequence. Based on the configuration of the MR scanner, a pulse sequence is created. The pulse sequence is transmitted from coils into the patient. The resulting response of tissue is measured by receiving radio frequency signals at the same or different coils), the body region, size, output task, and specification used by an agent of the reinforcement learned artificial intelligence in the optimization (Mailhe, [0141] The actions 1108 are determined based on the learned Markov decision process. The current image and meta data are used to determine a probability distribution of settings of a reconstruction parameter. The distribution provides the settings with greater and lesser probabilities of being rewarded to provide optimized reconstruction in a final MR image).
Mailhe does not disclose a size of the body region of the patient
However, Sreenivasan teaches a size of the body region of the patient (Sreenivasan, [Page 2] polyphonic, may be used to recreate spatial perception of e.g. noise, [Page 2] the scan simulation module, the patient monitoring module, and the patient profile generation module may be built-in units of ahead-mounted display (HDM). For example, the HMD is a virtual reality enabled HMD, which may have a built-in camera for monitoring the pupilsize of the patient and a processor for determining a stress level based on the monitored pupil size of the patient to create the patient profile.).
Before the effective filing date of the claimed invention, It would have been obvious to one of ordinary skill in the art to combine Mailhe and Sreenivasan’s teaching because Mailhe teaches a reinforcement learning framework for optimizing MR scanning and reconstruction parameters using patient specific input data but does not explicitly disclose incorporating a size of the body region of the patient as part of the optimization inputs. Sreenivasan teaches acquiring patient specific physiological and image based data, including parameters indiciative of patient size (e.g., pupil size and other image-derived characteristics), which inherently reflect spatial characteristics of the patient. A person of ordinary skill in the art would have been motivated to integrate such patient specific size related information from Sreenivasan into the system of Mailhe in order to improve the accuracy and effectiveness of MR scan parameter optimization by accounting for individual patient characteristics that affect imaging performance.
Regarding Claim 4, Mailhe in view of Sreenivasan discloses the method of claim 1 wherein establishing comprises establishing an agent of the reinforcement learned artificial intelligence (Mailhe, [0140] The trained recurrent reconstruction engine 1104 is configured by the training to respond to actions 1108 from the critic 1106. For each iteration of the reconstruction, one or more actions representing changes to be made in reconstruction parameters are received. The deep reinforcement machine-learnt network is trained to control the sequence of actions through the iterations based on the learned policy. The policy uses the current image and meta data to determine a change to be made in one or more settings of parameters of reconstruction) performing the optimization with a sequence of actions to adjust values of parameters of the MR pulse sequence (Mailhe, [0141] The current image and meta data are used to determine a probability distribution of settings of a reconstruction parameter. The distribution provides the settings with greater and lesser probabilities of being rewarded to provide optimized reconstruction in a final MR image. The action is determined by random or other sampling of the probability distribution. Probability distributions for more than one reconstruction parameter may be used. The sampling may select which type of action as well as the setting for the action to use. More than one action may be selected, such as altering settings for two or more reconstruction parameters for a given iteration).
Regarding Claim 5, Mailhe in view of Sreenivasan teaches the method of claim 1 wherein establishing comprises establishing values of multiple different parameters of the MR pulse sequence (Sreenivasan, [Page 5-6] In substep a3, the design parameters of the RF pulses and preferably the gradient waveforms are iteratively adjusted to maximize the merit function or minimize the cost function involving the recalculation of the FA distribution. Thus, N .sub.S, 0 pulse sequences are obtained, each sequence being optimal for a set of S0 objects).
Before the effective filing date of the claimed invention, It would have been obvious to one of ordinary skill in the art to combine Mailhe and Sreenivasan’s teaching because Mailhe teaches using a reinforcement learning framework to optimize MR reconstruction and scanning parameters but does not explicitly detail establishing values of multiple different MR pulse sequence parameters through iterative adjustment. Sreenivasan teaches iteratively adjusting multiple design parameters of MR pulse sequences including RF pulses and gradient waveforms, to optimize a merit or cost function and generation optimal pulse sequences. A person of ordinary skill in the art would have been motivated to combine Sreenivasan’s teaching of adjusting multiple MR pulse sequence parameters into the system of Mailhe in order to enhance the optimization process by enabling coordinated tuning of multiple parameters, thereby improving image quality and system performance.
Regarding Claim 6, Mailhe in view of Sreenivasan discloses the method of claim 5 wherein establishing comprises establishing values of all of the different parameters that can be set in MR scanner for the MR scanning of the patient (Mailhe, [0149] The user configures the MR scanner 100 to scan the patient, such as selecting one or more settings of scan parameters. The operator interface is used to set a scan parameter. The scan parameter results in a sample pattern. The under sampling is selectable).
Regarding Claim 7, Mailhe in view of Sreenivasan discloses the method of claim 5 wherein the multiple different parameters are parameters indicated by a user as to be established (Mailhe, [0132] In act 1402, user input of one or more settings are received. The user input is received through a user interface, such as with a keyboard and mouse or trackball. The user selects a preset or otherwise configures the MR scanner for scanning a particular patient. One or more individual scan parameters may be adjusted or set) and other parameters are fixed (Mailhe, [0118] Different arrangements may be used for different sample patterns, so the arrangement is another parameter that may be learned. In other embodiments, the arrangement is fixed or does not vary by sample pattern), wherein establishing comprises establishing the values of the multiple different parameters while maintaining values of the other parameters constant in the optimization (Mailhe, [0141] The sampling may select which type of action as well as the setting for the action to use. More than one action may be selected, such as altering settings for two or more reconstruction parameters for a given iteration, [0142] The action or actions are for any of the reconstruction parameters. For example, the gradient step size and/or relative contribution of different components (e.g., weights) are the reconstruction parameters corresponding to the action or actions. Alternatively or additionally, one or more reconstruction parameters for denoising due to the under sampled scan data are the reconstruction parameters corresponding to the action or actions. The shape of the threshold or other regularization parameter for controlling aliasing artifacts in the reconstruction are changed. The learned policy determines which of the changes and how much of the determined change to perform in each of the iterations).
Regarding Claim 8, Mailhe in view of Sreenivasan discloses the method of claim 1 wherein establishing comprises establishing where at least one of the multiple different parameters is constrained by hardware of the MR scanner and/or input by the user (Mailhe, [0134] In act 1404, the MR scanner or other MR system scans the patient with the MR compressed sequence. Based on the configuration of the MR scanner, a pulse sequence is created, [0149] The user configures the MR scanner 100 to scan the patient, such as selecting one or more settings of scan parameters. The operator interface is used to set a scan parameter), the optimization establishing values of the multiple different parameters as constrained (Mailhe, [0141] The actions 1108 are determined based on the learned Markov decision process. The current image and meta data are used to determine a probability distribution of settings of a reconstruction parameter. The distribution provides the settings with greater and lesser probabilities of being rewarded to provide optimized reconstruction in a final MR image. The action is determined by random or other sampling of the probability distribution. Probability distributions for more than one reconstruction parameter may be used. The sampling may select which type of action as well as the setting for the action to use. More than one action may be selected, such as altering settings for two or more reconstruction parameters for a given iteration).
Regarding Claim 9, Mailhe discloses the method of claim 1 wherein establishing comprises establishing by the reinforcement learned artificial intelligence where the optimization considers an end goal of the imaging (Mailhe, [0144] In act 1408, the image processor outputs a final reconstructed MR image. After the stop criterion shows sufficient or best fit of the reconstructed image to the under sampled scan data, the iterations cease. The last MR image is the optimized MR image)
Mailhe does not disclose the MR pulse sequence optimized for the end goal.
However, Sreenivasan teaches the MR pulse sequence optimized for the end goal (Sreenivasan, [Page 4] Step a) is to design a “optimal” pulse sequence for each of the first set of MRI objects SO; For example, the set SO may include N .sub.S0 = 50 people. The pulse sequence is considered to be optimal for the subject when minimizing or maximizing the cost function, respectively, representing the homogeneity or heterogeneity of the magnetized flip angle within the region of the subject's body obtained by reproducing the sequence of subject).
Before the effective filing date of the claimed invention, It would have been obvious to one of ordinary skill in the art to combine Mailhe and Sreenivasan’s teaching because Mailhe teaches optimization of MR imaging using a reinforcement learning framework that evaluates reconstructed images based on a criterion or quality measure but does not explicitly disclose optimizing the MR pulse sequence itself with respect to a defined imaging end goal. Sreenivasan teaches designing MR pulse sequences that are optimized according to a cost function or objective. A person of ordinary skill in the art would have been motivated to combine Sreenivasan’s teaching of optimizing MR pulse sequences based on an end goal into the system of Mailhe in order to improve the overall imaging performance by aligning the pulse sequence design with the desired imaging outcome.
Regarding Claim 10, Mailhe in view of Sreenivasan discloses the method of claim 9 wherein establishing comprises considering of the end goal as a type of map, segmentation, and/or detection (Mailhe, [0033] This disclosure describes examples based on machine learning. The processing is split into two phases: an offline training phase where the goal is to identify an optimal set of reconstruction parameters that can be applied to many different images, and an online processing phase in which new MR data are scanned and the goal is to reconstruct the images using the reconstruction parameters learned during the training phase, [0144] In act 1408, the image processor outputs a final reconstructed MR image. After the stop criterion shows sufficient or best fit of the reconstructed image to the under sampled scan data, the iterations cease. The last MR image is the optimized MR image, [0145] The output is to a display plane or buffer. Color mapping or other post reconstruction processing is used to generate the MR image. The display device reads from the display plane or buffer to display the image to the operator. In alternative or additional embodiments, the MR image is transferred over a network to other displays, a patient medical record, or memory).
Regarding Claim 11, Mailhe in view of Sreenivasan discloses the method of claim 1 wherein establishing comprises applying an agent of the reinforcement learned artificial intelligence in the optimization (Mailhe, [0140] The trained recurrent reconstruction engine 1104 is configured by the training to respond to actions 1108 from the critic 1106. For each iteration of the reconstruction, one or more actions representing changes to be made in reconstruction parameters are received. The deep reinforcement machine-learnt network is trained to control the sequence of actions through the iterations based on the learned policy. The policy uses the current image and meta data to determine a change to be made in one or more settings of parameters of reconstruction) where the agent selects a sequence of actions with respect to settings for different objects parameterizing the MR sequence (Mailhe, [0142] The action or actions are for any of the reconstruction parameters. For example, the gradient step size and/or relative contribution of different components (e.g., weights) are the reconstruction parameters corresponding to the action or actions. Alternatively or additionally, one or more reconstruction parameters for denoising due to the under sampled scan data are the reconstruction parameters corresponding to the action or actions. The shape of the threshold or other regularization parameter for controlling aliasing artifacts in the reconstruction are changed. The learned policy determines which of the changes and how much of the determined change to perform in each of the iterations).
Regarding Claim 12, Mailhe in view of Sreenivasan discloses the method of claim 1 wherein establishing comprises establishing over a sequence of different resolutions in actions steps from coarser to finer (Mailhe, [0059] To efficiently combine the images reconstructed from patches of different sizes, the patch based processing scheme is combined into a multiscale subsampled representation (such as with an orthogonal wavelet transform) which otherwise achieves perfect reconstruction. Effectively, this places more emphasis on reconstructing higher frequency components (such as sharp edges and fine patterns) with patches of smaller sizes, while low pass filtering in the down-sampled branch suppresses aliasing effects).
Pertinent Prior Art
The prior art made of record and not relied upon is considered pertinent to applicant’s disclose:
-US 20240168118 A1, describing systems and methods for dynamically adapting a sampling pattern during non-invasive measurements, including magnetic resonance (MR) imaging, where sampling patterns are modified based on acquired measurement data to improve image reconstruction.
-US 20250022599 A1, describing systems and methods for selecting and optimizing imaging protocols in a diagnostic imaging system, including magnetic resonance (MR) imaging, where imaging parameters and protocols are determined based on selected diagnostic tasks to improve imaging and analysis performance.
-US 11137462 B2, describing systems and methods for magnetic resonance (MR) imaging using machine learning to process imaging data, including quantifying and identifying features within a subject based on MR image data.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to IBRAHIM NAGI SHOHATEE whose telephone number is (571)272-6612. The examiner can normally be reached 8am-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Shelby Turner can be reached at (571) 272-6334. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/IBRAHIM NAGI SHOHATEE/Examiner, Art Unit 2857
/SHELBY A TURNER/Supervisory Patent Examiner, Art Unit 2857
Prosecution Insights