DETAILED ACTION
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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 06/05/2026 has been entered.
Response to Amendment
This office action is in response to the communications filed on 06/05/2026, concerning Application No. 17/689,032. The amendments to the claims filed on 06/05/2026 are acknowledged. Presently, claims 1-13 remain pending.
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d). The certified copy has been filed in parent Application No. KR10-2021-0031884, filed on 03/11/2021.
Claim Objections
Claims 1, 9, 11, and 13 are objected to because of the following informalities:
Claim 1, line 24, the limitation “the patient” should be changed to “a patient”;
Claim 9 is dependent from claim 1, and therefore, any corresponding limitation from claim 1 that is repeated in claim 9 should be amended to align with correct antecedent basis (i.e., changing “a” to “the” for respective limitations in claim 9);
Claim 11, line 18, the limitation “the patient” should be changed to “a patient”; and
Claim 13 is dependent from claim 11, and therefore, any corresponding limitation from claim 11 that is repeated in claim 13 should be amended to align with correct antecedent basis (i.e., changing “a” to “the” for respective limitations in claim 13).
Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 1-13 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
Claims 1 and 11 each recite the limitation “wherein the artificial neural network model is trained such that the maximum acoustic pressure calculated by applying the output CT image exhibits an error ratio of 4% or less on average across the M1, the V1, and the dACC, in comparison to a maximum acoustic pressure calculated by applying a corresponding actual CT image” (and similarly recited in dependent claims 9 and 13). However, looking at Applicant’s filed specification, it is not clear that this limitation directed to the AI model being trained such that the maximum acoustic pressure exhibits the claimed error rate ratio, etc., is actual step that is performed, or simply describing a statistic they calculated to show that their synthesized CT image is more accurate for ultrasound treatment. See paragraphs [0059]-[0060] of Applicant’s PGPub and Applicant’s Figure 7, wherein Applicant discusses comparing the generated CT image (sCT) and the real CT image (rCT), wherein Table 1 sets forth the error ratio from the comparison of sCT and rCT, and wherein paragraph [0060] sets forth “As shown in the above table, the maximum acoustic pressure exhibits an error ratio of 4% or less on average all over the targets. This signifies that the synthesized CT image rather than the real CT image is suitable for use in ultrasound treatment”.
The limitation of “wherein the artificial neural network model is trained such that the maximum acoustic pressure calculated by applying the output CT image exhibits an error ratio of 4% or less on average across the M1, the V1, and the dACC, in comparison to a maximum acoustic pressure calculated by applying a corresponding actual CT image” does not appear to be supported by the filed specification. Paragraphs [0059]-[0060] of Applicant’s PG-Pub appears to be the only place for support of the calculation of the maximum acoustic pressure, but it is not clear that it is the training of the AI model that is linked to the claimed error ratio. In other words, the specification does not appear to provide support for the training of the AI model to be specifically performed in order to achieve the claimed error ratio.
Therefore, this limitation for the claimed AI neural network model training being performed to provide the claimed error ratio fails to comply with the written description requirement of 35 USC 112(a), because the Applicant’s specification seems to only support that an error ratio as claimed was calculated to show that synthesized CT image provides more accurate results, but does not link training the artificial neural network model in order to provide the claimed error ratio. This is also true for the apparatus claims (dependent claims 9 and 13) as there is no mention that the artificial neural network is specifically “configured” to provide the claimed error ratio. Clarification is required.
Claims 2-10 and 12-13 are also rejected under 35 U.S.C. 112(a) due to the dependency on claims 1 and 11, respectively.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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 1-2, 6-8, and 11-12 are rejected under 35 U.S.C. 103 as being unpatentable over Lei et al. (NPL: “MRI-only based synthetic CT generation using dense cycle consistent generative adversarial networks.” American Association of Physicists in Medicine, Med. Phys. 46 (8), Aug. 2019, Retrieved from: https://pmc.ncbi.nlm.nih.gov/articles/PMC6692192/pdf/MP-46-3565.pdf; of record, hereinafter Lei) in view of Hynynen et al. (US 2018/0177491 A1, of record, hereinafter Hynynen), further in view of Zhang et al. (NPL: “Transcranial Focused Ultrasound Neuromodulation: A Review of the Excitatory and Inhibitory Effects on Brain Activity in Human and Animals.” frontiers in Human Neuroscience, vol. 15, 28 Sept. 2021, Retrieved from: https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2021.749162/full; of record, hereinafter Zhang), and even further in view of Liu et al. (NPL: “Synthetic CT Skull Generation for Transcranial MR Imaging–Guided Focused Ultrasound Interventions with Conditional Adversarial Networks.” arXiv:2202.10136v2, 22 Feb. 2022; of record, hereinafter Liu).
Regarding claims 1 and 11-12, Lei discloses a method for converting magnetic resonance imaging (MRI) to a computed tomography (CT) image based on artificial intelligence, performed by a processor, the method comprising: acquiring training data including an MRI image and an original CT image for machine learning; carrying out preprocessing of the training data; training an artificial neural network model using the training data, wherein the artificial neural network model generates a generated CT image corresponding to the MRI image, and compares the generated CT image with the original CT image included in the training data; receiving an input MRI image to be converted to a CT image; splitting the input MRI image into a plurality of three-dimensional (3D) patches, each having a predetermined same size; generating patches of a converted CT image corresponding to the plurality of 3D patches of the input MRI image by applying the plurality of 3D patches, as an input, to the trained artificial neural network model; and merging the patches of the converted CT image sequentially to generate an output CT image (see, e.g., Lei: Flow chart of Figure 1 on Page 3567, and Sections 1 and 2 on Pages 3565-3570, referring to the deep learning-based method consisting of a training stage and a synthesizing stage being used to generate patient specific synthetic CT (sCT) from routine MRI (i.e., original MRI), the “Patch extraction” of the MRI images, and patch fusion/merging to reconstruct the “Final sCT”; also see, e.g., Lei: Section 2F on Pages 3569-3570, referring to the patch size in the GAN-based method being set to 32x32x32, such that there is a predetermined patch size).
Lei does not specifically disclose [1] acquiring skull factor information and parameter information related to the ultrasound treatment of a target tissue based on the output CT image, [2] wherein the ultrasound treatment is performed using the acoustic property information acquired from the output CT image generated from the input MRI image, without the addition of any CT scan on the patient, [3] wherein the skull factor information is acquired based on at least one of a thickness, a location, and a shape of a skull of the patient, [4] wherein the skull factor information is used to acquire acoustic property information essential for the ultrasound treatment comprising a speed of sound traveling through the target tissue, a density of the target tissue, and an attenuation coefficient at the target tissue, [5] wherein the acoustic property information is used for transcranial ultrasound treatment of brain regions including at least one of primary motor cortex (M1), primary visual cortex (V1), and dorsal anterior cingulate cortex (dACC), [6] wherein the acoustic property information comprises at least one of a maximum acoustic pressure, a focal dice coefficient, and a focal distance error that are calculated differently for each target position including at least one of the M1, the V1, and the dACC, [7] wherein the acoustic property information is calculated for each of the target position from the output CT image generated by sequentially merging the converted CT patches, and [8] wherein the artificial neural network model is trained such that the maximum acoustic pressure calculated by applying the output CT image exhibits an error ratio of 4% or less on average across the M1, the V1, and the dACC, in comparison to a maximum acoustic pressure calculated by applying a corresponding actual CT image.
However, in the same field of endeavor of transcranial ultrasound imaging/treatment, Hynynen discloses acquiring skull factor information and parameter information related to the ultrasound treatment of a target tissue based on the output CT image, wherein the ultrasound treatment is performed using the acoustic property information acquired from the output CT image generated from the input MRI image, without the addition of any CT scan on the patient, wherein the skull factor information is acquired based on at least one of a thickness, a location, and a shape of a skull of the patient, wherein the skull factor information is used to acquire acoustic property information essential for the ultrasound treatment comprising a speed of sound traveling through the target tissue, a density of the target tissue, and an attenuation coefficient at the target tissue, wherein the acoustic property information is used for transcranial ultrasound treatment of brain regions (see, e.g., Para. [0060], “As shown in FIG. 1A, each of the transcranial ultrasound transducer array elements 100 may be oriented such that all of their far field regions spatially overlap within a least a portion of the brain (shown as focusing range or focusing region 50), thereby allowing for far-field focusing within this region via control of timing of the emission of ultrasound energy from the transcranial ultrasound transducer array elements 100 (e.g. operating the transcranial ultrasound transducer array as a phased array). In some example embodiments, the focusing region 50 may lie within a portion of the brain that is known to contain a target for therapy or imaging”, and Para. [0079-0083], and Para. [0097], and Para. [0104], “A CT scan (LightSpeed VCT, GE Healthcare, Chalfont St Giles, UK) of a human head can be obtained and used in each of the numerical simulations. The CT dataset (512×512×328 voxels with uniform voxels of size 625×625×625 mm.sup.3) can be used to extract density and morphology information. The density is obtained using a linear relation with the Hounsfield Units, using knowledge of the densities of brain tissue and air in the CT scan. The skull CT data is then segmented and interpolated such that the discretization in the numerical simulations is λ=10, where λ is the wavelength of the ultrasound in water” (emphasis added), where the claimed skull factor information corresponds to the disclosed “skull CT data” obtained from one CT image dataset, in which the disclosed skull CT data is used to extract density and morphology information, and where the one CT image dataset is disclosed to be from one initial CT scan and used in each of the numerical simulations / used to extract the desired information, and Para. [0105-0106], “A hybrid numerical model is employed to perform simulations involving the propagation of ultrasound bursts of variable length emitted from the transcranial ultrasound transducer array elements. The numerical method combines finite difference simulations with the grid method. This hybrid model calculates the pressure field in the brain and the particle displacement field in the cranial bone using finite difference methods, while coupling these different equations on the boundary using the grid method. […] where p denotes the acoustic pressure, α.sub.L is the longitudinal attenuation coefficient, c is the speed of sound, and ρ is the density”, where acoustic pressure, attenuation coefficient, speed of sound, and density can be calculated for simulations involving the propagation of ultrasound bursts from the transcranial ultrasound transducer array elements by using a numerical method, in which the initial CT scan of a human head / skull CT data can be obtained and used in each of the numerical simulations, see Para. [0104], as set forth above),
wherein the acoustic property information comprises at least one of a maximum acoustic pressure, a focal dice coefficient, and a focal distance error that are calculated differently for each target position including at least one of the brain regions (see, e.g., Para. [0101], “As shown in the examples provided below, the present example embodiments may be employed to generate high acoustic pressures at off-center targets, using fewer ultrasound elements than conventional transcranial ultrasound array devices”, and Para. [0105-0106], “A hybrid numerical model is employed to perform simulations involving the propagation of ultrasound bursts of variable length emitted from the transcranial ultrasound transducer array elements. The numerical method combines finite difference simulations with the grid method. This hybrid model calculates the pressure field in the brain and the particle displacement field in the cranial bone using finite difference methods, while coupling these different equations on the boundary using the grid method. […] where p denotes the acoustic pressure, α.sub.L is the longitudinal attenuation coefficient, c is the speed of sound, and ρ is the density”, and Figs. 12D-12F, with Para. [0129], “FIGS. 12D-F demonstrate the dependence of the peak acoustic pressure at each target location on the number of array elements”, where peak/maximum acoustic pressure at each target location is calculated), and
wherein the acoustic property information is calculated for each of the target position from the output CT image generated by sequentially merging the converted CT patches (see, e.g., Para. [0101], “As shown in the examples provided below, the present example embodiments may be employed to generate high acoustic pressures at off-center targets, using fewer ultrasound elements than conventional transcranial ultrasound array devices”, and Para. [0104], “A CT scan (LightSpeed VCT, GE Healthcare, Chalfont St Giles, UK) of a human head can be obtained and used in each of the numerical simulations. The CT dataset (512×512×328 voxels with uniform voxels of size 625×625×625 mm.sup.3) can be used to extract density and morphology information. […] The skull CT data is then segmented and interpolated”, where the claimed skull factor information corresponds to the disclosed “skull CT data” obtained from a CT image dataset, in which the disclosed skull CT data is used to extract density and morphology information, and Para. [0105-0106], “A hybrid numerical model is employed to perform simulations involving the propagation of ultrasound bursts of variable length emitted from the transcranial ultrasound transducer array elements. […] This hybrid model calculates the pressure field in the brain and the particle displacement field in the cranial bone using finite difference methods […] where p denotes the acoustic pressure, α.sub.L is the longitudinal attenuation coefficient, c is the speed of sound, and ρ is the density”, where acoustic pressure, attenuation coefficient, speed of sound, and density can be calculated for simulations involving the propagation of ultrasound bursts from the transcranial ultrasound transducer array elements by using a numerical method, in which a CT scan of a human head / skull CT data can be obtained and used in each of the numerical simulations, see Para. [0104], as set forth above, and Figs. 12D-12F, with Para. [0129], where peak/maximum acoustic pressure at each target location is calculated).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Lei by including [1] acquiring skull factor information and parameter information related to the ultrasound treatment of a target tissue based on the output CT image, [2] wherein the ultrasound treatment is performed using the acoustic property information acquired from the output CT image generated from the input MRI image, without the addition of any CT scan on the patient, [3] wherein the skull factor information is acquired based on at least one of a thickness, a location, and a shape of a skull of the patient, [4] wherein the skull factor information is used to acquire acoustic property information essential for the ultrasound treatment comprising a speed of sound traveling through the target tissue, a density of the target tissue, and an attenuation coefficient at the target tissue, [5] wherein the acoustic property information is used for transcranial ultrasound treatment of brain regions, [6] wherein the acoustic property information comprises at least one of a maximum acoustic pressure, a focal dice coefficient, and a focal distance error that are calculated differently for each target position including at least one of the brain regions, and [7] wherein the acoustic property information is calculated for each of the target position from the output CT image generated by sequentially merging the converted CT patches, as disclosed by Hynynen. One of ordinary skill in the art would have been motivated to make this modification in order to allow for improved correction for the bone induced effects on the wave propagation, as recognized by Hynynen (see, e.g., Para. [0058], [0074-0082], and [0104-0106]).
Lei modified by Hynynen still does not specifically disclose [5] wherein the acoustic property information is used for transcranial ultrasound treatment of brain regions specifically including at least one of primary motor cortex (M1), primary visual cortex (V1), and dorsal anterior cingulate cortex (dACC), and [8] wherein the artificial neural network model is trained such that the maximum acoustic pressure calculated by applying the output CT image exhibits an error ratio of 4% or less on average across the M1, the V1, and the dACC, in comparison to a maximum acoustic pressure calculated by applying a corresponding actual CT image.
However, in the same field of endeavor of transcranial ultrasound imaging/treatment, Zhang discloses wherein the acoustic property information is used for transcranial ultrasound treatment of brain regions including at least one of primary motor cortex (M1), primary visual cortex (V1), and dorsal anterior cingulate cortex (dACC) (see, e.g., Zhang: Page 7, Col. 2, lines 31-51 to Page 8, Col. 1, lines 1-2, “The excitatory effects of tFUS [transcranial focused ultrasound] were identified in healthy human subjects in 9 studies as indicated by fMRI and EEG. Six studies showed the inhibitory effects of TUS indicated by SEPs, motor induced potentials (MEPs), and intracortical facilitation (ICF). The modulated brain targets included the motor cortex, somatosensory cortex, thalamus, caudate nuclei, and visual cortex. […] Three studies conducted by Lee et al. (2015) showed excitatory effects of tFUS on the human primary and secondary somatosensory cortex and visual cortex. […] FUS targeting to the human visual cortex (V1) induced the perception of phosphine and activated a network of regions involved in visual and higher-order cognitive processes”, and Table 2 on Pages 5-6).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have further modified the method of Lei modified by Hynynen by including [5] wherein the acoustic property information is used for transcranial ultrasound treatment of brain regions specifically including at least one of primary motor cortex (M1), primary visual cortex (V1), and dorsal anterior cingulate cortex (dACC), as disclosed by Zhang. One of ordinary skill in the art would have been motivated to make this modification in order to identify the excitatory and/or inhibitory effects of transcranial focused ultrasound on desired brain targets, as recognized by Zhang (see, e.g., Zhang: Page 7, Col. 2, lines 31-51 to Page 8, Col. 1, lines 1-2, and Table 2 on Pages 5-6).
Lei modified by Hynynen and Zhang still does not specifically disclose [8] wherein the artificial neural network model is trained such that the maximum acoustic pressure calculated by applying the output CT image exhibits an error ratio of 4% or less on average across the M1, the V1, and the dACC, in comparison to a maximum acoustic pressure calculated by applying a corresponding actual CT image.
However, in the same field of endeavor of transcranial ultrasound imaging/treatment, Liu discloses wherein the artificial neural network model is trained such that the maximum acoustic pressure calculated by applying the output CT image exhibits an error ratio of 4% or less on average across the M1, the V1, and the dACC, in comparison to a maximum acoustic pressure calculated by applying a corresponding actual CT image (see, e.g., Fig. 4, and Page 6, Section 4. Discussion and Conclusions, lines 1-13, “In this work, we explored the feasibility of replacing real CT with synthetic CT images for TcMRgFUS planning. We showed that the synthesized CT skull generated by our proposed 3D patch-based cGAN are (1) spatially continuous and (2) highly comparable to real CTs when used with a TcMRgFUS simulator. Through Kranion, skull metric and element comparisons were shown to have strong correlation between rCT and sCT. Acoustic simulations revealed slight discrepancies between rCT and sCT in maximum pressure at the focus and focal location. […] Characterizing the sCTs using acoustic simulations provided further information about the clinical potential of sCTs. Koh et al. showed smaller percent error of maximum pressure at the focus (23.5±6.51% vs. 3.11±2.79%)” (emphasis added), where Liu does seem to be comparing real CTs to synthetic CTs generated using AI neural network models and calculating an error ratio, wherein it seems synthetic CTs may provide a small percent error of maximum pressure).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have further modified the method of Lei modified by Hynynen and Zhang by including [8] wherein the artificial neural network model is trained such that the maximum acoustic pressure calculated by applying the output CT image exhibits an error ratio of 4% or less on average across the M1, the V1, and the dACC, in comparison to a maximum acoustic pressure calculated by applying a corresponding actual CT image, as disclosed by Liu. One of ordinary skill in the art would have been motivated to make this modification in order to provide a smaller percent error of maximum pressure at the focus, as recognized by Liu (see, e.g., Page 6, Section 4. Discussion and Conclusions, lines 1-13).
Regarding claim 2, Lei modified by Hynynen, Zhang, and Liu discloses the method according to claim 1, as set forth above. Lei further discloses wherein the training the artificial neural network model comprises: a first process of generating a plurality of patches of the generated CT image corresponding to a plurality of patches of the MRI image included in the training data using a generator (see, e.g., Lei: Flow chart of Figure 1 on Page 3567, and Sections 1 and 2 on Pages 3565-3570, referring to the deep learning-based method consisting of a training stage and a synthesizing stage being used to generate patient specific synthetic CT (sCT) from routine MRI (i.e., original MRI), the “Patch extraction” of the MRI images, and patch fusion/merging to reconstruct the “Final sCT”); a second process of acquiring error data by comparing the plurality of patches of the generated CT image with a plurality of patches of the original CT image included in the training data using a discriminator; and a third process of training the generator using the error data (see, e.g., Lei: Page 3566, Col. 2, lines 10-17, “Generative adversarial networks (GAN) have been used in the generation of sCT by introducing an additional discriminator to distinguish the sCT from real CT, improving the final sCT imaging qualities in comparison to the previously deep learning-based methods.35 These models incorporate an adversarial loss term in addition to the conventional synthesis error, with the objective of producing more realistic CT data”).
Regarding claim 6, Lei modified by Hynynen, Zhang, and Liu discloses the method according to claim 1, as set forth above. Lei further discloses wherein the artificial neural network model generates the generated CT image corresponding to the MRI image through trained nonlinear mapping (see, e.g., Lei: Page 3566, Col. 1, lines 48-51, “Since deep learning-based methods can provide a more complex nonlinear mapping from input to output image through a multilayer and fully trainable model, these methods are becoming popular for the task of image synthesis”, and Page 3570, Col. 1, lines 4-6, “GAN-based methods incorporate an adversarial learning strategy into an end-to-end FCN to train nonlinear mapping from MRI patches to CT image patches”).
Regarding claim 7, Lei modified by Hynynen, Zhang, and Liu discloses the method according to claim 1, as set forth above. Lei further discloses wherein the carrying out preprocessing of the training data comprises: removing an unnecessary area for training by applying a mask to a region of interest in the MRI image and the original CT image included in the training data (see, e.g., Lei: Page 3575, Col. 1, lines 21-25, “We evaluated the surface distance of body outline and bone for all test brain and pelvic data. For body outline, we first used threshold 500 HU to get the binary mask of CT and sCT images, and then we used dilate, fill holes and erode operation to get the CT and sCT body outlines”, where a binary mask is applied to define the region of interest in the images).
Regarding claim 8, Lei modified by Hynynen, Zhang, and Liu discloses the method according to claim 1, as set forth above. Lei further discloses a computer program stored in a computer-readable recording medium, for performing the method for converting magnetic resonance imaging (MRI) to a computed tomography (CT) image based on artificial intelligence according to claim 1 (see, e.g., Lei: Page 3568, Col. 1, lines 38-40, “all training MRI and CT images were first rigidly registered by an intra-subject registration using a commercial software Velocity AI”; also see, e.g., Lei: Flow chart of Figure 1 on Page 3567, and Sections 1 and 2 on Pages 3565-3570, and the “Purpose” and “Methods and materials” sections in the Abstract, referring to the deep learning-based method for generation of sCT based on dense cycle-consistent generative adversarial networks (cycle GAN), in which the cycle GAN-based model was developed to generate sCT images in a patch-based framework).
Claims 3-5 are rejected under 35 U.S.C. 103 as being unpatentable over Lei (NPL) in view of Hynynen (US 2018/0177491 A1), Zhang (NPL), and Liu (NPL), as applied to claims 1-2 above, and further in view of Cui et al. (US 2022/0092787 A1, of record, hereinafter Cui).
Regarding claim 3, Lei modified by Hynynen, Zhang, and Liu discloses the method according to claim 2, as set forth above. Lei modified by Hynynen, Zhang, and Liu does not specifically disclose wherein the artificial neural network model is trained to reduce differences between the original CT image and the generated CT image by iteratively performing the first to third processes.
However, in the same field of endeavor of using neural network models in medical imaging systems, Cui discloses wherein the artificial neural network model is trained to reduce differences between the original image and the generated image by iteratively performing the first to third processes (see, e.g., Para. [0150], “The model training process may be a process for iteratively updating the model parameters. In some embodiments, a loss function of the preliminary metal detection model may include a multi-category comparison loss function of the improved neural network model for category semantic perception. In some embodiments, the loss function may be used to determine an error between an output of the trained metal detection model and the corresponding reference binary image. In some embodiments, the loss function may be used to determine an error backpropagation method, thereby driving an iterative training process”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have further modified the method of Lei modified by Hynynen, Zhang, and Liu by including wherein the artificial neural network model is trained to reduce differences between the original image and the generated image by iteratively performing the first to third processes, as disclosed by Cui. One of ordinary skill in the art would have been motivated to make this modification in order to improve the neural network model, as recognized by Cui (see, e.g., Para. [0150]).
Regarding claim 4, Lei modified by Hynynen, Zhang, and Liu discloses the method according to claim 2, as set forth above. Lei modified by Hynynen, Zhang, and Liu does not specifically disclose wherein the generator includes: at least one convolutional layer for receiving input MRI image data and outputting a feature map which emphasizes features of a region of interest; and at least one transposed convolutional layer for generating the generated CT image corresponding to the input MRI image based on the feature map.
However, in the same field of endeavor of using neural network models in medical imaging systems, Cui discloses wherein the generator includes: at least one convolutional layer for receiving input image data and outputting a feature map which emphasizes features of a region of interest (see, e.g., Para. [0138-0139], “FIG. 16 is a schematic diagram illustrating an exemplary structure of an improved neural network model for category semantic perception according to some embodiments of the present disclosure. As shown in FIG. 16, the improved neural network model for category semantic perception may include an initial feature extraction structure 1610, a fusion feature extraction structure 1620, and a feature fusion structure 1630… the initial feature extraction structure 1610 may be configured to initially extract edge features of the initial metal region from the X-ray image 1611. The initial feature extraction structure 1610 may include a first convolution layer 1612, at least three transition layers 1613, a plurality of residual blocks 1614, and a second convolution layer 1615”); and
at least one transposed convolutional layer for generating the generated image corresponding to the input image based on the feature map (see, e.g., Para. [0140-0141], “the fusion feature extraction structure 1620 may be used to perform a secondary feature extraction on the first initial feature generated by the residual block 1614 to further extract edge features of the initial metal region. The fusion feature extraction structure 1620 may include at least one up-sampling layer 1621. In some embodiments, each of the at least one up-sampling layer 1621 may include at least one transposed convolution 1622. Each of the up-sampling layer 1621 may connect to a residual block 1614 of the plurality of residual blocks 1614. In some embodiments, the up-sampling layer 1621 may be used to generate a fusion feature by performing a secondary feature extraction on the first initial feature generated by the residual block 1614… the feature fusion structure 1630 may connect to the second convolution layer 1615 and each up-sampling layer 1621. The feature fusion structure 1630 may be used to generate an initial binary image 1631 by fusing the second initial feature generated by the second convolution layer 1615 and the fusion feature generated by each of the at least one up-sampling layer 1621”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have further modified the method of Lei modified by Hynynen, Zhang, and Liu by including wherein the generator includes: at least one convolutional layer for receiving input image data and outputting a feature map which emphasizes features of a region of interest; and at least one transposed convolutional layer for generating the generated image corresponding to the input MRI image based on the feature map, as disclosed by Cui. One of ordinary skill in the art would have been motivated to make this modification in order to improve the neural network model, as recognized by Cui (see, e.g., Para. [0150]).
Regarding claim 5, Lei modified by Hynynen, Zhang, and Liu discloses the method according to claim 2, as set forth above. Lei modified by Hynynen, Zhang, and Liu does not specifically disclose wherein the discriminator includes at least one convolutional layer for receiving an input CT image data generated by the generator and outputting a feature map which emphasizes features of a region of interest.
However, in the same field of endeavor of using neural network models in medical imaging systems, Cui discloses wherein the discriminator includes at least one convolutional layer for receiving an input image data generated by the generator and outputting a feature map which emphasizes features of a region of interest (see, e.g., Para. [0138-0139], “FIG. 16 is a schematic diagram illustrating an exemplary structure of an improved neural network model for category semantic perception according to some embodiments of the present disclosure. As shown in FIG. 16, the improved neural network model for category semantic perception may include an initial feature extraction structure 1610, a fusion feature extraction structure 1620, and a feature fusion structure 1630… the initial feature extraction structure 1610 may be configured to initially extract edge features of the initial metal region from the X-ray image 1611. The initial feature extraction structure 1610 may include a first convolution layer 1612, at least three transition layers 1613, a plurality of residual blocks 1614, and a second convolution layer 1615”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have further modified the method of Lei modified by Hynynen, Zhang, and Liu by including wherein the discriminator includes at least one convolutional layer for receiving an input CT image data generated by the generator and outputting a feature map which emphasizes features of a region of interest, as disclosed by Cui. One of ordinary skill in the art would have been motivated to make this modification in order to improve the neural network model, as recognized by Cui (see, e.g., Para. [0150]).
Claims 9-10 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Lei (NPL) in view of Hynynen (US 2018/0177491 A1), Zhang (NPL), and Liu (NPL), as applied to claims 1 and 11 above, and further in view of Barnes et al. (US 2013/0158385 A1, of record, hereinafter Barnes).
Regarding claims 9 and 13, Lei modified by Hynynen, Zhang, and Liu discloses the method according to claim 1 and the method according to claim 11, respectively, as set forth above. As set forth above in the rejection of independent claims 1 and 11, Lei discloses using the method for converting MRI to a CT image based on artificial intelligence according to claims 1 and 11 (see, e.g., Lei: Flow chart of Figure 1 on Page 3567, and Sections 1 and 2 on Pages 3565-3570, referring to the deep learning-based method consisting of a training stage and a synthesizing stage being used to generate patient specific synthetic CT (sCT) from routine MRI (i.e., original MRI), the “Patch extraction” of the MRI images, and patch fusion/merging to reconstruct the “Final sCT”; also see, e.g., Lei: Section 2F on Pages 3569-3570, referring to the patch size in the GAN-based method being set to 32x32x32, such that there is a predetermined patch size); Hynynen discloses acquiring (via a processing unit) the skull factor information and the parameter information related to the ultrasound treatment of a target tissue based on the output CT image, wherein the ultrasound treatment is performed using the acoustic property information acquired from the output CT image generated from the input MRI image, without the addition of any CT scan on the patient, wherein the skull factor information is acquired based on at least one of a thickness, a location, and a shape of a skull of the patient, wherein the skull factor information is used to acquire the acoustic property information essential for the ultrasound treatment comprising the speed of sound traveling through the target tissue, the density of the target tissue, and the attenuation coefficient at the target tissue, wherein the acoustic property information is used for transcranial ultrasound treatment of the brain regions, wherein the acoustic property information comprises at least one of a maximum acoustic pressure, a focal dice coefficient, and a focal distance error that are calculated differently for each target position including at least one of the brain regions, and wherein the acoustic property information is calculated for each of the target position from the output CT image generated by sequentially merging the converted CT patches (see, e.g., Para. [0060], [0079-0083], and [0097], and Para. [0104], where the claimed skull factor information corresponds to the disclosed “skull CT data” obtained from a CT image dataset, in which the disclosed skull CT data is used to extract density and morphology information, and Para. [0105-0106], where acoustic pressure, attenuation coefficient, speed of sound, and density can be calculated for simulations involving the propagation of ultrasound bursts from the transcranial ultrasound transducer array elements by using a numerical method, in which a CT scan of a human head / skull CT data can be obtained and used in each of the numerical simulations, see Para. [0104], as set forth above, and further Para. [0101], and Figs. 12D-12F, with Para. [0129]); Zhang discloses wherein the acoustic property information is used for transcranial ultrasound treatment of brain regions including at least one of the M1, the V1, and the dACC (see, e.g., Zhang: Page 7, Col. 2, lines 31-51 to Page 8, Col. 1, lines 1-2, “The excitatory effects of tFUS [transcranial focused ultrasound] were identified in healthy human subjects in 9 studies as indicated by fMRI and EEG. Six studies showed the inhibitory effects of TUS indicated by SEPs, motor induced potentials (MEPs), and intracortical facilitation (ICF). The modulated brain targets included the motor cortex, somatosensory cortex, thalamus, caudate nuclei, and visual cortex. […] Three studies conducted by Lee et al. (2015) showed excitatory effects of tFUS on the human primary and secondary somatosensory cortex and visual cortex. […] FUS targeting to the human visual cortex (V1) induced the perception of phosphine and activated a network of regions involved in visual and higher-order cognitive processes”, and Table 2 on Pages 5-6); and Liu discloses wherein the artificial neural network model is trained such that the maximum acoustic pressure calculated by applying the output CT image exhibits an error ratio of 4% or less on average across the M1, the V1, and the dACC, in comparison to a maximum acoustic pressure calculated by applying a corresponding actual CT image (see, e.g., Fig. 4, and Page 6, Section 4. Discussion and Conclusions, lines 1-13, “In this work, we explored the feasibility of replacing real CT with synthetic CT images for TcMRgFUS planning. We showed that the synthesized CT skull generated by our proposed 3D patch-based cGAN are (1) spatially continuous and (2) highly comparable to real CTs when used with a TcMRgFUS simulator. Through Kranion, skull metric and element comparisons were shown to have strong correlation between rCT and sCT. Acoustic simulations revealed slight discrepancies between rCT and sCT in maximum pressure at the focus and focal location. […] Characterizing the sCTs using acoustic simulations provided further information about the clinical potential of sCTs. Koh et al. showed smaller percent error of maximum pressure at the focus (23.5±6.51% vs. 3.11±2.79%)” (emphasis added), where Liu does seem to be comparing real CTs to synthetic CTs generated using AI neural network models and calculating an error ratio, wherein it seems synthetic CTs may provide a small percent error of maximum pressure).
Lei modified by Hynynen, Zhang, and Liu still does not specifically disclose a magnetic resonance-guided ultrasound treatment device, comprising: a magnetic resonance imaging (MRI) image acquisition unit to acquire the MRI image of a patient; a display unit to display the target tissue for the ultrasound treatment on a display based on the MRI image; a computed tomography (CT) image generation unit to generate the CT image corresponding to the MRI image; and an ultrasound output unit to output the ultrasound set based on the factor information and the parameter information to the target tissue.
However, in the same field of endeavor of magnetic resonance imaging, Barnes discloses a magnetic resonance-guided ultrasound treatment device (system 10) (see, e.g., Abstract and Para. [0021]), comprising:
a magnetic resonance imaging (MRI) image acquisition unit to acquire an MRI image of a patient (see, e.g., Abstract and Para. [0021]);
a display unit (display 26) to display the target tissue for the ultrasound treatment on a display based on the MRI image (see, e.g., Para. [0024], “The display 26 is a monitor, LCD, projector, plasma display, CRT, printer, or other now known or later developed devise for outputting visual information. The display 26 receives images, graphics, or other information from the processor 24, memory 12, MR system 14, or ultrasound system 16. Where the ultrasound system 16 is for therapy only, the display 26 may be used for MR imaging and a user interface to guide steering of the therapy without displaying ultrasound images”);
the processing unit (processor 24) to specifically acquire factor information and parameter information related to the ultrasound treatment of the target tissue (see, e.g., Para. [0044], “The location for treatment and/or treatment parameters (e.g., frequency, duration, duty cycle, waveform, aperture, and/or amplitude) are determined by the processor 24”); and
an ultrasound output unit to output ultrasound set based on the factor information and the parameter information to the target tissue (see, e.g., Para. [0044], “The location for treatment and/or treatment parameters (e.g., frequency, duration, duty cycle, waveform, aperture, and/or amplitude) are determined by the processor 24. Corresponding control signals are provided to the ultrasound system 16. The processor 24 may generate the user interface for control of the ultrasound system 16”, and Para. [0046], “The processor 24 may be configured to control a mode of operation of the ultrasound system 16”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have further modified the method of Lei modified by Hynynen, Zhang, and Liu by including a magnetic resonance-guided ultrasound treatment device, comprising: a magnetic resonance imaging (MRI) image acquisition unit to acquire an MRI image of a patient; a display unit to display the target tissue for the ultrasound treatment on a display based on the MRI image; a computed tomography (CT) image generation unit to generate a CT image corresponding to the MRI image; and an ultrasound output unit to output ultrasound set based on the factor information and the parameter information to the target tissue, as disclosed by Barnes. One of ordinary skill in the art would have been motivated to make this modification in order to provide the desired system for therapeutic ultrasound for use with magnetic resonance imaging to guide steering of the therapy without displaying ultrasound images, as recognized by Barnes (see, e.g., Abstract and Para. [0024]).
Regarding claim 10, Lei modified by Hynynen, Zhang, Liu, and Barnes discloses the magnetic resonance-guided ultrasound treatment device according to claim 9, as set forth above. Lei modified by Hynynen, Zhang, and Liu does not specifically disclose wherein the ultrasound output unit is configured to output high-intensity focused ultrasound to thermally or mechanically remove the target tissue, or low-intensity focused ultrasound to stimulate the target tissue without damage.
However, in the same field of endeavor of magnetic resonance imaging, Barnes discloses wherein the ultrasound output unit is configured to output high-intensity focused ultrasound to thermally or mechanically remove the target tissue (see, e.g., Para. [0019], “A compact, highly integrated, high multi-dimensional element count high power focused ultrasound (FUS) system is contained within a patient applicator. The multi-dimensional FUS system, with collocated drive electronics and computational resources, results in a simplified MR compatible system. The FUS system may be used in other applications than MR, such as for ultrasound imaging or therapy”, and Para. [0066], “Any number of elements 54 may be used. In one embodiment, there are at least 1,600 elements. An efficient, high power, high channel count high intensity ultrasound array system may have more than 1,500 elements for providing up to 3.3 KVA power for system and all support functions and capable of producing greater than 150 acoustic watts of applied acoustic energy”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have further modified the method of Lei modified by Hynynen, Zhang, Liu, and Barnes by including wherein the ultrasound output unit is configured to output high-intensity focused ultrasound to thermally or mechanically remove the target tissue, as disclosed by Barnes. One of ordinary skill in the art would have been motivated to make this modification in order to provide the desired system for therapeutic ultrasound for use with magnetic resonance imaging to guide steering of the therapy without displaying ultrasound images, as recognized by Barnes (see, e.g., Abstract and Para. [0024]).
Response to Arguments
Applicant's arguments, see Remarks filed 06/05/2026, have been fully considered but they are not persuasive.
Applicant’s arguments with respect to independent claims 1 and 11 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument. Examiner emphasizes that independent claims 1 and 11 are now rejected under 35 U.S.C. 103 as being unpatentable over Lei et al. (NPL) in view of Hynynen et al. (US 2018/0177491 A1), further in view of Zhang et al. (NPL), and even further in view of Liu et al. (NPL: “Synthetic CT Skull Generation for Transcranial MR Imaging–Guided Focused Ultrasound Interventions with Conditional Adversarial Networks.” arXiv:2202.10136v2, 22 Feb. 2022; of record), as set forth above.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TAYLOR DEUTSCH whose telephone number is (571)272-0157. The examiner can normally be reached Monday-Friday 9am-5pm EST.
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, PASCAL BUI-PHO can be reached at (571)272-2714. 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.
/T.D./Examiner, Art Unit 3798
/PASCAL M BUI PHO/Supervisory Patent Examiner, Art Unit 3798
Prosecution Insights