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 .
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 18-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claims do not fall within at least one of the four categories of patent eligible subject matter because the computer program is not eligible patented subject matter. Claims 18-20 are directed to, a computer program. However, one of ordinary skilled in the art would broadly interpret, the claimed computer program to encompass ineligible transitory signal embodiments as "signal" per se and software per se, which does not fall within one of the statutory categories of the invention (i.e. process, machine, manufacture or a composition of matter) and broadly interpret the computer readable medium to typically cover forms of non-transitory tangible media and transitory propagating signals per se in view of the ordinary and customary meaning of computer readable media. Therefore, a transitory computer readable medium would reasonably be interpreted by one of ordinary skill in the art as signal, per se. Thus, subject matter "a computer program" is not limited to that which falls within a statutory category of invention. Therefore, the claim contains ineligible non-statutory subject matter. “ Products that do not have a physical or tangible form, such as information (often referred to as "data per se") or a computer program per se (often referred to as "software per se") when claimed as a product without any structural recitations see MPEP 2106.03.
Claim Objections
Claims 5-9, and 12-16 are objected to because of the following informalities: The limitation PET is not defined at least one time in claims referenced above with positron emission tomography (PET). Appropriate correction is required.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1-3, 10-11 and 20 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Lais (US 2024/0144482).
As per claims 1 and 20, Lais teaches, a medical image registration method and a computer apparatus (Lais, ¶[002] “The computer-implemented method further includes performing a first registration of the normalized displaced image relative to the normalized reference image.” This represents a medical image registration method, see fig.1B for medical images), comprising: obtaining a similarity weight distribution containing a similarity weight of each element (Lais, ¶[0073] “Hence, the alignment module 160 can join the N.sub.T tile deformation vector fields by determining, for each pixel within the defined region in common, a weighted average of the tile deformation vector fields overlapping at that pixel, and assigning, for each pixel within the defined region in common, the weighted average to the deformation vector field.” This represents obtaining a similarity weight distribution, and in region in common represents a similarity weight of each element, in this case pixel); adjusting, based on the similarity weight of each element, a contribution of a corresponding element in a similarity term of a loss function to the loss function (Lais, ¶[0085] “To train the alignment model, the training module 620 can iteratively determine a solution to an optimization problem with respect to a loss function based on a similarity metric of a pair of label maps and a deformation vector field associated with a training reference image and a training displaced image. Such a solution defines a trained machine-learning alignment model.” This represents adjusting, based on the similarity weight of each element, a contribution of a corresponding element in a similarity term of a loss function to the loss function, since we have an optimization problem with respect to a loss function based on a similarity metric in Lais), the similarity weight of the element having a positive correlation relationship with a registration requirement accuracy associated with a region in which the element is located (Lais, ¶[0182] “a weighted average of one of the first tile deformation vector field or the second tile deformation vector field and one of the respective adjacent tile deformation vector fields overlapping at the pixel; and assigning, for each pixel within the defined region in common, the weighted average to the deformation vector field.” vector fields overlapping at the pixel represent the correlation and the defined region in common is a registration requirement accuracy associated with a region in which the element is located since there is a region there is location representing which the element is located); and obtaining, based on an optimized loss function obtained by the adjustment, a target deformation field (Lais, ¶[0190] “Embodiment 15. The computer-implemented method of embodiment 13, wherein the loss function comprises a Dice similarity coefficient and the gradient of the deformation vector field, wherein the gradient is weighted by a regularization factor.” This represents a deformation field since the loss function comprises the deformation vector field).
As per claim 2, Lais teaches, the method of claim 1, further comprising: obtaining a regularization weight distribution containing a regularization weight of each element (Lais, ¶[0190] [0190] Embodiment 15. The computer-implemented method of embodiment 13, wherein the loss function comprises a Dice similarity coefficient and the gradient of the deformation vector field, wherein the gradient is weighted by a regularization factor.” This represents regularization weight distribution); and adjusting, based on the regularization weight of each element, a contribution of a corresponding element in a regularization term of the loss function to the loss function (Lais, ¶[0086-87] “The parameter λ.sub.reg is a regularization factor, and ∇{right arrow over (u)} is the gradient (that is, magnitude of change) of the inferred deformation vector field {right arrow over (u)}
¶[0087] In Eq. (1), the regularization term λ.sub.reg ∇{right arrow over (u)} discourages abrupt large deformations. The dice score Dice(s.sub.m, s.sub.f) is a similarity metric that assesses agreement of the pixel-wise class labels between the reference label map and displaced label map, instead of the pixel-wise color and intensity agreement between the training reference image and the training displaced image. The loss function L(m, f) suits the fact that the tissue slides to be aligned are not expected to be identical regarding fine grain details, such as the position of individual cells/nuclei. Instead, the matching is regarding the regions of cells.“ This represents adjusting, based on the regularization weight of each element with the loss function ¶[0085] “To train the alignment model, the training module 620 can iteratively determine a solution to an optimization problem with respect to a loss function based on a similarity metric of a pair of label maps and a deformation vector field associated with a training reference image and a training displaced image. Such a solution defines a trained machine-learning alignment model.”), the regularization weight of the element having a negative correlation relationship with a degree of freedom of tissue deformation associated with a region in which the element is located (Lais, ¶[0086] “Additionally, m and s.sub.m represent, respectively, a displaced image and the displaced label map associated with that image, after applying the inferred deformation vector field {right arrow over (y)}, output by the alignment model. The parameter λ.sub.reg is a regularization factor, and ∇{right arrow over (u)} is the gradient (that is, magnitude of change) of the inferred deformation vector field {right arrow over (u)}
¶[0087] In Eq. (1), the regularization term λ.sub.reg ∇{right arrow over (u)} discourages abrupt large deformations.” this represents negative correlation relationship with a degree of freedom of tissue deformation associated with a region in which the element is located, by having the regularization term λ.sub.reg ∇{right arrow over (u)} discourages abrupt large deformations).
As per claim 3, Lais teaches, the method of claim 1, wherein obtaining the similarity weight distribution comprises: performing a region segmentation on a medical image to be registered to obtain at least two first-type regions; each of the first-type regions being associated with a corresponding registration requirement accuracy; and determining, based on the registration requirement accuracy associated with the first-type region in which the element is located and the positive correlation relationship, the similarity weight of each element, and forming the similarity weight distribution.
As per claim 10, Lais teaches, the method of claim 3, wherein performing the region segmentation on the medical image to be registered comprises: determining a reference image among the at least two medical images to be registered (Lais, ¶[0087] “instead of the pixel-wise color and intensity agreement between the training reference image and the training displaced image.” This represents determining a reference image among the at least two medical images to be registered ); and performing the region segmentation on the reference image (Lais, ¶[0098] “The reference slide and the displaces slide can be obtained by consecutively slicing an organ of a subject, for example. As is described herein, a position of a pixel within the reference image can be defined with respect to a reference 2D coordinate system, and a position of a pixel within the particular displaced image need not be defined within the reference 2D coordinate system. Accordingly, a pixel at a particular position within the displaced image need not correspond to a pixel at that same position in the reference image.” This represents performing the region segmentation on the reference image because reference slide and the displaces slide can be obtained by consecutively slicing an organ of a subject, for example and this would ).
As per claim 11, Lais teaches the method of claim 1, wherein adjusting, based on the similarity weight of each element, the contribution of the corresponding element in the similarity term of the loss function to the loss function comprises: determining, for each element, based on a pixel value of the element in a reference image and a pixel value of the element in a motion image after a deformation field acts, a pixel difference term of the element in the similarity term; and assigning a corresponding similarity weight to the pixel difference term of each element (Lais, ¶[0190] “Embodiment 15. The computer-implemented method of embodiment 13, wherein the loss function comprises a Dice similarity coefficient and the gradient of the deformation vector field, wherein the gradient is weighted by a regularization factor.” And ¶[0193] “[0193] Embodiment 18. The computer-implemented method of any one of embodiments 13-17, wherein the generating the multiple pairs of training images comprises: configuring labels for respective pixels spanning an area of a defined size; and generating a base label map based on the configured labels, the base label map spanning the area of the defined size.” Pixels get labeled and it would be part of the embodiment 15 and this would represent a pixel difference term of the element in the similarity term; and assigning a corresponding similarity weight to the pixel difference term of each element).
Allowable Subject Matter
Claims 4-9 and 12-19 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims. While claims 18-19 are allowed they still have to overcome the 101 rejection above of a software per se rejection. Regarding the objecting allowability of claim 4, at the time of the effective filing date the region-weighted is innovative as it overcomes the traditional computational trade-off in medical image registration. Instead of using uniform mathematical weights across a full image, this method dynamically prioritizes critical tissues (like tumors or organs) for pinpoint alignment while applying more flexible, less computationally heavy alignments to surrounding, non-essential background areas. The relevant prior art typically treated all pixels/elements the same, resulting in blurring errors or over-taxed computational resources. This approach solved the problem by tying alignment constraints directly to the clinical importance of the segmented region. This establishes distinct Segmentation Maps and weighting zones appropriately, models could achieve higher overall accuracy without suffering from massive memory demands or unrealistic tissue deformation. Essentially prioritizing the injuries/lesions or organs. As per claim 5, being objected allowable as it was not found in the prior art, at the time of the effective filing date registration heavily relied on structural features like matching edges and tissue boundaries. This method, however, takes a functional approach by integrating positron emission tomography (PET) tracer concentration data. If an area has a high concentration of a radioactive tracer (e.g., hyperactive metabolic activity), the algorithm recognizes it as medically crucial and assigns a higher registration weight. With regards to claim 9, traditionally, an image similarity loss function (like Mean Squared Error or Normalized Cross-Correlation) treats every pixel or voxel equally. This method adjusts the weight of the similarity term depending on the specific region. This allows the algorithm to prioritize high-accuracy alignment in critical diagnostic areas (e.g., margins of a tumor) while relaxing constraints where precision is less vital. The prior art relies on spatially-varying methods heavily relied on strict, pre-labeled anatomical atlases. This method shifts to estimate these spatially-varying weights directly from the underlying data and features—improving robustness and adapting to real-time image shifts. As per claim 12, the prior art penalizes the "deformation field" (the mathematical grid used to align two images) uniformly across the entire image. This caused issues: if an organ moved dramatically or had a sharp anatomical boundary, the model would over-smooth the edges or cause what is referred to as “folding”. Earlier prior art (like also tried adjusting regularization, but they relied entirely on pre-existing anatomical segmentations. If the segmentation map was inaccurate, the regularization failed. This way this can all be labeled ahead of time by pixel to focus on those areas and if the organ is different then prior to analysis this is picked up. The algorithm assigns a weight to the spatial gradient term of each element. Elements are tiny subdivisions of the image grid. The algorithm determines the deformation field gradient at that exact location and adjusts the penalty weight which is useful, as if the gradient is high (indicating a sharp boundary or significant tissue deformation), the regularizer adjusts its contribution so the boundary isn't unrealistically smoothed out or distorted without reliance on possible bad image grids. As per claims 17-18, the problem that this solves is in attenuation coefficient images (like CT scans) contain a vast dynamic range, stretching from low-contrast soft tissues to high-contrast bones. Standard global enhancements often amplify noise or blow out details in areas where the contrast is naturally low. And instead of treating the image uniformly, the system dynamically segments pixel/voxel elements into distinct "attenuation coefficient intervals". By applying a mapping function with a steeper slope which was not found in the prior art, specifically to the interval with the minimum image contrast, the algorithm selectively boosts subtle details in ultra-low-contrast tissue zones while preventing over-saturation in naturally high-contrast areas. This provides an optimized, highly detailed preprocessed image tailored for accurate alignment. And further, most neural networks found in the prior art applied a single, global regularization weight to the entire image. However, human anatomy is not uniformly rigid. A global weight creates a major conflict: if it is too loose, soft organs warp unrealistically; if it is too tight, the algorithm cannot properly capture complex, localized tissue movements. And this method aims to solve this. This method then aims to solve element-wise (voxel-by-voxel) adaptive regularization. It computes the spatial gradient of the deformation field at every individual element. It then dynamically assigns a corresponding regularization weight to each element based on its unique directional changes, however as a reminder the 101 rejection above has to be overcome.
A close prior art that could have also been a reference and to consider when amending is, Fan (US 2020/0074626) ¶[0033] “[0033] Different from conventional multi resolution image registration algorithms in which deformation fields at lower-resolutions are typically used as initialization inputs to image registration at a higher spatial resolution, the disclosed example deep learning based method may jointly optimize deformation fields at all spatial resolutions with a typical feedforward and backpropagation based deep learning setting. As the optimization of the loss function proceeds, the parameters within the network may be updated through the feedforward computation and backpropagation procedure, leading to improved prediction of deformation fields. It is worth noting that no training deformation field information is needed for the optimization, and self-supervision through maximizing image similarity with smoothness regularization of deformation fields may be the only force to drive the optimization. The trained network can be directly used to register a pair of images, and any of them can be the fixed image.” And ¶[0024] “Given a pair of fixed image If and moving image Im, the task of image registration is to seek a spatial transformation that establishes pixel/voxel-wise spatial correspondence between the two images. Since the spatial correspondence can be gauged with a surrogate measure, such as an image intensity similarity, the image registration task can be formulated as an optimization problem to identify a spatial transformation that maximizes the image similarity measure between the fixed image and transformed moving image. For non-rigid image registration, the spatial transformation is often characterized by a dense deformation field Dv that encodes displacement vectors between spatial coordinates of If and their counterparts of Im.” Avoids image similarity but does not add all the details and analysis as discussed with regards to the dependent claims. And this also does ¶[0050] “[0050] The fully convolutional network may be configured to apply a displacement field for registering the first image and the second image on one or more of a voxel-by-voxel or a pixel-by-pixel basis.” On a pixel by pixel basis registering which is a key feature of the current invention. ¶[0056] Aspect 2. The method of Aspect 1, wherein the fully convolutional network is configured to apply a displacement field for registering the first image and the second image on one or more of a voxel-by-voxel or a pixel-by-pixel basis.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SANTIAGO GARCIA whose telephone number is (571)270-5182. The examiner can normally be reached Monday-Friday 9:30am-5:30pm.
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, Chineyere Wills-Burns can be reached at (571) 272-9752. 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.
/SANTIAGO GARCIA/Primary Examiner, Art Unit 2673
/SG/
Prosecution Insights