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 .
Response to Amendment
Applicant’s response, filed 26 May 2026, to the last office action has been entered and made of record.
In response to the amendments to the claims, they are acknowledged, supported by the original disclosure, and no new matter is added.
In response to the filed replacement drawings, the replacement drawings have overcome the objections to the drawings of the previous Office action, and the respective objections have been withdrawn.
In response to the amendments to the claims, specifically addressing the objection(s) to the claims of the previous Office action, the amended language has overcome the respective objection(s), and the objection(s) has been withdrawn.
In response to the amendments to the claims, specifically addressing the treatment of claim 14 under 35 U.S.C. § 112(f) / (pre-AIA ) sixth paragraph, of the previous Office action, the amended language has obviated the necessity to interpret the respective claim elements of claim 14 under 35 U.S.C. § 112(f) / (pre-AIA ) sixth paragraph, and the claim is treated with its broadest reasonable interpretation.
Amendments to the independent claims 1, 11, 14, and 20 have necessitated a new ground of rejection over the applied prior art. Please see below for the updated interpretations and rejections.
In response to the addition of new claim 21, it is acknowledged and made of record.
Response to Arguments
Applicant's arguments filed 26 May 2026 have been fully considered but they are not fully persuasive.
In response to Applicant’s arguments on p. 17-19 of Applicant’s reply, that the teachings of Poole fails to teach or suggest the amended features of “splitting the registered channel data along the symmetry plane or the symmetry axis into at least a first channel dataset and at least a second channel dataset, the first channel dataset including a first portion of the registered channel data corresponding to the first part of the symmetric organ, and the second channel dataset including a second portion of the registered channel data corresponding to the second part of the symmetric organ; [and] mirroring the second channel dataset along the symmetry plane or the symmetry axis to obtain a mirrored second channel dataset”, the Examiner respectfully disagrees.
Examiner notes the claims are treated with their broadest reasonable interpretations consistent with the specification. See MPEP 2111. Although the claims are interpreted in light of the specification, limitations from the specification are not read into the claims. See In re Van Geuns, 988 F.2d 1181, 26 USPQ2d 1057 (Fed. Cir. 1993). Furthermore, the test for obviousness is what the combined teachings of the references would have suggested to those of ordinary skill in the art. See In re Keller, 642 F.2d 413, 208 USPQ871 (CCPA 1981).
Poole is noted to further teach that semi-atlas channel data is added to the first and second folded data sets, where the semi-atlas channel data comprises channels which are representative of the x, y, and z coordinates for each voxel in the coordinate space of the folded datasets (see Poole [0102]-[0107]).
As the corresponding x, y, and z channels are added to the respective folded data sets, Poole’s teachings suggests that the semi-atlas channel data representing the coordinate space are similarly divided and folded to the corresponding folded datasets, such that coordinates channels corresponding to the inverted right part in the second folded data set are added to the second folded dataset, where the corresponding coordinates would also be folded to correspond with the inverted right part in the second folded data set. The suggested folding of respective coordinates corresponding with inverted right part in the second folded data set provides for the broadest reasonable interpretation of mirroring the second channel dataset along the symmetry plane or symmetry axis to obtain a mirrored second channel dataset.
Thus, one of ordinary skill in the art would have found it obvious from the teachings of the Poole reference to suggests the broadest reasonable interpretations of the claim features of “splitting the registered channel data along the symmetry plane or the symmetry axis into at least a first channel dataset and at least a second channel dataset, the first channel dataset including a first portion of the registered channel data corresponding to the first part of the symmetric organ, and the second channel dataset including a second portion of the registered channel data corresponding to the second part of the symmetric organ; [and] mirroring the second channel dataset along the symmetry plane or the symmetry axis to obtain a mirrored second channel dataset”.
Applicant’s arguments with respect to amended independent claims 1, 11, 14, and 20 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.
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 text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claims 1-2, 4-12, and 14-20 are rejected under 35 U.S.C. 103 as being unpatentable over Poole et al. (US 2018/0025255), herein Poole, in view of Han et al. (US 2019/0251694), herein Han.
Regarding claim 1, Poole discloses a method for generating training data for training a deep learning algorithm, the method comprising:
receiving medical imaging data of an examination area of a patient, the examination area including a first part of a symmetric organ and a second part of the symmetric organ (see Poole [0019]-[0022], where volumetric imaging data representative of an anatomical structure of a patient is obtained, where the anatomical structure is the brain or may be any anatomical structure that is substantially symmetrical; see Poole [0031], where training data sets are obtained comprising CT scans of the brains of a plurality of training subjects);
splitting the medical imaging data along a symmetry plane or a symmetry axis into a first dataset and a second dataset, the first dataset including a first portion of the medical imaging data corresponding to the first part of the symmetric organ, and the second dataset including a second portion of the medical imaging data corresponding to the second part of the symmetric organ (see Poole [0072]-[0073] and [0077], where a line of symmetry in transformed training data set is determined and taken to be a midline of the transformed training data set and used to extract a midline centered data set from the transformed training data set; see Poole [0086]);
mirroring the second dataset along the symmetry plane or the symmetry axis to obtain a mirrored second dataset (see Poole [0091], where in the folding stage, one of the left and the right part is inverted with respect to the midline, where the voxels of the inverted part overlay the voxels of the non-inverted part; see Poole [0097]-[0098], where the right part voxels and inverted left part voxels occupy exactly the same positions in space and may consider that each pair of mirror-image voxels, each having a respective intensity value, is replaced by a single right part voxel with two intensity values);
generating the training data by stacking the first dataset and the mirrored second dataset (see Poole [0115], where the folded datasets are concatenated to provide a single data set); and
providing the training data (see Poole [0120], where patches of the combined folded data sets are selected on which to train a detector).
While Poole does not explicitly teach splitting the midline centered data set into two separate data sets, Poole does teach that the determined midline may be considered to divide a first and second part of the midline centered data set (see Poole [0091]), and provides an implicit teaching for the broadest reasonable interpretation of “splitting the medical imaging data along a symmetry plane or a symmetry axis into a first dataset and a second dataset”. See also MPEP 2144.01.
At the time of filing, one of ordinary skill in the art would have found it obvious from Poole’s teachings that the determined midline divides the midline centered data set into a first and second part provides an implicit teaching for a first data set comprising the first part of the midline centered data set and a second data set comprising the second part of the midline centered data set.
This modification is rationalized as some teaching, suggestion, or motivation in the prior art that would have led one of ordinary skill to modify the prior art reference or to combine prior art reference teachings to arrive at the claimed invention.
In this instance, Poole teaches that the determined midline may be considered to divide a first and second part of the midline centered data set.
One of ordinary skill in the art would understand that the determined midline of the midline centered data allows for a divide between a first and second part of the midline centered data set, allowing for the first and second part of the midline centered data set to be separately considered.
Thus, it would have been within the general knowledge of one of ordinary skill in the art to reasonably expect that the midline dividing the first and second part of the midline centered data set provides an implicit teaching for a first data set comprising the first part of the midline centered data set and a second data set comprising the second part of the midline centered data set and providing the broadest reasonable interpretation of “splitting the medical imaging data along a symmetry plane or a symmetry axis into a first dataset and a second dataset”.
While Poole teaches that the folded data sets are concatenated to form a combined folded dataset and used to train a classifier comprising a convolutional neural network (see Poole [0109]-[0110], [0115]-[0117], [0120], [0122], [0125]-[0128], and [0131]-[0134]), where for each selected patch locations from the combined folded data set, two training patches are obtained, where the training patches have the corresponding listing orders for the intensities of the right part and left part as the combined folded data set (see Poole [0122)-[0124]; Poole does not explicitly disclose that the generating the training data by stacking the first dataset and mirrored second dataset is in a structure of the training data according to an ordering direction of the structure, the ordering direction corresponding to a readout order of the training data.
Han teaches in a related and pertinent method for training and using a trained machine learning classifier for atlas based segmentation (see Han Abstract), where the CNN model for implementing the image segmentation receives input stack of adjacent 2D images and outputs at least one feature vector or matrix representing the features of the input stack, where the vector contains information of the anatomical structures in input stack of images of the CNN model, and the stack of adjacent 2D images that contain dependent structure information are used for both training and as input of the CNN model to improve the accuracy of the output prediction (see Han [0070]-[0072] and [0082]).
At the time of filing, one of ordinary skill in the art would have found it obvious to apply the teachings of Han to the teachings of Poole, such that the image patches from the combined folded data set and corresponding training image patches include listing orders of entries for the intensities of the right part and left part, and that the image patches are represented as corresponding feature vectors to be output and used to train the CNN based classifier in the corresponding order.
This modification is rationalized as an application of a known technique to a known method ready for improvement to yield predictable results.
In this instance, Poole discloses a base method for generating training data and training a detector to detect abnormal voxels in medical volumetric imaging data obtained from a CT scanner, folded data sets are concatenated to form a combined folded dataset and used to train a detector of a classifier comprising a convolutional neural network, and where patches are selected from the combined folded data set with a corresponding listing order of the intensities of the right parts and left parts, in which training patches obtained from the patch locations have corresponding listing orders.
Han teaches a known technique for training a machine learning based classifier for atlas-based segmentation, where input stacks of images for a CNN model for implementing the image segmentation extracts feature vectors representing the features of the input stack containing the anatomical structures in the stack of images, and are used in the training of the CNN model.
One of ordinary skill in the art would have recognized that by applying Han's technique would allow for the method of Poole to extract corresponding feature vectors to represent the training image patches obtained from the combined folded data set to be used to train the CNN based classifier, predictably leading to an improved feature representation for the training data used to train the detector to detect abnormal voxels in medical volumetric imaging data for improving the accuracy of the output prediction.
Regarding claim 2, please see the above rejection of claim 1. Poole and Han disclose the method according to claim 1, further comprising:
receiving organ atlas data (see Poole [0048]-[0049], where set of reference landmarks are received for use in atlas alignment, where the atlas data set comprise a set of reference data); and
generating registered imaging data based on the medical imaging data and the organ atlas data (see Poole [0052]-[0054], where the training data and ground truth data sets are aligned to the atlas data set; see Poole [0065]-[0076], where the alignment process performs a rigid registration of the training data and the atlas data set using the relationship between corresponding landmark locations; and see Poole [0077]-[0081], where midline centered data set is extracted from the transformed training data set);
wherein the splitting splits the registered imaging data into the first dataset and the second dataset (see Poole [0090]-[0092] and [0096]-[0100], where the midline centered data set is divided by the midline into a first part and second part and two folded data sets are obtained), the first dataset including a first portion of the registered imaging data corresponding to the first part of the symmetric organ, and the second dataset including a second portion of the registered imaging data corresponding to the second part of the symmetric organ (see Poole [0098]-[0100], where the first folded data set includes intensity data from the left part of the midline centered data and the second folded data set includes intensity data from the right part of the midline centered data).
Regarding claim 4, please see the above rejection of claim 1. Poole and Han disclose the method according to claim 1, further comprising:
performing one of
receiving at least one channel including channel data (see Poole [0102], where semi-atlas channel data is added to first and second folded data set, the semi-atlas channel data comprises three extra channels representative of the x, y, and z coordinates for each voxel int the coordinate space of the folded data sets; suggesting at least that the extra channels are received to be added to the folded data set), or
calculating the at least one channel based on the medical imaging data, wherein the at least one channel includes channel data (see Poole [0093] and [0102]-[0103], where the symmetrical portion of the image is transformed to be aligned with the coordinate space of the atlas data set, in which the semi-atlas channel data represents the x, y, and z coordinates of the voxel in the coordinate space of the folded data sets; thus suggesting that the coordinates of the coordinate space of the folded data sets are calculated transformations to be aligned with the atlas data set, providing for the broadest reasonable interpretation of calculating the at least one channel based on the medical imaging data);
generating registered channel data based on the channel data and organ atlas data (see Poole [0102]-[0107], where the coordinate space of the folded data sets, including the added semi-atlas channel data is the same as the coordinate space of the atlas data set);
splitting the registered channel data along the symmetry plane or the symmetry axis into at least a first channel dataset and at least a second channel dataset, the first channel dataset including a first portion of the registered channel data corresponding to the first part of the symmetric organ, and the second channel dataset including a second portion of the registered channel data corresponding to the second part of the symmetric organ (see Poole [0102]-[0107], where the semi-atlas channel data comprises channels which are representative of the coordinates for each voxel in the coordinate space of the folded datasets; thus suggesting that the semi-atlas channel data representing the coordinate space are similarly divided and folded to the corresponding folded datasets);
mirroring the second channel dataset along the symmetry plane or the symmetry axis to obtain a mirrored second channel dataset (see Poole [0102]-[0107], where the semi-atlas channel data comprises channels which are representative of the coordinates for each voxel in the coordinate space of the folded datasets; thus suggesting that the semi-atlas channel data are similarly divided and folded to the corresponding folded datasets, e.g. the coordinates for an inverted right part in the second folded data set are also correspondingly folded to correspond to the inverted right part in the second folded data set); and
generating the training data by stacking the first dataset, the first channel dataset, the mirrored second dataset and the mirrored second channel dataset (see Poole [0109], where the folded data sets are concatenated to form the combined folded data set; see also Poole Fig. 8 and [0111]-[0115]).
Regarding claim 5, please see the above rejection of claim 4. Poole and Han disclose the method according to claim 4, wherein at least one of:
the at least one channel is a vessel channel, the channel data including vessel data, and the vessel data being based on a segmentation of a cerebrovascular vessel tree in the medical imaging data (see Poole [0050], where anatomical landmarks may be defined anatomically, in relation to anatomical structures such as bones, vessels or organs; see Poole [0106]-[0107], where semi-atlas channel data added to the folded data comprise label for each voxel corresponding to segmented atlas data indicating the anatomical structure to which they belong, where each segmented region of the atlas data set is used as an individual binary mask and the semi-atlas channel data comprises data from those binary masks); or
the at least one channel is a bone channel, the channel data including bone data, and the bone data being based on at least one of a bone mask or a segmentation of bones in the medical imaging data (see Poole [0050], where anatomical landmarks may be defined anatomically, in relation to anatomical structures such as bones, vessels or organs; see Poole [0106]-[0107], where semi-atlas channel data added to the folded data comprise label for each voxel corresponding to segmented atlas data indicating the anatomical structure to which they belong, where each segmented region of the atlas data set is used as an individual binary mask and the semi-atlas channel data comprises data from those binary masks).
Regarding claim 6, please see the above rejection of claim 4. Poole and Han disclose the method according to claim 4, wherein
the first dataset is prior to the first channel dataset in the readout order (see Poole Fig. 8 and [0113]-[0115], where the intensity channel comprising intensities of the left part is prior to the semi-atlas channel data of the left part in the first folded dataset of the combined folded data set);
the first channel dataset is prior to the mirrored second dataset in the readout order (see Poole Fig. 8 and [0113]-[0115], where the semi-atlas channel data of the left part in the first folded dataset of the combined folded data set is prior to the intensity channel comprising intensities of the inverted right part in the second folded dataset of the combined folded data set); and
the mirrored second dataset is prior to the mirrored second channel dataset in the readout order (see Poole Fig. 8 and [0113]-[0115], where the intensity channel comprising intensities of the inverted right part in the second folded dataset is prior to the semi-atlas channel data of the second folded dataset of the combined folded data set).
Regarding claim 7, please see the above rejection of claim 1. Poole and Han disclose the method according to claim 1, wherein
the medical imaging data is a first medical imaging data (see Poole [0031], where the training data sets are comprised of imaging data sets obtained from CT scans of the brains of a plurality of training subjects);
the patient is a first patient (see Poole [0031] and [0033], where obtaining image data sets from CT scans of brains of plurality of training subjects suggests at least a first patient);
the training data is first training data (see Poole [0031] and [0033], where the training data sets are comprised of imaging data sets obtained from CT scans of the brains of a plurality of training subjects; suggesting that the training data sets comprise multiple different training data sets corresponding to the brains of the plurality of training subjects and including a first training data set); and
the method further comprises generating second training data by stacking a third dataset and the mirrored second dataset, the third dataset being based on second medical imaging data of a second patient different from the first patient (see Poole [0031], where the training data sets are comprised of imaging data sets obtained from CT scans of the brains of a plurality of training subjects; see Poole [0116]-[0125], where the combined folded data set is used as a training data set to train the detector;
and see Poole [0130], where the detector is trained on patches derived from a plurality, or optionally, all of the training data sets; suggesting that a combined training data set includes training dataset based on CT scans and the combined folded data sets derived from CT scans the brains of different training subjects).
Regarding claim 8, please see the above rejection of claim 7. Poole and Han disclose the method according to claim 7, further comprising:
choosing the third dataset and the mirrored second dataset for generating the second training data based on boundary conditions, the boundary conditions corresponding to patient information data (see Poole [0031], where the training data sets have been identified as exhibiting signs of a pathology of interest or may be representative of anatomical structure, a different abnormality and/or a different modality).
Regarding claim 9, please see the above rejection of claim 1. Poole and Han disclose a method for training a deep learning algorithm to detect disease information in medical imaging data of an examination area of a patient, the method comprising:
receiving training data generated based on the method according to claim 1 (see above claim 1 rejection; see Poole [0120], where patches of the combined folded data sets are selected on which to train a detector);
receiving output training data including disease information corresponding to the training data (see Poole [0031]-[0035], where the training data sets have been identified as exhibiting signs of a pathology of interest; see Poole [0125]-[0134], where a detector comprising a classifier is trained to detect abnormal voxels, suggesting that a detection output is provided from the detector being trained);
training the deep learning algorithm based on the training data and the output training data to obtain a trained deep learning algorithm (see Poole [0122]-[0134], where the patches and corresponding ground truth data are used to train the detector to detect abnormal voxels, where the classifier of the detector to be trained comprises a convolutional neural network); and
providing the trained deep learning algorithm (see Poole [0133] and [0136], where a trained detector is output).
Regarding claim 10, please see the above rejection of claim 9. Poole and Han disclose the method according to claim 9, wherein the training of the deep learning algorithm is based on a symmetry of the training data and the output training data, the symmetry corresponding to an influence of switching a first dataset and a mirrored second dataset in the training data on the disease information in the output training data (see Poole [0031]-[0035], where the training data sets have been identified as exhibiting signs of a pathology of interest; see Poole [0125]-[0134], where a detector comprising a classifier is trained to detect abnormal voxels; see Poole [0125]-[0126], where the detector is trained on patches extracted from the combined folded data sets and trained such that it is left-right agnostic, where the first and second intensity data is used without taking into account whether the first intensity data came from the left or from the right side).
Regarding claim 11, Poole discloses a method for detecting disease information, the method comprising:
receiving medical imaging data of an examination area of a patient, the examination area including a first part of a symmetric organ and a second part of the symmetric organ (see Poole [0137], where novel data set is received by the detection circuitry; and see Poole [0139]-[0141], where a novel midline centered data set is extracted from transformed novel data sets and corresponds to symmetrical portion of the atlas data set and used to obtain a first part and second part);
splitting the medical imaging data along a symmetry plane or a symmetry axis into a first dataset and a second dataset, the first dataset including a first portion of the medical imaging data corresponding to the first part of the symmetric organ, and the second dataset including a second portion of the medical imaging data corresponding to the second part of the symmetric organ (see Poole [0138]-[0141], where the novel dataset is aligned with the atlas data set and a midline of the novel data set is determined and refined to extract a novel midline centered data set and corresponds to the same symmetrical portion of the atlas data set and used to obtain a first part and second part);
mirroring the second dataset along the symmetry plane or the symmetry axis to obtain a second mirrored dataset (see Poole [0141]-[0142], where the midline centered data set is folded at its midline to obtain a first part and a second part, producing two channel volume which are mirror imaged);
generating subject patient image data by stacking the first dataset and the mirrored second dataset (see Poole [0141]-[0142], where the first folded data set comprises a first channel representative of intensity of the left part of the midline centered data and a second channel representative of the intensity of the right part of the midline data set);
analyzing the subject patient image data by applying a trained deep learning algorithm to the subject patient image data in the readout order to obtain an analysis result, the trained deep learning algorithm having been trained using training data (see Poole [0144]-[0147], where the trained detector is applied to the first folded data set; see Poole [0122]-[0134], where the patches and corresponding ground truth data are used to train the detector to detect abnormal voxels, where the classifier of the detector to be trained comprises a convolutional neural network); and
detecting the disease information based on the analysis result(see Poole [0148]-[0150], where a probability that a voxel is abnormal is assigned to each voxel, and outputs a probability volume that comprises for each voxel of the novel midline centered data set, a probability that the voxel is abnormal).
At the time of filing, one of ordinary skill in the art would have found it obvious from Poole’s teachings that the determined midline divides the novel midline centered data set into a first and second part provides an implicit teaching for a first data set comprising the first part of the novel midline centered data set and a second data set comprising the second part of the novel midline centered data set.
This modification is rationalized as some teaching, suggestion, or motivation in the prior art that would have led one of ordinary skill to modify the prior art reference or to combine prior art reference teachings to arrive at the claimed invention.
In this instance, Poole teaches that the determined midline may be considered to divide a first and second part of the novel midline centered data set.
One of ordinary skill in the art would understand that the determined midline of the novel midline centered data allows for a divide between a first and second part of the novel midline centered data set, allowing for the first and second part of the novel midline centered data set to be separately considered.
Thus, it would have been within the general knowledge of one of ordinary skill in the art to reasonably expect that the midline dividing the first and second part of the novel midline centered data set provides an implicit teaching for a first data set comprising the first part of the novel midline centered data set and a second data set comprising the second part of the novel midline centered data set and providing the broadest reasonable interpretation of “splitting the medical imaging data along a symmetry plane or a symmetry axis into a first dataset and a second dataset”.
While Poole teaches that the first and second folded data sets, where folding produces a two channel volume, one of which channels is mirror imaged, and comprise of corresponding ordered channels representative of intensity of the left part of the midline centered data set and the intensity of the right part of the midline, and the trained detector is applied to the first and second folded data set, to which corresponding semi-atlas channels have been added (see Poole [0141]-[0142] and [0144]-[0147]);
Poole does not explicitly disclose that the generating the subject patient image data by stacking the first dataset and mirrored second dataset is in a structure of the subject patient image data according to an ordering direction of the structure, the ordering direction corresponding to a readout order of the subject patient image data.
Han teaches in a related and pertinent method for training and using a trained machine learning classifier for atlas based segmentation (see Han Abstract), where the CNN model for implementing the image segmentation receives input stack of adjacent 2D images and outputs at least one feature vector or matrix representing the features of the input stack, where the vector contains information of the anatomical structures in input stack of images of the CNN model, and the stack of adjacent 2D images that contain dependent structure information are used for both training and as input of the CNN model to improve the accuracy of the output prediction (see Han [0070]-[0072] and [0082]).
At the time of filing, one of ordinary skill in the art would have found it obvious to apply the teachings of Han to the teachings of Poole, such that the folded data sets applied to the trained detector include listing orders of entries for the intensities of the right part and left part, and that image data are represented as corresponding feature vectors to be output and processed by the trained CNN based classifier in the corresponding order.
This modification is rationalized as an application of a known technique to a known method ready for improvement to yield predictable results.
In this instance, Poole discloses a base method for detecting abnormal voxels in medical volumetric imaging data obtained from a CT scanner using a trained detector, where the first and second folded data sets comprise of corresponding ordered channels representative of intensity of the left part of the midline centered data set and the intensity of the right part of the midline, where folding produces a two channel volume, one of which channels is mirror imaged, and the trained detector is applied to the first and second folded data set, to which corresponding semi-atlas channels have been added.
Han teaches a known technique for training a machine learning based classifier for atlas-based segmentation, where input stacks of images for a CNN model for implementing the image segmentation extracts feature vectors representing the features of the input stack containing the anatomical structures in the stack of images, and are used in the training of the CNN model.
One of ordinary skill in the art would have recognized that by applying Han's technique would allow for the method of Poole to extract corresponding feature vectors to represent the folded image data sets to be processed by the trained detector, predictably leading to an improved feature representation for the folded data set to be processed by the trained detector to detect abnormal voxels in medical volumetric imaging data for improving the accuracy of the output prediction.
Regarding claim 12, please see the above rejection of claim 11. Poole and Han disclose the method according to claim 11, further comprising:
receiving a radiological finding based on the medical imaging data or the subject patient image data (see Poole [0151]-[0153], where for a given voxel, if the probability of being abnormal is above a threshold value, the voxel is classified as abnormal); and
determining view information by applying a view determining algorithm to the medical imaging data or to the subject patient image data, the view information including at least one of a projection or a projection angle corresponding to a region of the radiological finding (see Poole [0154]-[0162], the voxels that have been classified as abnormal are grouped into abnormal regions, the abnormal regions are ranked, and presented to a clinician or user, where an image from the novel data set is rendered with the abnormal region or regions displayed on the rendered image, and that a list of detections is displayed that is ranked by probability, and clicking on a listing of an abnormal region navigates to an MPR (multi-planar reformatting) view of the abnormal region, where each detected abnormal region is highlighted by colour in the MPR view).
Regarding claim 14, it recites a system performing the method of claim 11. Poole and Han teach a system performing the method of claim 11. Please see above for detailed claim analysis, with the exception to the following further limitations:
a first interface (see Poole Fig. 1 and [0024], where volumetric imaging data sets are provided to the computing apparatus from the memory), a processing circuitry (see Poole Fig. 1 and [0025]-[0028], where the computer apparatus includes a central processing unit, hard drive and other components of a PC including RAM and ROM, and that the disclosed circuitries are each implemented in the computing apparatus by means of a computer program having a computer-readable instructions that are executable to perform the disclosed methods), and a second interface (see Poole Fig. 1 and [0019], where the computing apparatus is connected to one or more display screens).
Please see the above rejection for claim 11, as the rationale to modify the teachings of Poole and Han are similar, mutatis mutandis.
Regarding claim 15, please see the above rejection of claim 1. Poole and Han discloses a non-transitory computer-readable storage medium storing computer-executable instructions that, when executed by a computer, cause the computer to carry out the method according to claim 1 (see Poole [0025]-[0028], where the computer apparatus includes a central processing unit, hard drive and other components of a PC including RAM and ROM, and that the disclosed circuitries are each implemented in the computing apparatus by means of a computer program having a computer-readable instructions that are executable to perform the disclosed methods).
Regarding claim 16, see above rejection for claim 2. It is a method claim reciting similar subject matter as claim 4. Please see above claim 4 for detailed claim analysis as the limitations of claim 16 are similarly rejected.
Regarding claim 17, see above rejection for claim 3. It is a method claim reciting similar subject matter as claim 4. Please see above claim 4 for detailed claim analysis as the limitations of claim 17 are similarly rejected.
Regarding claim 18, see above rejection for claim 17. It is a method claim reciting similar subject matter as claim 5. Please see above claim 5 for detailed claim analysis as the limitations of claim 18 are similarly rejected.
Regarding claim 19, see above rejection for claim 5. It is a method claim reciting similar subject matter as claim 6. Please see above claim 6 for detailed claim analysis as the limitations of claim 19 are similarly rejected.
Regarding claim 20, please see the above rejection of claim 1. Poole and Han disclose a method for detecting disease information, the method comprising:
receiving medical imaging data of an examination area of a patient, the examination area including a first part of a symmetric organ and a second part of the symmetric organ (see Poole [0137], where novel data set is received by the detection circuitry; and see Poole [0139]-[0141], where a novel midline centered data set is extracted from transformed novel data sets and corresponds to symmetrical portion of the atlas data set and used to obtain a first part and second part);
splitting the medical imaging data along a symmetry plane or a symmetry axis into a first dataset and a second dataset, the first dataset including a first portion of the medical imaging data corresponding to the first part of the symmetric organ, and the second dataset including a second portion of the medical imaging data corresponding to the second part of the symmetric organ (see Poole [0138]-[0141], where the novel dataset is aligned with the atlas data set and a midline of the novel data set is determined and refined to extract a novel midline centered data set and corresponds to the same symmetrical portion of the atlas data set and used to obtain a first part and second part);
mirroring the second dataset along the symmetry plane or the symmetry axis to obtain a mirrored second dataset (see Poole [0141]-[0142], where the midline centered data set is folded at its midline to obtain a first part and a second part, producing two channel volume which are mirror imaged);
generating subject patient image data by stacking the first dataset and the mirrored second dataset (see Poole [0141]-[0142], where the first folded data set comprises a first channel representative of intensity of the left part of the midline centered data and a second channel representative of the intensity of the right part of the midline data set);
analyzing the subject patient image data by applying a trained deep learning algorithm to the subject patient image data to obtain an analysis result, the trained deep learning algorithm having been trained using training data generated according to the method of claim 1 (see Poole [0144]-[0147], where the trained detector is applied to the first folded data set; see Poole [0122]-[0134], where the patches and corresponding ground truth data are used to train the detector to detect abnormal voxels, where the classifier of the detector to be trained comprises a convolutional neural network; see above rejection of claim 1); and
detecting the disease information based on the analysis result (see Poole [0148]-[0150], where a probability that a voxel is abnormal is assigned to each voxel, and outputs a probability volume that comprises for each voxel of the novel midline centered data set, a probability that the voxel is abnormal),
At the time of filing, one of ordinary skill in the art would have found it obvious from Poole’s teachings that the determined midline divides the novel midline centered data set into a first and second part provides an implicit teaching for a first data set comprising the first part of the novel midline centered data set and a second data set comprising the second part of the novel midline centered data set.
This modification is rationalized as some teaching, suggestion, or motivation in the prior art that would have led one of ordinary skill to modify the prior art reference or to combine prior art reference teachings to arrive at the claimed invention.
In this instance, Poole teaches that the determined midline may be considered to divide a first and second part of the novel midline centered data set.
One of ordinary skill in the art would understand that the determined midline of the novel midline centered data allows for a divide between a first and second part of the novel midline centered data set, allowing for the first and second part of the novel midline centered data set to be separately considered.
Thus, it would have been within the general knowledge of one of ordinary skill in the art to reasonably expect that the midline dividing the first and second part of the novel midline centered data set provides an implicit teaching for a first data set comprising the first part of the novel midline centered data set and a second data set comprising the second part of the novel midline centered data set and providing the broadest reasonable interpretation of “splitting the medical imaging data along a symmetry plane or a symmetry axis into a first dataset and a second dataset”.
Regarding claim 21, please see the above rejection of claim 1. Poole and Han disclose the method according to claim 1, wherein the structure of the training data is a vector or a tensor (see Han [0070]-[0072] and [0082]), where the CNN model for implementing the image segmentation receives input stack of adjacent 2D images and outputs at least one feature vector or matrix representing the features of the input stack, where the vector contains information of the anatomical structures in input stack of images of the CNN model, and the stack of adjacent 2D images that contain dependent structure information are used for both training and as input of the CNN model to improve the accuracy of the output prediction; where the combined teachings suggest that the image patches of the combined folded data set are represented as corresponding feature vectors to be output and used to train the CNN based classifier in the corresponding order).
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Poole and Han disclose as applied to claim 2 above, and further in view of Hemingway (US 2023/0139458, effectively filed 28 October 2021).
Regarding claim 3, please see the above rejection of claim 2. Poole and Han disclose the method according to claim 2, wherein the organ atlas data includes brain atlas data and the symmetric organ is a brain, (see Poole [0031] and [0033], where the training imaging data sets are imaging data of brains of a plurality of training subjects; see Poole [0047]-[0051], where the set of reference landmarks comprises landmarks in or near the brain).
While Poole teaches that a midline is determined for dividing the midline centered data of an anatomical structure into left and right parts (see Poole [0072]-[0073] and [0091]), and that the anatomical structure is the brain (see Poole [0031] and [0033]);
Poole and Han do not explicitly disclose the first part of the symmetric organ being a first cerebral hemisphere, and the second part of the symmetric organ being a second cerebral hemisphere
Hemingway teaches in a related and pertinent medical image processing apparatus s and method to generate symmetrized volume data based on mirrored volume data (see Hemingway Abstract), where image data that is representative of an anatomical structure of a patient, where the anatomical structure is the brain and is substantially symmetrical between the left and right sides of the brain (see Hemingway [0030]-[0032]), and a symmetry detection is performed to detect a symmetry plane that divides the volume into first and second sections, where the symmetry plane divides the head into a left half comprising a left hemisphere of the brain and a right half comprising a right hemisphere of the brain (see Hemingway [0043]).
At the time of filing, one of ordinary skill in the art would have found it obvious to combine the teachings of Poole and Han with the teachings of Hemingway such that when the midline determined to divide a midline centered data set corresponding to a brain anatomical structure into symmetrical left and right parts, the left part includes the left hemisphere of the brain and the right part includes the right hemisphere of the brain.
This modification is rationalized as combining prior art elements according to known methods to yield predictable results.
Poole and Han disclose determining a midline for dividing the midline centered data corresponding to a brain anatomical structure into symmetrical left and right parts.
Hemingway teaches detecting a symmetry plane that divides an imaging volume of a brain that is substantially symmetrical between the left and right sides into first and second sections, where the symmetry plane divides the head into a left half comprising a left hemisphere of the brain and a right half comprising a right hemisphere of the brain.
One of ordinary skill in the art could have combined the disclosed teachings of Poole, Han, and Hemingway, where dividing the midline centered data set corresponding to the brain anatomical structure into symmetrical left and right parts with a determined midline would indicate that the left part includes the left hemisphere of the brain and the right part includes the right hemisphere of the brain, and predictably result in separating the midline centered data set corresponding to the brain into symmetrical left and right parts.
Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over Poole and Han as applied to claim 12 above, and further in view of Yu et al. (US 2020/0334869), herein Yu.
Regarding claim 13, please see the above rejection of claim 12. Poole and Han do not explicitly disclose the method according to claim 12, wherein
the medical imaging data includes 2D-projections from different views onto the region of the radiological finding; and the method further comprises:
determining a coarse location of the region of the radiological finding by applying a back-projection on the 2D- projections.
Yu teaches in a related and pertinent systems and methods for reconstructing an image in an imaging system (see Yu Abstract), where the imaging device may generate or provide imaging data via scanning a subject, and can be a CT scanner (see Yu [0036]), where scan data representing an intensity distribution of energy detected at a plurality of detector elements is obtained from an imaging system performing a scan on a subject, the subject may include a specific portion of a body, such as a head, and the scan may be generated based on multiple projections at different angles around the subject using radiation beams (see Yu [0059]), where an image estimate is determined according to an image reconstruction algorithm, such as a filtered back projection algorithm (see Yu [0060]), and further iteratively update the image estimate according to an objective function to output a final image (see Yu [0066]-[0068]), and that final images may be used to reconstruct a 3D image based on a 3D reconstruction method, the 3D reconstruction method may include a multi-planar reconstruction (MPR) algorithm (see Yu [0056]).
At the time of filing, one of ordinary skill in the art would have found it obvious to apply the teachings of Yu to the teachings of Poole and Han, such that the volumetric imaging data obtained from the CT scanner is generated based on multiple projections at different angles around the patient using radiation beams, and that the detected and highlighted abnormal regions on the rendered MPR view are generated based on filtered back projection image reconstruction algorithm.
This modification is rationalized as an application of a known technique to a known method ready for improvement to yield predictable results.
In this instance, Poole and Han disclose a base method for generating training data and training a detector to detect abnormal voxels in medical volumetric imaging data obtained from a CT scanner, where the imaging data is representative of a symmetrical brain anatomical structure, and that detected abnormal regions are displayed to a clinician, where a list of detections is displayed that is ranked by probability, and clicking on a listing of an abnormal region navigates to an MPR (multi-planar reformatting) view of the abnormal region, and each detected abnormal region is highlighted by colour in the MPR view (see Poole [00156]).
Yu teaches known techniques for reconstructing an image in an imaging system, where the imaging device may generate or provide imaging data via scanning a subject with a CT scanner, where the scan data representing an intensity distribution of energy detected at a plurality of detector elements is obtained from an imaging system performing a scan on a subject, the subject may include a specific portion of a body, such as a head, and the scan may be generated based on multiple projections at different angles around the subject using radiation beams, and an image estimate is determined according to an image reconstruction algorithm, such as a filtered back projection algorithm, and further iteratively update the image estimate according to an objective function to output a final image, and that final images may be used to reconstruct a 3D image based on a 3D reconstruction method, the 3D reconstruction method may include a multi-planar reconstruction (MPR) algorithm.
One of ordinary skill in the art would have recognized that by applying Yu’s techniques would allow for the method of Poole and Han to generate the volumetric imaging data obtained from the CT scanner based on multiple projections at different angles around the patient using radiation beams, and that the detected and highlighted abnormal regions on the rendered MPR view are generated based on filtered back projection image reconstruction algorithm, predictably leading to an improved method for generating volumetric imaging data and reconstructing images of the abnormal regions from obtained CT scans as taught by Yu.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TIMOTHY WING HO CHOI whose telephone number is (571)270-3814. The examiner can normally be reached 9:00 AM to 5:00 PM.
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, VINCENT RUDOLPH can be reached at (571) 272-8243. 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.
/TIMOTHY CHOI/Examiner, Art Unit 2671
/VINCENT RUDOLPH/Supervisory Patent Examiner, Art Unit 2671
