DETAILED ACTION
Acknowledgements
This office action is in response to the claims filed May 04, 2026.
Claims 1-23 are pending
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(s)
Claims 1-23 are pending.
Claim Rejection - 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 1-23 are rejected to under 35 U.S.C 101 as not being directed to eligible subject matter the grounds set out in detail below:
Independent Claims 1, 19, and 20:
Eligibility Step 1 (does the subject matter fall within a statutory category?): Independent Claims 1, 19, and 20 falls within the statutory category of method.
Eligibility Step 2A-1 (does the claim recite an abstract idea, law of nature, or natural phenomenon?): Independent claims 1, 19, and 20 claimed invention is directed to a judicial exception.
The claim elements in the independent claim 1 which set forth the abstract idea are:
A method for generating a prediction of a future health-related event associated with an individual, the method comprising:
identifying health-related data associated with the individual, the health-related data comprising a sequence of health-related events, wherein each health-related event is associated with one of a plurality of event types;;
generating, for each health-related event in the sequence of health-related events, a respective embedded representation of the health-related event by processing the health-related event using an embedding function corresponding to the respective event type to generate one or more embeddings in a shared embedding
generating, …[…]…and conditioned on at least the respective embedded representations of the health-related events, an output embedded representation of the future health-related event;
and processing at least the output embedded representation of the future health-related event…[…]…to generate the prediction of the future health-related event wherein the future health-related event is associated with one of the plurality of event types..
The claim elements in the independent claims 19 which set forth the abstract idea are:
A method for generating a prediction of one or more missing health-related events associated with an individual, the method comprising:
identifying incomplete health-related data associated with the individual, the incomplete health-related data comprising a sequence of health-related events with one or more gaps to be fille, wherein the gap is in between two respective health-related events in the sequence;
updating the sequence of health-related events by applying a mask for each of the one or more gaps by including a mask token in between the two respective health-related events in the sequence;
generating, for each health-related event or mask token in the sequence of health-related events, a respective embedded representation of the health-related event or the mask token;
generating, for each mask token in the sequence of health-related events, …[…]…and conditioned on at least the respective embedded representations of the health-related events, an output embedded representation of a missing health-related event;
and processing at least the output embedded representation of each missing health-related event …[…]…to generate the prediction of the one or more missing health-related events.
The claim elements in the independent claims 20 which set forth the abstract idea are:
A method for detecting one or more missing health-related events associated with an individual, the method comprising:
identifying health-related data associated with the individual at a first time, the health-related data comprising a sequence of health-related events;
generating a sequence of predictions of future health-related events that are likely to occur within a window of time after the first time;
identifying updated health-related data associated with the individual at a second time, updated health-related data comprising a sequence of health-related events that have occurred between the first time and the second time;
and processing the sequence of predictions of future health-related events and the updated health-related data to identify one or more health-related events in the sequence of future health-related events as missing health-related events, wherein the missing health-related events do not have a match within the updated health-related data and providing data representing the missing health-related events for presentation.
which falls within “certain methods of organizing human activity” as following rules or instructions to predict a future and/or missing health related event based on data. See MPEP § 2106.04(a)(2).
Eligibility Step 2A-2 (does the claim recite additional elements that integrate the judicial exception into a practical application?): For Independent Claim 1, 19, and 20 this judicial exception is not integrated into a practical application.
In Claim 1, 19, and 20 the additional elements are:
one or more de-embedding machine learning models
a sequence processing neural network
a user device
Examiner takes the applicable considerations stated in MPEP 2106.04 (d) and analyzes them below in light of the instant applications disclosure and claim elements as a whole.
The additional element, (a) and (b) are merely applying the abstract idea as “apply-it” or an equivalent (e.g. using) to analyze data
The additional element, (c) is merely applying the abstract ideas as “apply-it” or an equivalent to output data
Accordingly, claims 1, 19, and 20 does not integrate the abstract idea into a practical application.
Eligibility Step 2B (Does the claim amount to significantly more?): The independent claims 1, 19, and 20 do not include additional elements that are sufficient to amount to significantly more than the judicial exception because as analyzed above in step 2A prong 2 above, these additional elements, whether viewed individually or as an ordered combination, amount to no more than applying the abstract idea and thus insufficient to provide “significantly more”. Therefore, the claims do not amount to significantly more and the claims are ineligible.
Dependent Claims 2-18 and 21-23:
Eligibility Step 1 (does the subject matter fall within a statutory category?):The dependent claims 2-18 and 21-23 fall within the statutory category of method.
Eligibility Step 2A-1 (does the claim recite an abstract idea, law of nature, or natural phenomenon?): Dependent claims 2-18 and 21-23 claimed invention are directed to a judicial exception.
Dependent claims 2-18 and 21-23 continue to limit the abstract idea in the independent claim by (1) types of health related data (2) further limiting the rules to analyze the data thus, inheriting the same abstract idea which falls within “certain methods of organizing human activity” as following rules or instructions to predict a future and/or missing health related event based on data. See MPEP § 2106.04(a)(2).
Eligibility Step 2A-2 (does the claim recite additional elements that integrate the judicial exception into a practical application?): In Claims 2-18 and 21-23 this judicial exception is not integrated into a practical application.
In Claims 2-18 and 21-23 the additional elements not already recited in the independent claims are:
an embedding neural network
a generative neural network
Examiner takes the applicable considerations stated in MPEP 2106.04 (d) and analyzes them below in light of the instant applications disclosure and claim elements as a whole.
The additional element, (a) and (b) are merely applying the abstract idea as “apply-it” or an equivalent (e.g. using) to analyze data
Eligibility Step 2B (Does the claim amount to significantly more?): Dependent claims 2-18 and 21-23, do not include additional elements that are sufficient to amount to significantly more than the judicial exception because as analyzed above in step 2A prong 2 above, these additional elements, whether viewed individually or as an ordered combination, amount to no more than apply it thus insufficient to provide “significantly more”. Therefore, the claims do not amount to significantly more and the claims are ineligible.
Claim Rejections - 35 USC § 102
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 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 1-8 and 11-17 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Tomasev et. al (hereinafter Tomasev) (US11302446B2)
As per claim 1, Tomasev teaches:
A method for generating a prediction of a future health-related event associated with an individual, the method comprising: (abstract discloses, “Methods, systems, and apparatus, including computer programs encoded on computer storage media, for predicting future adverse health events using neural networks” and see Col. 1 lines 33-35 discloses, “This specification describes a system that makes predictions that characterize the likelihood that a specific adverse health event will occur to a patient in the future.”)
identifying health-related data associated with the individual, the health-related data comprising a sequence of health-related events, wherein each health-related event is associated with one of a plurality of event types; (Col. 1 lines 50-66 discloses, “Thus in one aspect there is provided a system comprising one or more computers and one or more storage devices storing instructions that, when executed by the one or more computers, cause the one or more computers to perform operations comprising receiving electronic health record data for a patient, the electronic health data comprising a plurality of features representing health events in an electronic health record for the patient, each of the plurality of features belonging to a vocabulary of possible features that comprises a plurality of possible numerical features and a plurality of possible discrete features. The operations may further comprise generating, from the electronic health record data, an input sequence comprising a respective feature representation at each of a plurality of time steps, wherein the plurality of time steps comprises a respective time window time step for each of a plurality (succession) of time windows.” And see Col. 14 lines 23-29 discloses, “For example, the system can map each feature to a corresponding high-level concept, e.g., procedure, diagnosis prescription, laboratory test, vital sign, admission, transfer and so on. The system can then include in the feature representation at each time step a histogram of frequencies of each high-level concept among the features that occurred at the time step.” And see The deep embedding neural network 240 includes multiple fully-connected layers and is configured to embed the features in the feature representation in an embedding space. In other words, the deep embedding neural network 240 maps the feature representation into an ordered collection of numeric values, e.g., a vector, in an embedding space that has a fixed dimensionality. By doing so, the embedding neural network 240 transforms the high-dimensional and sparse input feature representation into a lower-dimensional continuous representation that makes subsequent prediction easier. In some implementations, there are residual connections between the fully-connected layers.” / examiner notes event types is interpreted as defined in the instant application disclosure e.g. page 7 lines 6-9 as labs, EHR data, procedures etc.)
generating, for each health-related event in the sequence of health-related events, a respective embedded representation of the health-related event by processing the health-related event using an embedding function corresponding to the respective event type to generate one or more embeddings in a shared embedding space; (Col. 4 lines 4-10 discloses, “In implementations the neural network comprises a deep embedding neural network to embed the features in the feature representation in an embedding space. The neural network output may then be generated from the embedded features. For example the deep embedding neural network may comprise a plurality of fully-connected layers (though other architectures may be used).” And see Col. 3 lines 46-67 and see Col. 4 lines 17-27 and see Col. 10 lines 19-46 / examiner notes as previously cited the features are a sequence of health related events and the disclosure is utilizing a temporal convolutional embedding space which is shared learned representation space as each sequence is processed using this same space and the embedding neural networks are representing feature vector from high to low dimensional space which one of ordinary skill would understand is using embedding math functions to do this semantically as disclosed. )
generating, using a sequence processing neural network and conditioned on at least the respective embedded representations of the health-related events, an output embedded representation of the future health-related event; (Col. 4 lines 4-10 discloses, “In implementations the neural network comprises a deep embedding neural network to embed the features in the feature representation in an embedding space. The neural network output may then be generated from the embedded features. For example the deep embedding neural network may comprise a plurality of fully-connected layers (though other architectures may be used).” And see Col. 7 lines In particular, the system 100 receives electronic health record data 102 for a patient, generates an input sequence 122 from the electronic health record data 102, and then processes the input sequence 122 using a neural network 110 to generate a neural network output 132. For example, the neural network output 132 can include a score, e.g., a probability, that characterizes a predicted likelihood that the adverse health event will occur to the patient within a fixed time period after the last time window in the electronic health record data 102. As another example, the neural network output 132 can include scores for multiple time periods, each starting after the last time window in the health record data 102 and each being a different length. The score for each time period can be a score, e.g., a probability, that characterizes the predicted likelihood that the adverse health event will occur to the patient within the corresponding time period after the last time window in the health record data 102.” And see Col 4 lines 20-23 discloses, “For example in some implementations the neural network may comprise a recurrent neural network e.g., with a plurality of recurrent neural network (RNN) layers and, optionally, highway connections.”)
and processing at least the output embedded representation of the future health-related event using one or more de-embedding machine learning models to generate the prediction of the future health-related event wherein the future health-related event is associated with one of the plurality of event types. (Col. 4 lines 4-10 discloses, “In implementations the neural network comprises a deep embedding neural network to embed the features in the feature representation in an embedding space. The neural network output may then be generated from the embedded features. For example the deep embedding neural network may comprise a plurality of fully-connected layers (though other architectures may be used).” And see Col. 7 lines In particular, the system 100 receives electronic health record data 102 for a patient, generates an input sequence 122 from the electronic health record data 102, and then processes the input sequence 122 using a neural network 110 to generate a neural network output 132. For example, the neural network output 132 can include a score, e.g., a probability, that characterizes a predicted likelihood that the adverse health event will occur to the patient within a fixed time period after the last time window in the electronic health record data 102. As another example, the neural network output 132 can include scores for multiple time periods, each starting after the last time window in the health record data 102 and each being a different length. The score for each time period can be a score, e.g., a probability, that characterizes the predicted likelihood that the adverse health event will occur to the patient within the corresponding time period after the last time window in the health record data 102.” And see Col 4 lines 20-23 discloses, “For example in some implementations the neural network may comprise a recurrent neural network e.g., with a plurality of recurrent neural network (RNN) layers and, optionally, highway connections.” And see Col. 10 lines 47-56 discloses, the lower dimensionality steps and see Col. 7 lines 57-65 and Col. 6 lines 59-62 discloses that the future event is associated with the EHR feature representation events)
As per claim 2, Tomasev teaches:
The method of claim 1, wherein the health-related data comprises time information for each health-related event in the sequence of health-related events. (Col. 3 lines 23-45 discloses, “A further characteristic of EHR data which presents a challenge for neural network processing is that for a large proportion of the features there may be no very exact time stamp…[…]…Thus in implementations the plurality of time steps may include one or more surrogate time steps, each associated with a plurality of time window time steps that immediately precede the surrogate time step in the input sequence. For example there may be a surrogate time step at the conclusion of each day. Then generating the feature representation may comprise, for each of the surrogate time steps, determining whether the EHR data identifies any features (i) as occurring during a time interval spanned by the time windows corresponding to the plurality of time window time steps associated with the surrogate time steps without (ii) identifying a specific time window during which the feature occurred, and if so generating the feature representation for the surrogate time step from such features. In implementations the neural network processes such features at a current time step but the neural network output is not used for predicting an adverse health event until at least the next time step.”)
As per claim 3, Tomasev teaches:
The method of claim 1, wherein generating an output embedded representation of the future health-related event comprises: processing a sequence of the respective embedded representations for the health-related events, wherein the respective embedded representations for the health-related events are ordered within the sequence according to a time at which each health-related event occurred. (Col. 7 lines57-60 discloses, “In the example of FIG. 1 and for ease of description, the electronic health record data 102 is depicted as a sequential representation of health events, with events being ordered by the time that the events occurred and represented by circles.” )
As per claim 4, Tomasev teaches:
The method of claim 1, wherein the shared embedding space is shared across the plurality of event types. (Col. 4 lines 4-10 discloses, “In implementations the neural network comprises a deep embedding neural network to embed the features in the feature representation in an embedding space. The neural network output may then be generated from the embedded features. For example the deep embedding neural network may comprise a plurality of fully-connected layers (though other architectures may be used).” And see Col. 3 lines 46-67 and see Col. 4 lines 17-27 and see Col. 14 lines 23-29 discloses, “For example, the system can map each feature to a corresponding high-level concept, e.g., procedure, diagnosis prescription, laboratory test, vital sign, admission, transfer and so on. The system can then include in the feature representation at each time step a histogram of frequencies of each high-level concept among the features that occurred at the time step.” And see The deep embedding neural network 240 includes multiple fully-connected layers and is configured to embed the features in the feature representation in an embedding space. In other words, the deep embedding neural network 240 maps the feature representation into an ordered collection of numeric values, e.g., a vector, in an embedding space that has a fixed dimensionality. By doing so, the embedding neural network 240 transforms the high-dimensional and sparse input feature representation into a lower-dimensional continuous representation that makes subsequent prediction easier. In some implementations, there are residual connections between the fully-connected layers.” / examiner notes event types is interpreted as defined in the instant application disclosure e.g. page 7 lines 6-9 as labs, EHR data, procedures etc. and the features are a sequence of health related events and the disclosure is utilizing a temporal convolutional embedding space which is shared learned representation space as each sequence is processed using this same space )
As per claim 5, Tomasev teaches:
The method of claim 4, wherein the respective embedded representations for two or more of the plurality of event types comprise a different number of embeddings in the shared embedding space. (Col. 10 lines 12-56 and see Col. 14 lines 23-29 discloses, fixed dimensionality for values or vectors of features which have are a plurality of event types as previously cited)
As per claim 6, Tomasev teaches:
The method of claim 1, wherein generating, for each health-related event in the sequence of health-related events, a respective embedded representation of the health-related event comprises: for each health-related event in the sequence of health-related events: determining a respective event type for the health-related event; and processing the health-related event using an embedding function corresponding to the respective event type to generate the respective embedded representation of the health-related event. (see Col. 14 lines 23-29 and see Col. 10 and Col. 12-14 discloses, the use of neural network embedding functions and layers to take event types utilizing embedding and embedding space to determine scores and representations of the health related event for the future. )
As per claim 7, Tomasev teaches:
The method of claim 6, wherein for one or more of the respective event types, the corresponding embedding function is a learned function. (Col. 4 lines 23-25 discloses, “In other implementations the neural network may comprise a temporal convolutional neural network.” And see Col. 10 lines 19-22 discloses, “The neural network 110 includes a deep embedding neural network 240, a deep recurrent neural network 250, a set of main output layers 260, and optionally a set of auxiliary output layers 270.” / instant application defines the embedding neural network as being an e.g. CNN with layers see page 27 lines 20-25 instant application spec. )
As per claim 8, Tomasev teaches:
The method of claim 7, wherein the learned function is an embedding neural network. (Col. 4 lines 23-25 discloses, “In other implementations the neural network may comprise a temporal convolutional neural network.” And see Col. 10 lines 19-22 discloses, “The neural network 110 includes a deep embedding neural network 240, a deep recurrent neural network 250, a set of main output layers 260, and optionally a set of auxiliary output layers 270.”)
As per claim 11, Tomasev teaches:
The method of claim 6, further comprising, for each health-related event in the sequence of health-related events: obtaining a corresponding event type encoding for the respective event type characterizing the respective event type; and updating the respective embedded representation using the corresponding event type encoding. (Col. 2 lines 37-48 discloses, “In implementations, therefore, presence features are generated which enable the neural network to distinguish between the absence of a numerical feature (value) and an actual value of zero. Put differently, a presence feature may be considered to capture a feature associated with an act of making a measurement, whatever the outcome. Thus, for example, a presence feature may be a binary feature. The presence features may also encode discrete features such as the implementation of diagnostic or other medical procedure codes. This approach facilitates better use of EHR data by a neural network for predicting the likelihood of an adverse health event.” And see Col. 10 lines 29-35 and Col. 13 lines 8-23 / examiner notes the disclosure teaches numerical features or values also taught as vectors encoded. The event types and encoding is interpreted as it is defined in the instant application specification page 28 lines 26-30 as feature vectors)
As per claim 12, Tomasev teaches:
The method of claim 6, further comprising, for each health-related event in the sequence of health-related events: updating the respective embedded representation by applying a corresponding learned transformation for the respective event type to the respective embedded representation. (Col. 12 lines 16-26 discloses, minimizing a loss function on a trained multi layer neural network such as cross entropy)
As per claim 13, Tomasev teaches:
The method of claim 1, wherein the sequence processing neural network has been trained to minimize a loss function that measures a distance between output embedded representations of the future health-related event generated by the sequence processing neural network for training samples, and target output embedded representations for the training samples. (Col. 12 lines 16-26 discloses, minimizing a loss function on a trained neural network such as cross entropy)
As per claim 14, Tomasev teaches:
The method of claim 1, wherein each of the one or more de-embedding machine learning models corresponds to a respective event type of the plurality of event types, and wherein processing at least the output embedded representation of the future health-related event using one or more de-embedding machine learning models to generate the prediction of the future health-related event comprises: obtaining a target event type; and processing the output embedded representation of the future health-related event using the de-embedding machine learning model corresponding to the target event type to generate the prediction of the future health-related event. (Col. 9 lines 62-67 and Col. 10 lines 1-67 and see Col. 11 lines 1-30 discloses the use of an RNN for sequential time based event data into vectors and de-embedding into scores for prediction of future health related event in which the historical event data is taken into account when generating predictions output by the RNN using embedded representations of predicted future health related events and see Col. 11 and Col. 12 and see Col. 14 lines 23-29 / examiner notes the target prediction is interpreted as the target event type as the event type is broadly defined in the instant application specification e.g. see page 7 of instant spec. and the cited prior art defines the event type as being labs, prescription diagnosis, etc. as the features analyzed to make target future event types )
As per claim 15, Tomasev teaches:
The method of claim 14, wherein each of the one or more de-embedding machine learning models corresponding to a respective event type has been trained to map an embedded representation of a health-related event to a prediction of the health-related event (Col. 9 lines 62-67 and Col. 10 lines 1-67 and see Col. 11 lines 1-30 discloses the use of an RNN for sequential time based event data into vectors and de-embedding into scores for prediction of future health related event in which the historical event data is taken into account when generating predictions output by the RNN using embedded representations of predicted future health related events and see Col. 11 and Col. 12 and see Col. 14 lines 23-29)
As per claim 16, Tomasev teaches:
The method of claim 1, wherein each of the one or more de-embedding machine learning models corresponds to a respective event type of the plurality of event types, and wherein processing at least the output embedded representation of the future health-related event using one or more de-embedding machine learning models to generate the prediction of the future health-related event comprises: processing the output embedded representation using a learned classifier function to generate a distribution over a plurality of event types; selecting a particular event type based on the distribution; and processing the output embedded representation using the de-embedding machine learning model corresponding to the particular event type to generate the prediction of the future health-related event. (Col. 9 lines 62-67 and Col. 10 lines 1-67 and see Col. 11 lines 1-30 discloses the use of an RNN for sequential time based event data into vectors and de-embedding into scores for prediction of future health related event in which the historical event data is taken into account when generating predictions output by the RNN using embedded representations of predicted future health related events and see Col. 11 and Col. 12 and see Col. 14 lines 23-29 / examiner notes the target prediction is interpreted as the target event type as the event type is broadly defined in the instant application specification e.g. see page 7 of instant spec. and the cited prior art defines the event type as being labs, prescription diagnosis, etc. as the features analyzed to make target future event type and using the distribution of probabilities and assigned scores of features representing event types for prediction )
As per claim 17, Tomasev teaches:
The method of claim 1, wherein the one or more de-embedding machine learning models comprise a generative neural network, and wherein processing at least the output embedded representation of the future health-related event using one or more de-embedding machine learning models to generate the prediction of the future health-related event comprises: processing the output embedded representation of the future health-related event using the generative neural network to generate the prediction of the future health-related event, wherein the generative neural network has been trained to generate a prediction of a future health-related event conditioned on the output embedded representation of the future health-related event. (Col. 9 lines 62-67 and Col. 10 lines 1-67 and see Col. 11 lines 1-30 discloses the use of an RNN for sequential time based event data into vectors and de-embedding into scores for prediction of future health related event in which the historical event data is taken into account when generating predictions output by the RNN using embedded representations of predicted future health related events)
As per claim 20, Tomasev teaches:
A method for detecting one or more missing health-related events associated with an individual, the method comprising: (abstract discloses, “Methods, systems, and apparatus, including computer programs encoded on computer storage media, for predicting future adverse health events using neural networks” and see Col. 1 lines 33-35 discloses, “This specification describes a system that makes predictions that characterize the likelihood that a specific adverse health event will occur to a patient in the future.” / examiner notes page 2 lines 11-13 of the instant application specification defines a future event which can be also a missing health related event)
identifying health-related data associated with the individual at a first time, the health-related data comprising a sequence of health-related events; (Col. 1 lines 50-66 discloses, “Thus in one aspect there is provided a system comprising one or more computers and one or more storage devices storing instructions that, when executed by the one or more computers, cause the one or more computers to perform operations comprising receiving electronic health record data for a patient, the electronic health data comprising a plurality of features representing health events in an electronic health record for the patient, each of the plurality of features belonging to a vocabulary of possible features that comprises a plurality of possible numerical features and a plurality of possible discrete features. The operations may further comprise generating, from the electronic health record data, an input sequence comprising a respective feature representation at each of a plurality of time steps, wherein the plurality of time steps comprises a respective time window time step for each of a plurality (succession) of time windows.”)
generating a sequence of predictions of future health-related events that are likely to occur within a window of time after the first time; (Col. 4 lines 28-42)
identifying updated health-related data associated with the individual at a second time, updated health-related data comprising a sequence of health-related events that have occurred between the first time and the second time; (Col. 13 lines 54-67 and Col. 14 lines 1-20)
and processing the sequence of predictions of future health-related events and the updated health-related data to identify one or more health-related events in the sequence of future health-related events as missing health-related events, herein the missing health-related events do not have a match within the updated health-related data and providing data representing the missing health-related events for presentation to a user device. (see Col. 9 lines 18-23 and see Col. 11 lines 47-67 and Col. 12 lines 1-7 / examiner notes that the interpretation under BRI is taken of a match within the health data as comparison to ground truth data as identifying if there is or is not a match to the updated health related future events and the claim construction states future health related events as missing health events)
Claim 19 is rejected under 35 U.S.C. 102(a)(2) as being anticipated by FOSCHINI et. al (hereinafter FOSCHINI) (US20250069750A1)
As per claim 19, FOSCHINI teaches:
A method for generating a prediction of one or more missing health-related events associated with an individual, the method comprising: (abstract discloses, “Disclosed is a method comprising accessing, by a machine learning system, a set of data records for a plurality of users, the data records representative of physical statistics measured for each of the plurality of users over a time period. At least a subset of the data records comprises patterns of missing data for at least a portion of the time period.”)
identifying incomplete health-related data associated with the individual, the incomplete health-related data comprising a sequence of health-related events with one or more gaps to be filled, wherein the gap is in between two respective health-related events in the sequence; [0083] In a second operation 720, the system may identify gaps in the collected wearable data likely caused by natural missingness (naturally-occurring patterns of missing wearable device data). These patterns may be present due to patterns of wearable device disuse or downtime that may occur over a duration of typical use by a subject. Upon determining these patterns, the system may mask one or more portions of data collected from a subject, creating gaps to make the data appear similar to subject data with patterns of natural missingness.
updating the sequence of health-related events by applying a mask for each of the one or more gaps by including a mask token in between the two respective health-related events in the sequence; ([0076] In a second operation 520, the self-supervised learning system may generate a training data set by using patterns of missing information (i.e., "missingness") in some of the data records of the set to mask other records of the set. For example, in some embodiments, the self-supervised learning system applies a missingness of each data record of the set to a next data record of the set, and repeats the process over one or more iterations to generate the training set. In other embodiments, the self-supervised learning system identifies pairs of data records of the set based a level of similarity and/or a level of overlap in missingness, and applies the missingness of a first data record of the pair to a second data record of the pair to generate the training set. And see [0120] discloses, “Embodiments of the disclosure may generate synthetic data by placing synthetic data values in gaps within a collected sequence of wearable sensor data (e.g., at places where data is missing or has been masked), while removing non-synthetic wearable sensor data values. This may be performed until the entire wearable dataset comprises synthetic data.” And see [0102] discloses, “The encoder sub-system 1070 may generate representations from the data. The encoder sub-system may comprise one or more machine learning algorithms. In some embodiments, one or more of the machine learning algorithms comprises a neural network (or artificial neural network (ANN)). A neural network may be a convolutional neural network (CNN) or recurrent neural network (RNN). A neural network may be a multilayer perceptron (MLP).” / examiner notes synthetic data in gaps is a type of data imputation where the mask token is the missing data)
generating, for each health-related event or mask token in the sequence of health-related events, a respective embedded representation of the health-related event or the mask token; ([0005] discloses, “The method comprises providing a set of time series wearable sensor data. The method also comprises generating a plurality of embeddings from the time series wearable data”)
generating, for each mask token in the sequence of health-related events, using a sequence processing neural network and conditioned on at least the respective embedded representations of the health-related events, an output embedded representation of a missing health-related event; ([0120] discloses, “Embodiments of the disclosure may generate synthetic data by placing synthetic data values in gaps within a collected sequence of wearable sensor data (e.g., at places where data is missing or has been masked), while removing non-synthetic wearable sensor data values. This may be performed until the entire wearable dataset comprises synthetic data.” And see [0102] discloses, “The encoder sub-system 1070 may generate representations from the data. The encoder sub-system may comprise one or more machine learning algorithms. In some embodiments, one or more of the machine learning algorithms comprises a neural network (or artificial neural network (ANN)). A neural network may be a convolutional neural network (CNN) or recurrent neural network (RNN). A neural network may be a multilayer perceptron (MLP).” / examiner notes synthetic data in gaps is a type of data imputation where the mask token is the missing data)
and processing at least the output embedded representation of each missing health-related event using one or more de-embedding machine learning models to generate the prediction of the one or more missing health-related events. ( [0043] Each physical statistic monitored by the self-supervised learning system 110 can be affected based on the behavior of a user 120 (e.g., whether the user is exercising, is asleep, etc.) and/or a health condition of the user 120 (e.g., whether the user is exhibiting normal health, has the flu, is suffering from allergies, etc.). As such, by analyzing the physical statistics monitored for a given user, predictions can be made relating to user behavior and/or user health condition. For example, in some embodiments, the physical statistic data module 210 can be used for training models to reflect different behaviors and/or health conditions, and to predict behaviors and/or health conditions for an individual user based on received physical statistic data for the user, in real time or near-real time. In some embodiments, the physical statistic data module 210 continuously receives physical statistic data from health sensors 125 or users and preprocesses the physical statistic data for evaluation in real time or near-real time (for example, for predicting the health condition of a user).” And see [0108] The head 1080 may process the representations 1080 to perform a downstream task. The head 1080 may comprise one or more machine learning algorithms configured to perform the downstream task. For example, the head 1080 may comprise one or more supervised and/or unsupervised machine learning algorithms The head 1080 may comprise, for example, support vector machines (SVM), a logistic regression, or a decision tree algorithm (e.g., gradient boosted trees, Adaboost, XGBoost, or random forests). The head may comprise one or more layers.” And see [0109] The head 1080 may comprise an activation function to produce a prediction output. The head may perform a regression task. The head may perform a classification task. The head may comprise a binary classifier. The head may comprise a multiclass classifier.”)
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 9-10 and 18 are rejected to under 35 U.S.C. 103 as being unpatentable over Tomasev et. al (hereinafter Tomasev) (US11302446B2) in view of Kang et. al (hereinafter Kang) (CN117316362A)
As per claim 9, Tomasev does not teach
The method of claim 8, wherein the embedding neural network is pre-trained and frozen prior to training the sequence processing neural network.
However, Kang teaches:
The method of claim 8, wherein the embedding neural network is pre-trained and frozen prior to training the sequence processing neural network. (page 3 para. 6 discloses, “Further, in S6, the input constructed based on the discharge medical order text features based on transfer learning is the patient discharge summary text, and the intermediate representation is obtained as the embedding of the text from the clinical text ICD coding model of the multi-filter residual convolutional neural network, and then Clinical text ICD encoding model pre-trained in multi-filter residual convolutional neural network using patient's unlabeled data, clinical text via frozen multi-filter residual convolutional neural network using discharge summary text of patient vt times The weight of the ICD encoding model is obtained. After the intermediate representation is obtained, it is passed through the SUM layer and used as the user's multi-view medical order text feature. It is transmitted to the output layer together with the module features from other views and then fed into the classifier for prediction.”)
It would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Tomasev’s teachings of a surgical system as previously cited with Kang’s teachings as previously cited, the motivation being Tomasev teaches the concern of overfitting and accuracy of modelling (e.g. see Col 1 and 10) thus it would be obvious to accelerate training, reduce computational costs, and prevent overfitting by combining with Kang with no unpredictable results to substitute one machine learning for Kang’s.
As per claim 10, Tomasev does not teach
The method of claim 8, wherein the embedding neural network is pre-trained, and wherein parameters for the embedding neural network are updated during training of the sequence processing neural network.
However, Kang does teach:
The method of claim 8, wherein the embedding neural network is pre-trained, and wherein parameters for the embedding neural network are updated during training of the sequence processing neural network. ((page 3 para. 6 discloses, “Further, in S6, the input constructed based on the discharge medical order text features based on transfer learning is the patient discharge summary text, and the intermediate representation is obtained as the embedding of the text from the clinical text ICD coding model of the multi-filter residual convolutional neural network, and then Clinical text ICD encoding model pre-trained in multi-filter residual convolutional neural network using patient's unlabeled data, clinical text via frozen multi-filter residual convolutional neural network using discharge summary text of patient vt times The weight of the ICD encoding model is obtained. After the intermediate representation is obtained, it is passed through the SUM layer and used as the user's multi-view medical order text feature. It is transmitted to the output layer together with the module features from other views and then fed into the classifier for prediction.” And see page 9 and 10 / examiner under BRI interprets the multi filter residual convolutional neural network to update its parameters through optimization such as batch as taught in the cited prior art)
It would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Tomasev’s teachings of a surgical system as previously cited with Kang’s teachings for the same reasons given above for claim 9.
As per claim 18, Tomasev does not teach
The method of claim 1, wherein identifying health-related data associated with the individual comprises receiving data representing one or more health-related events of the sequence from a user.
However, Kang does teach:
The method of claim 1, wherein identifying health-related data associated with the individual comprises receiving data representing one or more health-related events of the sequence from a user. (page 9 para. 3 discloses, “Label=D={x1,x2,...,xn} represents the ICD code manually entered by the doctor for this visit”)
It would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Tomasev’s teachings of a surgical system as previously cited with Kang’s teachings for the same reasons given above for claim 9 as the claim recited choice data analyzed in the same manner.
Claims 21-23 are rejected to under 35 U.S.C. 103 as being unpatentable over Tomasev et. al (hereinafter Tomasev) (US11302446B2) in view of Saripalli et. al (hereinafter Saripalli) (US11404145B2)
As per claim 21, Tomasev does not teach:
The method of claim 1, wherein clinically similar health-related events are located close to each other in the shared embedding space
However, Saripalli does teach:
The method of claim 1, wherein clinically similar health-related events are located close to each other in the shared embedding space. (Col. 17 lines 40-61 discloses, “For example, the K-means/medoids 732 can cluster the data 710, 720 according to certain similarity. The Gaussian mixture model (GMM) 734 can be used to cluster the data 710, 720 according to arbitrary cluster shapes determined by Gaussian parameters of the distribution of the data 710, 720, for example. Other models, such as DBSCAN, etc., can be used to aggregate the data 710, 720. One or more models 732-734 can be selected or activated to the process the data 710, 720. In certain examples, the recommender system 736 predicts a ranking, rating, or preference associated with the data 710, 720, output of a model 732, 734, etc. The recommender system 736 processes the input data 710, 720 and output of the model 732, 734 to apply collaborative filtering and/or content-based filtering to generate a predictive or recommended output. For example, collaborative filtering builds a model based on past behavior as well as similar 55 decisions made by other users. This model is then used to predict items ( or ratings for items) that the user may have an interest in. The predicted output provides one or more samples of interest 740 for further processing to predict and/or categorize an event and/or classify a patient based on 60 the event( s ), etc.”)
It would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Tomasev’s teachings as previously cited with Saripalli’s teachings as previously cited, the motivation being Tomasev teaches the concern of efficiently using neural networks utilizing embedding spaces (e.g. see Col 2), thus it would be obvious to increase computational efficiencies and improve training by clustering similar features together, with no unpredictable results.
As per claim 22, Tomasev teaches:
The method of claim 1, wherein the plurality of event types comprises two or more of: symptoms, tests, labs, test results, lab results, diagnoses, medications, prescriptions, referrals, actions, procedures, outcomes, patient information records, health-related images, health-related electronic documents, genomic data, sensor data, prescription fulfillment data, or air quality data. (see Col. 14 lines 23-29 discloses, “For example, the system can map each feature to a corresponding high-level concept, e.g., procedure, diagnosis prescription, laboratory test, vital sign, admission, transfer and so on. The system can then include in the feature representation at each time step a histogram of frequencies of each high-level concept among the features that occurred at the time step.”)
As per claim 23, Tomasev does not teach:
The method of claim 22, wherein the plurality of event types comprises an image, and wherein the image is an X-ray image or diagnostic image.
However, Saripalli does teach:
The method of claim 22, wherein the plurality of event types comprises an image, and wherein the image is an X-ray image or diagnostic image. (Col. 4 lines 39-61 discloses, “Medical data can be obtained from imaging devices, sensors, laboratory tests, and/or other data sources. Alone or in combination, medical data can assist in diagnosing a patient, treating a patient, forming a profile for a patient population, influencing a clinical protocol, etc. However, to be useful, medical data must be organized properly for analysis and correlation beyond a human's ability to track and reason. Computers and associated software and data constructs can be implemented to transform disparate medical data into actionable results. For example, imaging devices (e.g., gamma camera, positron emission tomography (PET) scanner, computed tomography (CT) scanner, X-Ray machine, magnetic resonance (MR) imaging machine, ultrasound scanner, etc.) generate two-dimensional (2D) and/or three-dimensional (3D) medical images (e.g., native Digital Imaging and Communications in Medicine (DICOM) images) representative of the parts of the body (e.g., organs, tissues, etc.) to diagnose and/or treat diseases. Other devices such as electrocardiogram (ECG) systems, echoencephalograph (EEG), pulse oximetry (SpO2) sensors, blood pressure measuring cuffs, etc., provide one-dimensional waveform and/or time series data regarding a patient.”)
It would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Tomasev’s teachings as previously cited with Saripalli’s teachings as previously cited, the motivation being Tomasev teaches the concern of accurately predicting events (e.g. see Col 1), thus it would be obvious to increase modelling accuracy by including other choice data such as images which would improve training quality of data features, with no unpredictable results.
Response to Arguments Regarding 35 U.S.C § 101 Rejection
Applicant’s arguments on remarks pages 1-7 are as follows: Claims 1-20 were rejected under 35 U.S.C. § 101 as allegedly reciting non-patentable subject matter. Applicant respectfully disagrees with the rejection. In particular, Applicant respectfully submits that the subject matter of claim 1 solves a technical problem in the field of healthcare monitoring, by generating predictions of health events. As described in the Specification: A neural network can effectively be used to analyze a sequence of health events, e.g., a sequence of health events derived from an electronic medical record for a current patient. For example, the system described in this specification can generate one or more predicted future health-related events for a sequence of health events by generating a respective embedded representation of each health-related event. The system can generate an output embedded representation of a predicted future health-related event using a sequence processing neural network. The system can generate the predicted future health-related event using one or more de-embedding machine learning models. As another example, the system described in this specification can identify missing health-related events in the sequence of health events. For example, the system can generate a sequence of predicted future health-related events. The system can receive updated health-related data. The system can identify health-related events in the sequence of predicted future health-related events as missing health-related events if equivalents do not occur in the updated health-related data. [ page 1, line 29 - page 2, line 13 ] A doctor or other healthcare professional can be provided with information characterizing the output of the neural network or outputs derived from outputs generated by the neural network, improving the healthcare professional's ability to provide quality healthcare to the professional's patients or to perform risk assessment. For example, the system can simulate the forward evolution of a medical case associated with a patient or in aggregate. For example, the system can simulate the forward evolution of a medical case associated with a patient by processing health-related data for the patient to generate one or more predicted future health-related events.
The system can simulate the forward evolution of multiple patients in aggregate by processing health-related data for the multiple patients to generate one or more predicted future health-related events for each of the multiple patients. [ page 2, lines 14-24 ] In some examples, the system can provide a user interface that allows a healthcare professional to provide a future window of time, and the system can provide predictions of future health events likely to occur within the window of time. For example, the system can process health-related data and any previously predicted future health-related events for the patient to generate one or more predicted future health-related events until one of the predicted future health- related events occurs after the window of time. Generating a prediction of a future health-related event can also aid in administrative reasoning, for healthcare quality and patient safety monitoring, or in population management. For example, the system can be used for healthcare monitoring to identify unlikely proposed health-related events in the context of the health data, such as commission events or suboptimal decisions. As another example, the system can be used for healthcare monitoring to identify missing events that are likely to have occurred but did not occur. [ page 3, lines 16-27 ] (emphasis added) Thus, as described above, the claimed invention solves a technical problem in the technical field of healthcare monitoring and prediction by generating predictions of future health- related events in a way that produces structured clinical data. The claimed invention processes health-related events using embedding functions corresponding to their respective event types to map the health-related events into a shared embedding space, and uses a sequence processing neural network and one or more de-embedding machine learning models to generate a prediction of a future health-related event, where the future health-related event is associated with a respective event type. The system can thus provide for simulating the forward evolution of a medical case. Further, the claims recite the features that confer these advantages. For example, independent claim 1 recites: " generating, for each health-related event in the sequence of health- related events, a respective embedded representation of the health-related event, by processing the health-related event using an embedding function corresponding to the respective event type to generate one or more embeddings in a shared embedding space; generating, using a sequence processing neural network and conditioned on at least the respective embedded representations of the health- related events, an output embedded representation of the future health-related event; and processing at least the output embedded representation of the future health-related event using one or more de-embedding machine learning models to generate the prediction of the future health-related event, wherein the future health- related event is associated with one of the plurality of event types" (emphasis added) The claims thus recite a specific combination of steps for healthcare monitoring and prediction.
Applicant respectfully submits that the subject matter of claim 19 solves a technical problem in the field of healthcare monitoring, by identifying missing health-related events. As described in the Specification: In some examples, the system can provide a user interface that allows a healthcare professional to provide a future window of time, and the system can provide predictions of future health events likely to occur within the window of time. For example, the system can process health-related data and any previously predicted future health-related events for the patient to generate one or more predicted future health-related events until one of the predicted future health- related events occurs after the window of time. Generating a prediction of a future health-related event can also aid in administrative reasoning, for healthcare quality and patient safety monitoring, or in population management. For example, the system can be used for healthcare monitoring to identify unlikely proposed health-related events in the context of the health data, such as commission events or suboptimal decisions. As another example, the system can be used for healthcare monitoring to identify missing events that are likely to have occurred but did not occur. [ page 3, lines 16-27 ] Missing events can be difficult to infer or detect because practice patterns can vary, and healthcare systems often do not have fixed or consistent sampling intervals. The system can be used to infer missing events in a patient's medical record. For example, the system can identify a serialized medical history missing some health events, and reason about the missing health events. Missing events can include an omission event, which can include an action that was not proposed or taken but is predicted to be high probability. For example, the system can determine that a particular action, such as a thyroid stimulating hormone (TSH) blood test, has a high likelihood of having occurred in the medical history, but was not in the medical history. As another example, the system can determine that a potassium medication order is likely after a low blood potassium level result from a lab test, but the potassium medication order was not in the medical history. The system can thus prompt the healthcare professional to consider taking the particular action, reducing the likelihood of negative impacts on a patient's health due to omission events. [ page 4, lines 19-31 ] (emphasis added) Thus, as described above, the claimed invention solves a technical problem in the technical field of healthcare monitoring. For example, missing events can be difficult to infer or detect because healthcare systems often do not have fixed or consistent sampling intervals. The claimed invention addresses this issue by identifying incomplete health-related data that includes a sequence of health-related events with one or more gaps to be filled, updating the sequence of health-related events by replacing each gap with a mask token, and generating a prediction of the missing health-related event for each gap. The system can thus generate a prediction about missing health events from a medical history, which can aid in healthcare quality and patient safety monitoring, or in population management. Further, the claims recite the features that confer these advantages. For example, independent claim 19 recites: " identifying incomplete health-related data associated with the individual, the incomplete health-related data comprising a sequence of health-related events with one or more gaps to be filled, wherein the gap is in between two respective health-related events in the sequence; updating the sequence of health-related events by applying a mask for each of the one or more gaps by including a mask token in between the two respective health-related events in the sequence; generating, for each mask token in the sequence of health-related events, using a sequence processing neural network and conditioned on at least the respective embedded representations of the health-related events, an output embedded representation of a missing health-related event; and processing at least the output embedded representation of each missing health-related event using one or more de-embedding machine learning models to generate the prediction of the one or more missing health-related events" (emphasis added) Claim 19 thus recites recite a specific combination of steps for healthcare monitoring.
Applicant respectfully submits that the subject matter of claim 20 solves a technical problem in the field of healthcare monitoring, by identifying missing health-related events. As described in the Specification: As another example, the system described in this specification can identify missing health-related events in the sequence of health events. For example, the system can generate a sequence of predicted future health-related events. The system can receive updated health-related data. The system can identify health-related events in the sequence of predicted future health-related events as missing health-related events if equivalents do not occur in the updated health-related data. [ page 2, lines 8 - 13 ] In some examples, the system can provide a user interface that allows a healthcare professional to provide a future window of time, and the system can provide predictions of future health events likely to occur within the window of time. For example, the system can process health-related data and any previously predicted future health-related events for the patient to generate one or more predicted future health-related events until one of the predicted future health-related events occurs after the window of time. Generating a prediction of a future health- related event can also aid in administrative reasoning, for healthcare quality and patient safety monitoring, or in population management. For example, the system can be used for healthcare monitoring to identify unlikely proposed health-related events in the context of the health data, such as commission events or suboptimal decisions. As another example, the system can be used for healthcare monitoring to identify missing events that are likely to have occurred but did not occur. [ page 3, lines 16-27 ] Missing events can be difficult to infer or detect because practice patterns can vary, and healthcare systems often do not have fixed or consistent sampling intervals. The system can be used to infer missing events in a patient's medical record. For example, the system can identify a serialized medical history missing some health events, and reason about the missing health events. Missing events can include an omission event, which can include an action that was not proposed or taken but is predicted to be high probability. For example, the system can determine that a particular action, such as a thyroid stimulating hormone (TSH) blood test, has a high likelihood of having occurred in the medical history, but was not in the medical history. As another example, the system can determine that a potassium medication order is likely after a low blood potassium level result from a lab test, but the potassium medication order was not in the medical history. The system can thus prompt the healthcare professional to consider taking the particular action, reducing the likelihood of negative impacts on a patient's health due to omission events. [ page 4, lines 19-31 ] (emphasis added) Thus, as described above, the claimed invention solves a technical problem in the technical field of healthcare monitoring. For example, missing events can be difficult to infer or detect because healthcare systems often do not have fixed or consistent sampling intervals. The claimed invention addresses this issue by generating a sequence of predictions of future health- related events and identifying predicted health-related events that do not have a corresponding match within the health-related events of updated health-related data as missing health-related events. The system can then provide data representing the missing health-related events for presentation on a user device. The system can thus reduce the likelihood of negative impacts on a patient's health due to omission events. Further, the claims recite the features that confer these advantages. For example, independent claim 20 recites: " identifying health-related data associated with the individual at a first time, the health-related data comprising a sequence of health-related events; generating a sequence of predictions of future health-related events that are likely to occur within a window of time after the first time; identifying updated health-related data associated with the individual at a second time, updated health-related data comprising a sequence of health-related events that have occurred between the first time and the second time; [[and]] processing the sequence of predictions of future health-related events and the updated health-related data to identify one or more health-related events in the sequence of future health-related events as missing health-related events, wherein the missing health-related events do not have a match within the updated health- related data; and providing data derived from the missing health-related events for presentation" (emphasis added) The claims thus recite a specific combination of steps for healthcare monitoring.
Examiner appreciates applicants argument but does not find them persuasive. MPEP 2106.04(a)(2) (II) states, “Finally, the sub-groupings encompass both activity of a single person (for example, a person following a set of instructions or a person signing a contract online) and activity that involves multiple people (such as a commercial interaction), and thus, certain activity between a person and a computer (for example a method of anonymous loan shopping that a person conducts using a mobile phone) may fall within the “certain methods of organizing human activity” grouping. It is noted that the number of people involved in the activity is not dispositive as to whether a claim limitation falls within this grouping.” And see MPEP 2106.04(d)
Improving the healthcare professional's ability to provide quality healthcare to the professional's patients or to perform risk assessment and aid in healthcare quality and patient safety monitoring, or in population management is not reasonably understood to be a problem arising in technology but rather a problem medical education and diagnostics. The claimed invention is confined to a general purpose computer and are using the computer as a tool and any improvement present is an improvement to the abstract idea of predicting a future and/or missing health related event based on data within certain methods of organizing human activity. Further the claims do not recite or reflect a solution to the technical problem in the technical field of healthcare monitoring and prediction as the claims do not recite additional elements which provide an improvement to the technology of machine learning or monitoring or prediction as the claim is confined to the computer environment but rather provide improvement to the abstract idea of predicting future health related events through following rules and instructions for data review steps which a human could do to diagnose what may happen in the future or what may be missing events defined by applicant as events that did not occur but were likely to occur as this is what a doctor should do when diagnosing and prognosing a patient and following through with treatment care. If applicant’s line of reasoning were correct the invention in Alice Corp. would have been subject matter eligible because it was an improvement to the technology of settlement mitigation.
Response to Arguments 35 U.S.C. § 102 and 35 U.S.C. § 103
Applicant’s arguments on remarks pages 7-10 are as follows: Claims 1-8 and 11-17 were rejected under 35 U.S.C. § 102(a)(1) over Tomasev et al., U.S. Patent No. US 11,302,446 B2. Claims 9-10 and 18 were rejected under 35 U.S.C. § 103 over Tomasev in view of Kang et al., Chinese Patent No. CN 117316362 A. Claim 1 as amended recites, "generating, for each health-related event in the sequence of health-related events, a respective embedded representation of the health-related event, by processing the health-related event using an embedding function corresponding to the respective event type to generate one or more embeddings in a shared embedding space." The Action cited col. 14, lines 23-29 as disclosing a similar feature of claim 4. The cited portion of Tomasev reads as follows: For example, the system can map each feature to a corresponding high-level concept, e.g., procedure, diagnosis, prescription, laboratory test, vital sign, admission, transfer and so on. The system can then include in the feature representation at each time step a histogram of frequencies of each high-level concept among the features that occurred at the time step. [Tomasev, col. 14, lines 23-29 ] Applicant respectfully submits that the cited portion of Tomasev does not disclose or suggest "generating, for each health-related event in the sequence of health-related events, a respective embedded representation of the health-related event, by processing the health-related event using an embedding function corresponding to the respective event type to generate one or more embeddings in a shared embedding space." In particular, the cited portion of Tomasev does not disclose or suggest, for each health- related event, processing the health-related event using an embedding function corresponding to the respective event type associated with the health-related event. Tomasev describes "includ[ing] in the feature representation at each time step a histogram of frequencies of each high-level concept among the features that occurred at the time step." The Action has not shown that the cited portion of Tomasev discloses "for each health-related event in the sequence of health-related events", "processing the health-related event using an embedding function corresponding to the respective event type to generate one or more embeddings in a shared embedding space," as recited in amended claim 1. Thus, amended claim 1 and its dependent claims are patentable over Tomasev.
Claim 19 was rejected under 35 U.S.C. 102(a)(2) over Foschini et al., U.S. Patent No. US 20250069750 A1. Claim 19 as amended recites, "updating the sequence of health-related events by applying a mask for each of the one or more gaps by including a mask token in between the two respective health-related events in the sequence." The cited portion of Foschini reads as follows: In a second operation 520, the self-supervised learning system may generate a training data set by using patterns of missing information (i.e., "missingness") in some of the data records of the set to mask other records of the set. For example, in some embodiments, the self-supervised learning system applies a missingness of each data record of the set to a next data record of the set, and repeats the process over one or more iterations to generate the training set. In other embodiments, the self-supervised learning system identifies pairs of data records of the set based a level of similarity and/or a level of overlap in missingness, and applies the missingness of a first data record of the pair to a second data record of the pair to generate the training set. [Foschini, paragraph 0076 ] Applicant respectfully submits that the cited portion of Foschini does not disclose or suggest "updating the sequence of health-related events by applying a mask for each of the one or more gaps by including a mask token in between the two respective health-related events in the sequence." In particular, the cited portion of Foschini does not disclose or suggest including a mask token in between the two respective health-related events in the sequence. Foschini describes "using patterns of missing information (i.e., 'missingness') in some of the data records of the set to mask other records of the set" to replicate periods when wearable devices are inactive or not in use, to generate training data. Thus, Foschini removes portions of continuous time-series sensor data to create empty gaps within continuous data, rather than "updating the sequence of health- related events by applying a mask for each of the one or more gaps by including a mask token in between the two respective health-related events in the sequence," as recited in amended claim 19. Thus, amended claim 19 is patentable over Foschini.
Claim 20 was rejected under 35 U.S.C. § 102(a)(1) over Tomasev. Claim 20 as amended recites, "processing the sequence of predictions of future health- related events and the updated health-related data to identify one or more health-related events in the sequence of future health-related events as missing health-related events, wherein the missing health-related events do not have a match within the updated health-related data; and providing data representing the missing health-related events for presentation to a user device." The cited portion of Tomasev reads as follows: As an example, the auxiliary outputs can include a respective predicted maximum future observed value for each of one or more medical tests that are correlated with the adverse health event or, more generally, any statistics of future observed values, e.g., mean, median, maximum, minimum, or mode. For example, when the adverse health event is AKI, the auxiliary outputs predict the maximum future observed value of a set of laboratory tests over the same set of time intervals as the future AKI predictions. The laboratory tests predicted are ones that are known to be relevant to kidney function: specifically, the tests can include one or more of creatinine, urea nitrogen, sodium, potassium, chloride, calcium or phosphate. This multi-task approach can in some instances result in better generalization and more-robust representations, especially under class imbalance. In particular, by training the neural network to accurately predict the results of relevant medical tests, the system causes the neural network to generate intermediate representations that more robustly represent the features that are relevant to whether the adverse health event will occur in the future. During training, the system also receives the targets that should have been generated by the neural network (the "main targets" and the "auxiliary targets") and computes a loss based on an error between the main predictions and the main targets and another loss based on an error between the auxiliary targets and the auxiliary predictions. In other words, the system receives the ground truth outputs that reflect the actual future health of the patient. [Tomasev, col. 11 lines 47-57 and col. 12 lines 1-7 ] Applicant respectfully submits that the cited portion of Tomasev does not disclose or suggest "processing the sequence of predictions of future health-related events and the updated health-related data to identify one or more health-related events in the sequence of future health- related events as missing health-related events, wherein the missing health-related events do not have a match within the updated health-related data; and providing data representing the missing health-related events for presentation to a user device." In particular, the cited portion of Tomasev does not disclose or suggest processing the sequence of predictions of future health-related events and the updated health-related data to identify one or more health-related events in the sequence of future health-related events as missing health-related events, and providing data representing the missing health-related events for presentation to a user device. Tomasev describes "comput[ing] a loss based on an error between the main predictions and the main targets." The Action has not shown that the cited portion of Tomasev discloses "processing the sequence of predictions of future health-related events and the updated health-related data to identify one or more health-related events in the sequence of future health-related events as missing health-related events, wherein the missing health-related events do not have a match within the updated health-related data; and providing data representing the missing health-related events for presentation to a user device," as recited in amended claim 20.Thus, amended claim 20 is patentable over Tomasev.
Examiner appreciates applicant’s arguments but does not find them persuasive. Tomasev teaches in Col. 4 lines 4-10 “In implementations the neural network comprises a deep embedding neural network to embed the features in the feature representation in an embedding space. The neural network output may then be generated from the embedded features. For example the deep embedding neural network may comprise a plurality of fully-connected layers (though other architectures may be used).” And see Col. 3 lines 46-67 and see Col. 4 lines 17-27 and see Col. 10 lines 19-46 which discloses the embedding neural network 240 transforms the high-dimensional and sparse input feature representation into a lower-dimensional continuous representation that makes subsequent prediction easier and .examiner notes as previously cited the features are a sequence of health related events and the disclosure is utilizing a temporal convolutional embedding space which is shared learned representation space as each sequence is processed using this same space and the embedding neural networks are representing feature vector from high to low dimensional space which one of ordinary skill would understand is using embedding math functions to do this semantically as disclosed.
Per claim 19 examiner notes Foschini is cited by examiner as teaching “updating the sequence of health-related events by applying a mask for each of the one or more gaps by including a mask token in between the two respective health-related events in the sequence." In cited portions [0076] In a second operation 520, the self-supervised learning system may generate a training data set by using patterns of missing information (i.e., "missingness") in some of the data records of the set to mask other records of the set. For example, in some embodiments, the self-supervised learning system applies a missingness of each data record of the set to a next data record of the set, and repeats the process over one or more iterations to generate the training set. In other embodiments, the self-supervised learning system identifies pairs of data records of the set based a level of similarity and/or a level of overlap in missingness, and applies the missingness of a first data record of the pair to a second data record of the pair to generate the training set. And see [0120] discloses, “Embodiments of the disclosure may generate synthetic data by placing synthetic data values in gaps within a collected sequence of wearable sensor data (e.g., at places where data is missing or has been masked), while removing non-synthetic wearable sensor data values. This may be performed until the entire wearable dataset comprises synthetic data.” And see [0102] discloses, “The encoder sub-system 1070 may generate representations from the data. The encoder sub-system may comprise one or more machine learning algorithms. In some embodiments, one or more of the machine learning algorithms comprises a neural network (or artificial neural network (ANN)). A neural network may be a convolutional neural network (CNN) or recurrent neural network (RNN). A neural network may be a multilayer perceptron (MLP).” . examiner notes synthetic data in gaps is a type of data imputation where the mask token is the missing data thus Foschini does teach at the level of breadth claimed.
As per claim 20, examiner notes for "processing the sequence of predictions of future health-related events and the updated health-related data to identify one or more health-related events in the sequence of future health-related events as missing health-related events, wherein the missing health-related events do not have a match within the updated health-related data; and providing data representing the missing health-related events for presentation to a user device," Tomasev teaches in Col. 9 lines 18-23 and see Col. 11 lines 47-67 and Col. 12 lines 1-15 teaches a comparison to ground truth data as identifying if there is or is not a match to the updated health related future events as it is a type of matching evaluating accuracy of model output for the future prediction and updating if not correct and the claimed construction states future health realated events as missing health events and this feature mapping is presented to a display under BRI is taken as a match.
Examiner maintains the 35 U.S.C. 103 rejection.
Prior Art Cited But Not Relied Upon
Corrado et. al – (US10402721)
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for using recurrent neural networks to analyze health events. One of the methods includes: processing each of a plurality of initial temporal sequences of health events to generate, for each of the initial temporal sequences, a respective network internal state of a recurrent neural network for each time step in the initial temporal sequence; storing, for each of the initial temporal sequences, one or more of the network internal states for the time steps in the temporal sequence in a repository; obtaining a first temporal sequence; processing the first temporal sequence using the recurrent neural network to generate a sequence internal state for the first temporal sequence; and selecting one or more initial temporal sequences that are likely to include health events that are predictive of future health events in the first temporal sequence.
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 Ashley Elizabeth Evans whose telephone number is (571) 270-0110. The examiner can normally be reached Monday – Friday 8:00 AM – 5:00 PM.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Mamon Obeid can be reached on (571) 270-1813. The fax phone number for the organization where this application or proceeding is assigned 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. Should you have questions on access to the Patent Center, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
/ASHLEY ELIZABETH EVANS/Examiner, Art Unit 3687
/MAMON OBEID/Supervisory Patent Examiner, Art Unit 3687
Prosecution Insights