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 .
Applicant’s arguments filed in the reply on May 12, 2026 were received and fully considered. Claims 1 and 5 were amended. Claims 2-4 were cancelled. Please see below for more detail.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1 and 5-19 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
Regarding Claim 1, the claim(s) recites “(d) applying the subcutaneous EEG measurement data to the trained machine learning algorithm with the computer system” which amounts to an abstract idea (mental process).
This judicial exception is not integrated into a practical application because:
- The claims fail to outline an improvement to the technical field.
- The claims fail to apply the judicial exception to effect a particular treatment.
- The claims fail to apply the judicial exception with a particular machine.
- The claims fail to effect a transformation or reduction of a particular article to a different state or thing.
Next, the claim as a whole is analyzed to determine whether any element or a combination of elements, integrates judicial exception into a practical application.
For this part of the 101 analysis, the following additional limitations are considered:
“(a) recording subcutaneous EEG measurement data with the subcutaneous EEG device, wherein the subcutaneous EEG measurement data comprise EEG signals measured subcutaneously from a subject;”
“(b) accessing a trained machine learning algorithm with a computer system, wherein the trained machine learning algorithm has been trained on training data in order to predict a likelihood of seizure onset occurring within the EEG signals contained in the subcutaneous EEG measurement data;”
“(c) transmitting the subcutaneous EEG measurement data from the subcutaneous EEG device to the computer system;”
“generating output as an indication of seizure onset occurring in the subcutaneous EEG measurement data.”
“wherein the trained machine learning algorithm is trained on the training data using a multi-stage training process;
wherein the multi-stage training process includes training an initial machine learning algorithm on first training data and retraining the initial machine learning algorithm on second training data,
generating output as the trained machine learning algorithm; and
wherein the first training data comprise scalp-recorded EEG data and the second training data comprise subcutaneously recorded EEG data acquired from subjects.”
The additional elements are insufficient to amount to significantly more than the judicial exception because they seem to merely generally link the use of the judicial exception to a particular technological environment.
Moreover, the claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because they pertain merely to insignificant extrasolution data gathering activities and generic postsolution activity of generating an output.
Furthermore, EEG measurements devices are general fields of use and generic computer elements used to perform generic computer functions don’t add significantly more and are well-understood, routine, and previously known to the industry.
None of these limitations, considered as an ordered combination provide eligibility because the claim taken as a whole, does not amount to significantly more than the underlying abstract idea of accessing a seizure prediction algorithm and applying newly recorded subcutaneous EEG data to generate an output of seizure prediction and does not purport to improve the functioning of the signal processing, or to improve any other technology or technical field. Use of a generic signal processing does not amount to significantly more than the abstract idea itself.
Dependent claims 5-19 also do not recite patent eligible subject matter as they merely further limit the abstract idea, recite limitations that do not integrate the claims into a practical application for similar reasons as set forth above, and/or do not recite significantly more than the identified abstract idea for substantially similar reasons as set forth above.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1, 12-16, and 18-19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Gu et al (US 2020/0397363) (“Gu”) in view of Kremen et al (US 2020/0337645) (“Kremen”).
Regarding Claim 1, while Gu teaches a method for predicting a seizure onset in electroencephalography (EEG) measurement data recorded with a subcutaneous EEG measurement device (Abstract, Fig. 10, [0028] implantable medical device includes the physiological detecting components for electroencephalogram, thus reflecting a subcutaneous EEG measurement device), the method comprising:
(a) recording subcutaneous EEG measurement data with the subcutaneous EEG device, wherein the subcutaneous EEG measurement data comprise EEG signals measured subcutaneously from a subject (Fig. 3 and 10, [0041] detecting unit 110 of implantable medical device records patient physiological data, [0069]-[0078] S301);
(b) accessing a trained machine learning algorithm with a computer system, wherein the trained machine learning algorithm has been trained on training data in order to predict a likelihood of seizure onset occurring within the EEG signals contained in the subcutaneous EEG measurement data ([0069]-[0078] S315 and S302, at an updated loop of the system, a customized prediction algorithm / trained machine learning algorithm is accessed, the customized prediction algorithm trained on internal and external data in order to predict a likelihood of seizure onset occurring within the EEG signals contained in the subcutaneous EEG measurement data);
(c) transmitting the subcutaneous EEG measurement data from the subcutaneous EEG device to the computer system (Figs. 3, 8, and 10, [0043] where the detecting unit 110 originally recording the EEG transmits the EEG to the control unit 120 / computer system, the control unit performs the predictions and steps of Fig. 10);
(d) applying the subcutaneous EEG measurement data to the trained machine learning algorithm with the computer system (Figs. 8 and 10, [0069]-[0078], S303), generating output as an indication of seizure onset occurring in the subcutaneous EEG measurement data (Figs. 8 and 10, [0069]-[0078], S304, [0052] output of prediction);
wherein the trained machine learning algorithm is trained on the training data using a multi-stage training process (Fig. 10, the updating loops of the seizure prediction, customizing the prediction algorithm to the patient is seen as multiple stages of training).
wherein the multi-stage training process includes training an initial machine learning algorithm on first training data and retraining the initial machine learning algorithm on second training data, generating output as the trained machine learning algorithm (Fig. 10, [0069]-[0078] while it is not explicitly stated that the prediction algorithm of S302 is originally a machine learning algorithm, by step S311-S315, the verified customized prediction algorithm is trained by machine learning on first training data set. And in the subsequent loops, newly detected physiological signals are added and used as training data in step S310 for a second verified customized prediction algorithm, thus reflecting a second training data, and again outputting a trained machine learning algorithm after retraining),
Gu fails to teach wherein the first training data comprise scalp-recorded EEG data and the second training data comprise subcutaneously recorded EEG data acquired from subjects.
However Kremen teaches a system for classifying brain state based on EEG data (Abstract, [0007] interested in performing classification of a brain state with data from beyond scalp EEG) comprising machine learning prediction ([0031]), where the machine learning classification training is performed on multiple sets of patient EEG data ([0031] “In some embodiments, the receiving, selecting, calculating, classifying and providing steps are performed by the one or more processors 108 using an operational mode comprising: a fully automated and unsupervised mode (e.g., FIG. 4); or a semi-automated mode; or an active learning mode—unsupervised first and supervised by redefining clusters displayed to user (user can reassign each part of the data into different class and retrain); or a supervised mode—fully supervised and trained by expert or trained on known scalp electrophysiology data in parallel with any simultaneous data (e.g. intracranial, epidural, subscalp, EEG, video recording, EMG, actigraphy, etc.); or other desired mode.” Where the machine learning described herein has been trained on known scalp electrophysiology data, along with other EEG data sources such as subcutaneous EEG, and teaches that machine learning may involve retraining).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to perform the multi-stage EEG training of Gu of multiple EEG training datasets by utilizing multiples types of EEG data such as both scalp and subcutaneous EEG, as taught by Kremen, as a way to create an accurate machine learning prediction for seizure despite differences in data characteristics. Furthermore, it would be obvious that despite the passage only discussing both datasets in the context of parallel training, the stated goal of the invention is to utilize datasets from beyond the scalp so these datasets would be relevant in retraining the algorithm as well.
Regarding Claim 12, Gu and Kremen teach the method of claim 1, wherein the computer system is local to the subcutaneous EEG device (See Claim 1 Rejection).
Regarding Claim 13, Gu and Kremen teach the method of claim 1, wherein the computer system is physically separate from the subcutaneous EEG device ([0052] external monitoring device 20 may also fulfill the limitations of a computer system, by its processing unit 230, accessing of updated seizure prediction algorithm from external machine learning device 30, and receipt of internal EEG measurement transmitted by the implantable medical device).
Regarding Claim 14, Gu and Kremen teach the method of claim 1, further comprising generating an alarm to a user when the trained machine learning algorithm generates output indicating a seizure onset is likely to occur based on the subcutaneous EEG measurement data input to the trained machine learning algorithm (See Claim 1 Rejection, [0052]).
Regarding Claim 15, Gu and Kremen teach the method of claim 14, wherein the alarm comprises an auditory alarm (See Claim 14 Rejection, [0052]).
Regarding Claim 16, Gu and Kremen teach the method of claim 14, wherein the alarm comprises a visual alarm (See Claim 14 Rejection, [0052]).
Regarding Claim 18, Gu and Kremen teach the method of claim 1, further comprising:
providing a user interface to the subject, via the computer system, that is configured to receive feedback on the indication of seizure onset occurring in the subcutaneous EEG measurement data (Fig. 8, [0053], [0058]-[0060]);
receiving user feedback data, via the computer system, wherein the user feedback data indicates whether a seizure event occurred following the indication of seizure onset occurring in the subcutaneous EEG measurement data (Fig. 8, [0053], [0058]-[0060]); and
retraining the machine learning algorithm based on the user feedback data ([0060]).
Regarding Claim 19, Gu and Kremen teach the method of claim 18, wherein the machine learning algorithm is retrained based on the user feedback data using an active learning technique (Fig. 8, [0053], [0058]-[0060]).
Claim(s) 5-7, 10-11, and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Gu in view of Kremen and further in view of Chan et al (US 2021/0282701) (“Chan”).
Regarding Claim 5, while Gu and Kremen teach the method of claim 1, their combined efforts fail to teach wherein the initial machine learning algorithm is retrained using transfer learning on the second training data.
However Chan teaches a seizure prediction system based on machine learning (Abstract, [0120]) and teaches that a transfer learning step facilitates model convergence despite differences in training data sets ([0120] noted on training data for different patients).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to perform the EEG training of Gu and Kremen with transfer learning to compensate for dynamic variation between EEG monitoring locations as taught in Kremen based on the characteristics of transfer learning taught by Chan.
Regarding Claim 6, Gu, Kremen, and Chan teach the method of claim 5, and Gu teaches wherein the initial machine learning algorithm is trained using deep learning ([0075]), and Chan teaches that a deep learning algorithm can specifically be a multi-layer long short-term memory (LSTM) network ([0064], [0110], [0111]).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to specifically set the deep learning of Gu as a multi-layer long short-term memory (LSTM) network as taught by Chan to create a standardized framework for generating machine learning prediction algorithms, ensuring greater consistency across trials.
Regarding Claim 7, Gu, Kremen, and Chan teach the method of claim 6, wherein the multi-layer LSTM network comprises two LSTM network layers (See Claim 6 Rejection, [0110]-[0111]) and Chan teaches that additional LSTM network layers can be used (Claim 38).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, for the multi-layer long short-term memory (LSTM) network taught by Chan to be a three-layer LSTM network instead of two to better capture the complexity of dynamic variation and evolution over time in EEG seizure characteristics ([0110]-[0111]).
Regarding Claim 10, Gu, Kremen, and Chan teach the method of claim 7, and Gu teaches wherein the initial machine learning algorithm is trained using a neural network having at least one convolutional layer ([0075]).
Regarding Claim 11, Gu, Kremen, and Chan teach the method of claim 7, and Chan teaches wherein the initial machine learning algorithm is trained using a neural network having fully connected layers (See Claim 7 Rejection, [0111] the layers of the LSTM network are described as fully connected).
Regarding Claim 17, Gu, Kremen, and Chan teach the method of claim 11, and Gu further teaches the method comprising:
providing a user interface to the subject, via the computer system, that is configured to receive feedback on the indication of seizure onset occurring in the subcutaneous EEG measurement data (Fig. 8, [0053], [0058]-[0060]);
receiving user feedback data, via the computer system, wherein the user feedback data indicates whether a seizure event occurred following the indication of seizure onset occurring in the subcutaneous EEG measurement data (Fig. 8, [0053], [0058]-[0060]); and
performing corrective action on the prediction algorithm based on the user feedback ([0060]), and Chan further that a refinement based seizure detection ([0040]-[0045]) can output seizure alarm can be based on threshold value ([0046]).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, that if the output predictions of seizure are receiving feedback based on seizures that do not arrive as taught by Gu, and one uses a threshold taught by Chan to make those predictions, then one of the corrective actions of Gu will be to modify the threshold. The teaching of Chan provides a clarification on how the alarm step can be generated (i.e. threshold) to provide a reviewable metric and Gu’s correction only strengthens the threshold’s benefit for the seizure prediction.
Claim(s) 8 is/are rejected under 35 U.S.C. 103 as being unpatentable over Gu in view of Kremen and further in view of Chan and further in view of Supratak et al (“TinySleepNet: An Efficient Deep Learning Model for Sleep Stage Scoring based on Raw Single-Channel EEG”) (“Supratak”).
Regarding Claim 8, while Gu, Kremen, and Chan teach the method of claim 7, their combined efforts fail to teach wherein the first and second LSTM network layers are non-trainable and the third LSTM network layer is trainable.
However Supratak teaches an EEG-deep learning system for automated classification of brain data (p641, Abstract) and further teaches a deep learning model that utilizes multiple non-trainable layers (p642, Fig. 1).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to specifically layers within the LSTM network of Gu, Kremen, and Chan as non-trainable as taught by Supratak as this will reduce the amount of layers that need to be trained, and thus the training time of the overall model. And identifying two layers as non-trainable while maintaining one layer as trainable is recognized as an optimization between improved accuracy by providing the model more capability of customizing to a subject versus a reduced wait time in machine learning training by reducing the information that can be modified and thus reducing the potential complexity of the algorithm modification.
Claim(s) 9 is/are rejected under 35 U.S.C. 103 as being unpatentable over Gu in view of Kremen and further in view of Chan and further in view of Firouzi et al (US 2021/0000444) (“Firouzi”).
Regarding Claim 9, while Gu, Kremen, and Chan teach the method of claim 7, and Chan teaches wherein the initial machine learning algorithm is trained using a recurrent neural network ([0110] beyond the LSTM network, another RNN network 902b as part of the seizure prediction process), their combined efforts fail to teach the recurrent neural network comprising at least one gated recurrent unit (GRU) layer.
However Firouzi teaches a machine learning seizure analysis system (Abstract) based on patient brain data where patient brain data for seizure analysis is performed by machine learning ([0158]) and example machine learning algorithms include recurrent neural networks such as GRUs ([0159).
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to specifically set the second RNN of Chan as a GRU as taught by Firouzi to create a standardized framework for generating machine learning prediction algorithms, ensuring greater consistency across trials.
Response to Arguments
Applicant’s amendments and arguments filed May 12, 2026 with respect to the 35 USC 101 rejections have been fully considered, but are not persuasive. Applicant argues that the use of scalp-recorded EEG data to train an initial machine learning algorithm and then use subcutaneously recorded EEG to retrain the initial machine learning algorithm improves upon the prior art by applying abundant scalp EEG data and refining the machine learning algorithm with scarcer subcutaneous EEG data to develop an accurate seizure classifier. Examiner respectfully disagrees. While this concept may reflect an improvement in the art, the improvement is not captured by the current claim language. There is no limit placed on the amount of training data applied. For example, if an hour of scalp-recorded EEG data was the first training dataset and 30 seconds of subcutaneous-recorded EEG data was the second training dataset, it is unclear if there would be a large change in classifier accuracy from such a small sample. Theoretically, a reduced training time with high accuracy is achieved by a combined training dataset from combining the scalp and subcutaneous data, where the combined training dataset is smaller than an equivalent scalp training dataset that would need more data to reflect the same level of accuracy. However, that is not the given claim language. The rejection stands.
Applicant’s amendments and arguments filed May 12, 2026 with respect to the 35 USC 103 rejections have been fully considered, but are not persuasive. Applicant argues that Kremen’s parallel training approach is fundamentally different from the sequential multi-stage approach process recited in amended claim 1. Examiner respectfully disagrees. Upon reconsideration, the multi-stage training process with two training datasets is outlined by Gu itself, as written in the rejection above. And Kremen teaches both machine learning training done in parallel and with a sequential retraining in [0031]. Further, Kremen may not be concerned about data scarcity, but does teach that next generation brain monitor can be implanted and require training for accurate prediction, thus necessitating datasets from other monitoring locations ([0007]). In totality, we can see that a next-generation implanted brain monitor is relevant to Gu’s seizure-based brain monitor, that these next-generation brain monitors benefit from prediction algorithms generated from multiple locations, and that parallel training and retraining are both considered forms of developing machine learning in this context. It would be obvious to apply these datasets to the retraining in Gu to achieve the same benefit of readying the prediction algorithms for new generations of monitors.
Applicant also argues that Kremen is directed to behavioral state classification not seizure prediction with different goals and thus Gu and Kremen are not addressing the data scarcity problem through a multi-stage approach and do not provide motivation to combine to do so. Examiner respectfully disagrees for the reasons given above.
Consequently, Claims 5-19 remain rejected due to their dependency on rejected independent claim 1.
Conclusion
THIS ACTION IS MADE FINAL. 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 JAIRO H PORTILLO whose telephone number is (571)272-1073. The examiner can normally be reached M-F 9:00 am - 5:15 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, Jacqueline Cheng can be reached at (571)272-5596. 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.
/JAIRO H. PORTILLO/
Examiner
Art Unit 3791
/PUYA AGAHI/Primary Examiner, Art Unit 3791
Prosecution Insights