DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on May 12, 2026 has been entered.
Response to Amendment
In the amendment filed on May 12, 2026, the following has occurred: claim(s) 1, 3-4, 6-10, 14, 16-20 have been amended. Now, claim(s) 1-20 are pending.
Notice to Applicant
The Examiner has withdrawn the 35 U.S.C. 101 rejection(s) for the current claims 1-20 because of the claimed portion of “fine-tuning a pre-trained large language model (LLM) to generate search queries from natural language requests, wherein the pre-trained LLM is fine- tuned on a training dataset that includes the database schema, the data dictionary, the object attributes, a prompt template comprising partial instructions for the LLM, and a plurality of example natural language requests and corresponding expected database search queries, to adjust a plurality of parameters of the LLM that minimize error between actual database search queries generated by the LLM and the expected database search queries, wherein the fine-tuned LLM is used to generate a database search query from a natural language request, wherein the database search query is executed against the medical imaging database to identify content relevant to the natural language request” in independent claims 1, 10, and 20. The newly amended claims overcome the 35 U.S.C. 101 rejection(s) as the amended claimed portion integrates the abstract idea into a practical application as the amended claimed portion provides a technical improvement of solving the technical problem of “ghosting” in LLMs in the technical field of large language models (LLMs).
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.
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.
Claims 1-5, 10-15, 20 are rejected under 35 U.S.C. 103 as being unpatentable over Arkoff et al. (U.S. Patent Publication No. 12,001,464) in view of Leary et al. (U.S. Patent Pre-Grant Publication No. 2024/0095463).
As per independent claim 1, Arkoff discloses a system comprising:
one or more processors (See col. 10, ll. 59-67, col. 11, ll. 1-4: The circuitry 202 may include one or more specialized processing units, which may be implemented as an integrated processor or a cluster of processors that perform the functions of the one or more specialized processing units, collectively); and
one or more memories storing processor-executable instructions that, when executed by the one or more processors (See col. 11, ll. 5-18: The memory 204 may include suitable logic, circuitry, interfaces, and/or code that may be configured to store the program instructions to be executed by the circuitry 202), cause the one or more processors to perform operations comprising:
receiving object attributes associated with medical images stored in the medical imaging database, wherein the object attributes include data that is descriptive of the medical images and corresponding studies (See col. 15, ll. 12-21, col. 18, ll. 8-19: The third data source may be associated with a patient and may correspond to medical data that may be obtained from medical records of the patient, which the Examiner is interpreting the medical data to encompass object attributes associated with medical images, and interpreting CT reports and pathology reports to encompass data that is descriptive of the medical images and corresponding studies); and
fine-tuning a pre-trained large language model (LLM) to generate search queries from natural language requests (See col. 7, ll. 52-67, col. 8, ll. 1-37: Examples of different types of the one or more LLMs may include a Generative Pre-trained Transformer (GPT), which the Examiner is interpreting a GPT to encompass a pre-trained LLM, and interpreting tunable (recited as tuncable in Arkoff in col. 8, ll. 28, the Examiner is interpreting this recitation as a spelling error) during training of the network to encompass fine-tuning to generate search queries from natural language requests), wherein the pre-trained LLM is fine-tuned on a training dataset that includes: the database schema, the data dictionary, the object attributes, a prompt template comprising partial instructions for the LLM, and a plurality of example natural language requests and corresponding expected database search queries (See col. 8, ll. 52-67, col. 9, ll. 1-48: Each of the one or more LLMs may correspond to a sophisticated AI system trained on vast amounts of text data, and each of the one or more LLMs may learn to predict and generate text by analyzing patterns and relationships within the massive corpus of text, which the Examiner is interpreting the training on vast amounts of text data to encompass a prompt template comprising partial instructions for the LLM, and a plurality of example natural language requests and corresponding expected database search queries), to adjust a plurality of parameters of the LLM that minimize error between actual database search queries generated by the LLM and the expected search queries (See col. 7, ll. 52-65, col. 8, ll. 26-63: In training of the neural network, one or more parameters of each node of the corresponding neural network may be updated based on whether an output of the final layer for a given input (from the training dataset) matches a correct result based on a loss function for the corresponding neural network, which the Examiner is interpreting the process may be repeated for the same or a different input to encompass adjust a plurality of parameters of the LLM that minimize error between actual database search queries generated by the LLM and the expected search queries as the process is repeated to minimize a training error, and interpreting the set of parameters to encompass a plurality of parameters),
wherein the fine-tuned LLM is used to generate a database search query from a natural language request (See col. 7, ll. 52-65, col. 14, ll. 4-25: The system may be configured to apply a NLP model on the metadata to determine the at least one keyword, which the Examiner is interpreting the metadata to determine the at least one keyword to encompass the fine-tuned LLM is used to generate a database search query from a natural language request), wherein the database search query is executed against the medical imaging database to identify content relevant to the natural language request (See col. 7, ll. 52-65, col. 14, ll. 11-25: The circuitry may be configured to select the first MDG database of the one or more MDG databases based on the determined at least one keyword, which the Examiner is interpreting the first MDG database to encompass the medical imaging database, and interpreting each of the one or more MDG databases based on the determined at least one keyword to encompass identify content relevant to the natural language request.)
While Arkoff teaches the system as described above, Arkoff may not explicitly teach receiving a database schema and a data dictionary associated with a medical imaging database, wherein the database schema describes a structural representation of a logical and physical layout of the medical imaging database.
Leary teaches a system for receiving a database schema and a data dictionary associated with a medical imaging database (See [0041]: A retrieval set tag can be used to specify a database, dictionary, library, or other set of data or facts to be used for a specific inferencing task associated with an endpoint, which the Examiner is interpreting a retrieval set tag can be used to specify a database or dictionary to encompass a database schema and a data dictionary associated with a medical imaging database), wherein the database schema describes a structural representation of a logical and physical layout of the medical imaging database (See [0029]: A data marshalling selection can be made that can specify how to map from input fields with specific types (e.g., for structured language protocols such as JSON or Protobuf) into, and out of, the text strings to be processed by the LLMs, which the Examiner is interpreting the specific types to encompass a structural representation of a logical and physical layout of the medical imaging database as the endpoints described may also include an indicator for the size of LLM to use for a request sent to this endpoint, as well as one or more associated inference parameters (e.g., number of samples or temperature).)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed to modify the system of Arkoff to include receiving a database schema and a data dictionary associated with a medical imaging database, wherein the database schema describes a structural representation of a logical and physical layout of the medical imaging database as taught by Leary. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify Arkoff with Leary with the motivation of providing computationally cost effective solutions (See Detailed Description of Leary in Paragraph [0021]).
Claim(s) 10 and 20 mirror claim 1 only within different statutory categories, and are rejected for the same reason as claim 1.
The addition of “wherein the medical imaging database maintains medical images and non-image data associated with a plurality of clinical studies” of claim 10 is encompassed by Arkoff in col. 15, ll. 12-21, col. 18, ll. 8-19: The third data source may be associated with a patient and may correspond to medical data that may be obtained from medical records of the patient, which the Examiner is interpreting the medical data to encompass object attributes associated with medical images, and interpreting CT reports and pathology reports to encompass non-image data associated with a plurality of clinical studies.
The addition of “wherein the database search query is executed against the medical imaging database to identify content relevant to the natural language request” of claim 20 is encompassed by Arkoff in col. 6, ll. 5-19, col. 15, ll. 22-32: The system may extract, from the first MDG database, raw structured data associated with the determined metadata, which the Examiner is interpreting raw structured data to encompass the content relevant to the natural language request comprises structured data.)
As per claim 2, Arkoff/Leary discloses the system of claim 1 as described above. Arkoff may not explicitly teach wherein the medical imaging database is a picture archiving and communication system (PACS), and wherein the object attributes comprise digital imaging and communications in medicine (DICOM) tags.
Leary teaches a system wherein the medical imaging database is a picture archiving and communication system (PACS) (See [0110]: Machine learning models may be trained at facility using data (such as imaging data) generated at facility (and stored on one or more picture archiving and communication system (PACS) servers at facility), may be trained using imaging or sequencing data 1408 from another facility(ies), or a combination thereof, which the Examiner is interpreting the PACS server to encompass the medical imaging database is a PACS), and wherein the object attributes comprise digital imaging and communications in medicine (DICOM) tags See [0120]: A data augmentation service may further be included that may provide GPU accelerated data (e.g., DICOM, RIS, CIS, REST compliant, RPC, raw, etc.) extraction, resizing, scaling, and/or other augmentation, which the Examiner is interpreting the GPU accelerated data to encompass the object attributes comprise DICOM tags.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed to modify the system of Arkoff to include the medical imaging database is a picture archiving and communication system (PACS), and wherein the object attributes comprise digital imaging and communications in medicine (DICOM) tags as taught by Leary. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify Arkoff with Leary with the motivation of providing computationally cost effective solutions (See Detailed Description of Leary in Paragraph [0021]).
Claim(s) 12 mirrors claim 2 only within a different statutory category, and is rejected for the same reason as claim 2.
As per claim 3, Arkoff/Leary discloses the system of claim 1 as described above. Arkoff further teaches wherein the object attributes are stored in an auxiliary database (See col. 18, ll. 29-53: A set of data sources for obtaining medical data to be stored in one or more medical governance databases, which the Examiner is interpreting to encompass the object attributes are stored in an auxiliary database), wherein the processor-executable instructions further cause the system to determine a second database schema and a second data dictionary associated with the auxiliary database, and wherein the LLM is fine-tuned further using the second database schema and the second data dictionary (See col. 10, ll. 20-39: The one or more MDG databases may be utilized to train the one or more LLMs, and the training is iterative, which the Examiner is interpreting one or more MDG databases to encompass the second database schema and the second data dictionary (col. 7, ll. 35-51: Each MDG database is arranged systematically.)
Claim(s) 13 mirrors claim 3 only within a different statutory category, and is rejected for the same reason as claim 3.
As per claim 4, Arkoff/Leary discloses the system of claim 1 as described above. Arkoff further teaches wherein to generating the database search query includes:
receiving the natural language request via a user input (See col. 10, ll. 10-22: The system may be configured to receive the user input including at least one search query to retrieve first medical data from one or more MDG databases); and providing the natural language request as an input to the fine-tuned LLM, wherein the fine-tuned LLM outputs the database search query (See col. 10, ll. 23-39: The system may apply the one or more LLMs on the received at least one search query, and the system may determine metadata associated with the at least one search query based on the application of the one or more LLMs on the received search query, which the Examiner is interpreting the application of the one or more LLMs to encompass providing the natural language request as an input to the fine-tuned LLM.)
Claim(s) 14 mirrors claim 4 only within a different statutory category, and is rejected for the same reason as claim 4.
As per claim 5, Arkoff/Leary discloses the system of claim 1 as described above. Arkoff further teaches wherein the content relevant to the natural language request comprises at least one of medical images, structured data, or unstructured data associated with at least one study in the medical imaging database, wherein the instructions further cause the system to present the content relevant to the natural language request via a graphical user interface (GUI) (See Fig. 2 and col. 6, ll. 5-19, col. 15, ll. 22-32: The system may extract, from the first MDG database, raw structured data associated with the determined metadata, which the Examiner is interpreting raw structured data to encompass the content relevant to the natural language request comprises structured data., and a display device may be used.)
Claim(s) 15 mirrors claim 5 only within a different statutory category, and is rejected for the same reason as claim 5.
As per claim 11, Arkoff/Leary discloses the computer-implemented method of claim 10 as described above. Arkoff further teaches wherein the object attributes comprise at least one of: patient attributes including demographics, a medical state, and a medical history of a patient associated each of the plurality of clinical studies; study attributes indicative of an imaging procedure that was used to capture the medical images associated with each of the plurality of clinical studies (See col. 15, ll. 12-21, col. 18, ll. 8-19: The third data source may be associated with a patient and may correspond to medical data that may be obtained from medical records of the patient, which the Examiner is interpreting CT reports and pathology reports to encompass study attributes indicative of an imaging procedure that was used to capture the medical images associated with each of the plurality of clinical studies); or image attributes that describe each of the medical images and their associated acquisition parameters for each of the plurality of clinical studies
Claims 6 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Arkoff et al. (U.S. Patent Publication No. 12,001,464) in view of Leary et al. (U.S. Patent Pre-Grant Publication No. 2024/0095463) in further view of He et al. (U.S. Patent Pre-Grant Publication No. 2024/0362286).
As per claim 6, Arkoff/Leary discloses the system of claim 1 as described above. Arkoff/Leary may not explicitly teach wherein the content relevant to the natural language request is ranked according to relevance by the fine-tuned LLM.
He teaches a system wherein the content relevant to the natural language request is ranked according to relevance by the fine-tuned LLM (See [0131]: Ranking algorithms that take into account factors like term frequency and document relevance, which the Examiner is interpreting to encompass the claimed portion when combined with Arkoff/Leary.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed to modify the system of Arkoff/Leary to include the content relevant to the natural language request is ranked according to relevance by the fine-tuned LLM as taught by He. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify Arkoff/Leary with He with the motivation of improving search results for a user (See Detailed Description of He in Paragraph [0036]).
Claim(s) 16 mirrors claim 6 only within a different statutory category, and is rejected for the same reason as claim 6.
Claims 7 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Arkoff et al. (U.S. Patent Publication No. 12,001,464) in view of Leary et al. (U.S. Patent Pre-Grant Publication No. 2024/0095463) in further view of Lehmann et al. (U.S. Patent Pre-Grant Publication No. 2024/0370464).
As per claim 7, Arkoff/Leary discloses the system of claim 1 as described above. Arkoff/Leary may not explicitly teach wherein the LLM is a first LLM, and wherein the operations further comprise:
training a second LLM to encode documents containing unstructured data as vector embeddings, wherein the second LLM is trained using a second training dataset comprising a plurality of text documents that are representative of unstructured data stored in the medical imaging database,
wherein the trained second LLM is used to generate a second database search query for the medical imaging database from the natural language request, wherein the second database search query is used in conjunction with the database search query to identify the content relevant to the natural language request.
Lehmann teaches a system wherein the LLM is a first LLM (See [0105]: Providing the embeddings to a large language model and receiving an identifier of the first cluster from the large language model), and wherein the operations further comprise:
training a second LLM to encode documents containing unstructured data as vector embeddings, wherein the second LLM is trained using a second training dataset comprising a plurality of text documents that are representative of unstructured data stored in the medical imaging database (See [0037]-[0038], [0051]: The cognition processing system is configured to receive unstructured content and process the unstructured input into different domains of knowledge, the LLM can be trained based on particular knowledge sets, which the Examiner is interpreting unstructured content to encompass unstructured data to encompass encode documents containing unstructured data as vector embeddings ([0052]), and interpreting the LLM can be trained based on particular knowledge sets to encompass a second training dataset comprising a plurality of text documents that are representative of unstructured data stored in the medical imaging database ([0112]),
wherein the trained second LLM is used to generate a second database search query for the medical imaging database from the natural language request, wherein the second database search query is used in conjunction with the database search query to identify the content relevant to the natural language request (See [0060]-[0061], [0070]: The classification engine may use a large language model (LLM) to extract key terms from the content using the various attention techniques, and a self-attention technique can be used to identify the relevant terms within the document, and a cross-attention technique can be used to compare the relevant terms to other taxonomies to identify a classification, which the Examiner is interpreting identify the relevant terms within the document to encompass identify the content relevant to the natural language request to encompass identify the content relevant to the natural language request when combined with Arkoff/Leary, and interpreting an attention function can be described as mapping a query and a set of key-value pairs to an output, where the query, keys, values, and output are all vectors to encompass a second database search query for the medical imaging database from the natural language request.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed to modify the system of Arkoff/Leary to include the LLM is a first LLM, and wherein the operations further comprise: training a second LLM to encode documents containing unstructured data as vector embeddings, wherein the second LLM is trained using a second training dataset comprising a plurality of text documents that are representative of unstructured data stored in the medical imaging database, wherein the trained second LLM is used to generate a second database search query for the medical imaging database from the natural language request, wherein the second database search query is used in conjunction with the database search query to identify the content relevant to the natural language request as taught by Lehmann. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify Arkoff/Leary with Lehmann with the motivation of improve the performance of downstream tasks (See Detailed Description of Lehmann in Paragraph [0045]).
Claim(s) 17 mirrors claim 7 only within a different statutory category, and is rejected for the same reason as claim 7.
Claims 8-9 and 18-19 are rejected under 35 U.S.C. 103 as being unpatentable over Arkoff et al. (U.S. Patent Publication No. 12,001,464) in view of Leary et al. (U.S. Patent Pre-Grant Publication No. 2024/0095463) in view of Lehmann et al. (U.S. Patent Pre-Grant Publication No. 2024/0370464) in further view of He et al. (U.S. Patent Pre-Grant Publication No. 2024/0362286).
As per claim 8, Arkoff/Leary discloses the system of claim 1 and Arkoff/Leary/Lehmann discloses the system of claim 7 as described above. Arkoff/Leary may not explicitly teach wherein the operations further comprise:
receiving a plurality of unstructured data documents from the medical imaging database;
encoding each of the plurality of unstructured data documents as a vector embedding using the trained second LLM;
and generating a vector database that includes, for each of the plurality of unstructured data documents, the vector embedding and an identifier for an associated study in the medical imaging database, wherein the second database search query is executed against the vector database.
Lehmann teaches a system wherein the operations further comprise:
receiving a plurality of unstructured data documents from the medical imaging database (See [0037]: The cognition processing system is configured to receive unstructured content and process the unstructured input into different domains of knowledge, which the Examiner is interpreting receive unstructured content to encompass the claimed portion);
encoding each of the plurality of unstructured data documents as a vector embedding using the trained second LLM (See [0044]: The encoder takes the input text and converts the input text into a sequence of hidden representations and captures the meaning of the text at different levels of abstraction.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed to modify the system of Arkoff/Leary to include receiving a plurality of unstructured data documents from the medical imaging database, encoding each of the plurality of unstructured data documents as a vector embedding using the trained second LLM as taught by Lehmann. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify Arkoff/Leary with Lehmann with the motivation of improve the performance of downstream tasks (See Detailed Description of Lehmann in Paragraph [0045]).
While Arkoff/Leary/Lehmann discloses the system as described above, Arkoff/Leary/Lehmann may not explicitly teach generating a vector database that includes, for each of the plurality of unstructured data documents, the vector embedding and an identifier for an associated study in the medical imaging database, wherein the second database search query is executed against the vector database.
He teaches a system to generating a vector database that includes, for each of the plurality of unstructured data documents, the vector embedding and an identifier for an associated study in the medical imaging database, wherein the second database search query is executed against the vector database (See [0110], [0130]-[0132]: The search manager may store the document vectors in a database, and index the document vectors into a searchable document index, which the Examiner is interpreting store the document vectors in a database to encompass a vector database, and interpreting the search query to encompass the second database search query executed against the vector database.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed to modify the system of Arkoff/Leary/Lehmann to include generating a vector database that includes, for each of the plurality of unstructured data documents, the vector embedding and an identifier for an associated study in the medical imaging database, wherein the second database search query is executed against the vector database as taught by He. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify Arkoff/Leary/Lehmann with He with the motivation of improving search results for a user (See Detailed Description of He in Paragraph [0036]).
As per claim 9, Arkoff/Leary discloses the system of claim 1, Arkoff/Leary/Lehmann discloses the system of claim 7, and Arkoff/Leary/Lehmann/He discloses the system of claim 8 as described above. Arkoff/Leary/Lehmann may not explicitly teach wherein generating the second database search query includes: receiving the natural language request via a user input; and providing the natural language request as an input to the trained second LLM, wherein the trained second LLM outputs a vector embedding of the natural language request as the second database search query, and wherein the vector embedding of the natural language request is executed against the vector database using a similarity search.
He teaches a system wherein generating the second database search query includes:
receiving the natural language request via a user input (See [0132]-[0133]: The search manager may receive a search query in a natural language representation); and
providing the natural language request as an input to the trained second LLM, wherein the trained second LLM outputs a vector embedding of the natural language request as the second database search query, and wherein the vector embedding of the natural language request is executed against the vector database using a similarity search (See [0122], [0196]-[0197]: The search model may receive as input the information blocks of the electronic document and output contextualized embeddings corresponding to each of the information blocks to form a set of document vectors, which the Examiner is interpreting the information blocks to encompass the natural language request, and the semantically similar document content to encompass the natural language request is executed against the vector database using a similarity search.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed to modify the system of Arkoff/Leary/Lehmann to include generating the second database search query includes: receiving the natural language request via a user input; and providing the natural language request as an input to the trained second LLM, wherein the trained second LLM outputs a vector embedding of the natural language request as the second database search query, and wherein the vector embedding of the natural language request is executed against the vector database using a similarity search as taught by He. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify Arkoff/Leary/Lehmann with He with the motivation of improving search results for a user (See Detailed Description of He in Paragraph [0036]).
Claim(s) 19 mirrors claim 9 only within a different statutory category, and is rejected for the same reason as claim 9.
As per claim 18, Arkoff/Leary discloses the computer-implemented method of claim 10 and Arkoff/Leary/Lehmann discloses the computer-implemented method of claim 17 as described above. Arkoff/Leary may not explicitly teach further comprising:
receiving, by the one or more processors, a plurality of unstructured data documents from the medical imaging database;
synthesizing, by the one or more processors, pixel data for each medical image in a selected series to generate a corresponding synthesized content document;
encoding, by the one or more processors and using a large multimodal model (LMM), each corresponding synthesized content document as a vector embedding using the trained second LLM; and
generating, by the one or more processors, a vector database that includes, for each of the plurality of unstructured data documents, the vector embedding and an identifier for an associated study in the medical imaging database, wherein the second database search query is executed against the vector database.
Lehmann teaches a computer-implemented method further comprising:
receiving, by the one or more processors, a plurality of unstructured data documents from the medical imaging database (See [0037]: The cognition processing system is configured to receive unstructured content and process the unstructured input into different domains of knowledge, which the Examiner is interpreting receive unstructured content to encompass the claimed portion);
synthesizing, by the one or more processors, pixel data for each medical image in a selected series to generate a corresponding synthesized content document (See [0044]-[0045]: An embedding is a representation of a discrete object, such as a word, a document, or an image, as a continuous vector in a multi-dimensional space, an embedding captures the semantic or structural relationships between the objects, such that similar objects are mapped to nearby vectors, and dissimilar objects are mapped to distant vectors, which the Examiner is interpreting embedding captures the semantic or structural relationships between the objects to encompass pixel data for each medical image in a selected series to generate a corresponding synthesized content document);
encoding, by the one or more processors and using a large multimodal model (LMM), each corresponding synthesized content document as a vector embedding using the trained second LLM (See [0044]: The encoder takes the input text and converts the input text into a sequence of hidden representations and captures the meaning of the text at different levels of abstraction.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed to modify the method of Arkoff/Leary to include receiving, by the one or more processors, a plurality of unstructured data documents from the medical imaging database; synthesizing, by the one or more processors, pixel data for each medical image in a selected series to generate a corresponding synthesized content document; encoding, by the one or more processors and using a large multimodal model (LMM), each corresponding synthesized content document as a vector embedding using the trained second LLM as taught by Lehmann. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify Arkoff/Leary with Lehmann with the motivation of improve the performance of downstream tasks (See Detailed Description of Lehmann in Paragraph [0045]).
While Arkoff/Leary/Lehmann teaches the computer-implemented method as described above, Arkoff/Leary/Lehmann may not explicitly teach generating, by the one or more processors, a vector database that includes, for each of the plurality of unstructured data documents, the vector embedding and an identifier for an associated study in the medical imaging database, wherein the second database search query is executed against the vector database.
He teaches a computer-implemented method further comprising: generating, by the one or more processors, a vector database that includes, for each of the plurality of unstructured data documents, the vector embedding and an identifier for an associated study in the medical imaging database, wherein the second database search query is executed against the vector database (See [0110], [0130]-[0132]: The search manager may store the document vectors in a database, and index the document vectors into a searchable document index, which the Examiner is interpreting store the document vectors in a database to encompass a vector database, and interpreting the search query to encompass the second database search query executed against the vector database.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed to modify the method of Arkoff/Leary/Lehmann to include generating, by the one or more processors, a vector database that includes, for each of the plurality of unstructured data documents, the vector embedding and an identifier for an associated study in the medical imaging database, wherein the second database search query is executed against the vector database as taught by He. One of ordinary skill in the art before the effective filing date of the claimed invention would have been motivated to modify Arkoff/Leary/Lehmann with He with the motivation of improving search results for a user (See Detailed Description of He in Paragraph [0036]).
Response to Arguments
In the Remarks filed on May 12, 2026, the Applicant argues that the newly amended and/or added claims overcome the 35 U.S.C. 101 rejection(s) and 35 U.S.C. 103 rejection(s). The Examiner acknowledges that the newly added and/or amended claims overcome the 35 U.S.C. 101 rejection(s). However, the Examiner does not acknowledge that the newly added and/or amended claims overcome the 35 U.S.C. 103 rejection(s).
The Applicant argues that:
(1) the Final Office Action alleges that Arkoff's broad reference to ''vast amounts of text data'' in col. 7, In. 52-col. 8, In. 4, reproduced below, teaches or suggests that such data includes the ''training dataset that includes the database schema, the data dictionary, the object attributes, a prompt template comprising partial instructions for the LLM, and a plurality of example natural language requests and corresponding expected database search queries'' as claimed. However, the Office does not provide any reason as to why one of ordinary skill in the art would be motivated to select the specific training dataset, as claimed, from the vast amounts of text data for training an LLM. Nevertheless, to expedite prosecution and further distinguish the claims from Arkoff, claim 1 has been further amended to clarify that the pre-trained LLM is ''fine-tuned'' on the claimed ''training dataset that includes the database schema, the data dictionary, the object attributes, a prompt template comprising partial instructions for the LLM, and a plurality of example natural language requests and corresponding expected database search queries." One of ordinary skill in the art would understand that Arkoff's broad reference to training LLMs (e.g., BERT, GPT) on ''vast amounts of text data'' refers to training an LLM from scratch, resulting in a pre-trained model that would be ''capable of understanding, generating, and processing human-like language at an extensive scale," but that such pre-trained models would then need to be further fine-tuned with specific data to enhance performance in specific applications. Applicant's own specification in para. [0035] states that ''[i]t should be appreciated that ''training'' an LLM, in this context, does not necessarily refer to training an LLM from scratch, but rather can be viewed as ''fine-tuning'' an LLM to a particular use-case." As best understood, the cited portions of Arkoff do not appear to teach or suggest any fine-tuning of LLMs that are pre-trained on ''vast amounts of text data.";
(2) Arkoff does not teach or suggest ''fine-tuning the pre-trained LLM on a training dataset that includes the database schema, the data dictionary, the object attributes, a prompt template comprising partial instructions for the LLM, and a plurality of example natural language requests and corresponding expected database search queries," as recited by amended claim 1. The other cited references do not remedy the deficiencies of Arkoff with respect to this feature, nor are they cited for this purpose. Consequently, the cited references, even in combination, at least fail to teach or suggest ''fine-tuning the pre-trained LLM on a training dataset that includes the database schema, the data dictionary, the object attributes, a prompt template comprising partial instructions for the LLM, and a plurality of example natural language requests and corresponding expected database search queries," as recited by amended claim 1. Claims 10 and 20 recite similar features. Accordingly, Applicant respectfully requests that the rejection of claims 1, 10, and 20, and their respective dependent claims, under 35 U.S.C. § 103 be withdrawn.
In response to argument (1), the Examiner does not find the Applicant’s argument(s) persuasive. The Examiner maintains that Arkoff teaches “training dataset that includes the database schema, the data dictionary, the object attributes, a prompt template comprising partial instructions for the LLM, and a plurality of example natural language requests and corresponding expected database search queries” as Arkoff in col. 8, ll. 52-67, col. 9, ll. 1-48: Each of the one or more LLMs may correspond to a sophisticated AI system trained on vast amounts of text data, and each of the one or more LLMs may learn to predict and generate text by analyzing patterns and relationships within the massive corpus of text, which the Examiner is interpreting the training on vast amounts of text data to encompass a prompt template comprising partial instructions for the LLM, and a plurality of example natural language requests and corresponding expected database search queries as the neural network can be tuned (col. 8, ll. 28) with a set of parameters. The 35 U.S.C. 103 rejection(s) stand.
In response to argument (2), the Examiner does not find the Applicant’s argument(s) persuasive. The Examiner maintains that Arkoff does teach “fine-tuning the pre-trained LLM on a training dataset that includes the database schema, the data dictionary, the object attributes, a prompt template comprising partial instructions for the LLM, and a plurality of example natural language requests and corresponding expected database search queries,” in col. 8, ll. 52-67, col. 9, ll. 1-48 as each f the one or more LLMs may correspond to a sophisticated AI system trained on vast amounts of text data, and each of the one or more LLMs may learn to predict and generate text by analyzing patterns and relationships within the massive corpus of text, which the Examiner is interpreting the training on vast amounts of text data to encompass a prompt template comprising partial instructions for the LLM, and a plurality of example natural language requests and corresponding expected database search queries as the neural network can be tuned (col. 8, ll. 28) with a set of parameters. The 35 U.S.C. 103 rejection(s) stand.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Grbic et al. (U.S. Patent Pre-Grant Publication No. 2025/0068668), describes method involves receiving prompts comprising patient data retrieved from patient databases and instructions.
Sun (U.S. Patent Pre-Grant Publication No. 2025/0258774), describes a Data Storage Device (DSD) includes a first memory storing reference files used to derive vector embeddings in a vector database, and a query vector embedding is received from a host and one or more vector embeddings similar to the query vector embedding are identified in the vector database.
Reicher et al. (US. Patent Pre-Grant Publication No. 2025/0078987), describes a model generation module or device may, for each of the training medical imaging exams: use finding item criteria to reorganize text of the training report into a list of finding items, each associated with text extracted from the training report text, use natural language processing to analyze the resultant text associated with each finding item to determine an associated classification of each finding item, store, in a training dataset, the training medical imaging exam, the associated finding items, the matching text, and the classifications resulting from the analysis of the matching text.
Welter et al. (“Towards case-based medical learning in radiological decision making using content-based image retrieval”), the IBCR-RE paradigm incorporates a novel combination of essential aspects of diagnostic learning in radiology.
Zhang et al. (“Applications and Challenges for Large Language Models: From Data Management Perspective”), describes promising categories of data management applications where LLMs can be adapted, including data generation, data transformation, data integration, and data exploration. We then discuss the corresponding challenges for such adaption.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Bennett S Erickson whose telephone number is (571)270-3690. The examiner can normally be reached Monday - Friday: 9:00am - 5:00pm.
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, Robert Morgan can be reached at (571) 272-6773. 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.
/Bennett Stephen Erickson/Primary Examiner, Art Unit 3683
Prosecution Insights