DETAILED ACTION
This is responsive to application 18/776,169 filed on 07/17/2024 in which claims 1-20 are presented for examination.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claim 1:
Step 1: Is the claim to a process, machine, manufacture or composition of matter?” Yes, it’s a method.
Step 2a Prong 1 (judicial exception)
Step 2A (1): “Does the claim recite an abstract idea, law of nature, or natural phenomenon? Yes , the claim comes under mental processes.
Claim 1 recites:
“A method for generating structured metadata from a plurality of published case reports, the method comprising: identifying a plurality of relevant case reports from a database of published case reports; extracting relevant text from one or more of the relevant case reports; extracting a plurality of entities from the relevant text, wherein each of the entities has an entity type and corresponds to at least a part of the relevant text; predicting a relationship between one or more pairs of the extracted entities; grouping two or more of the entities into a group based on the entity types of one or more of the entities and the relationship; and mapping one or more of one or more of the entities and the predicted relationship to a database of medical terminology.”
All the limitations above are abstract idea related to the mental process (concepts performed in the human mind (including an observation, evaluation, judgment, opinion)). Claim language pertains to analyzing medical case reports and extracting/(choosing relevant information) . one can use pen and paper to extract relevant data from a case report. Grouping related data can also be done and mapping/matching the data to relevant medical terminology, and can be done using pen and paper.
Step 2A(2): Prong Two: evaluate whether the claim recites additional elements that integrate the exception into a practical application of the exception. NO
The claim does recite additional elements; however they don’t integrate the exception into a practical application of the exception.
database (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f))
Step 2B: evaluate whether the claim recites additional elements that amount to an inventive concept (aka “significantly more”) than the recited judicial exception? NO
As discussed previously with respect to Step 2A Prong Two, the additional element in the claim amounts to no more than mere instructions to apply the exception using a generic computer component.
The same analysis applies here in 2B, i.e., mere instructions to apply an exception using a generic computer component cannot integrate a judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Dependent claims 2-17 further narrows the abstract idea described in claim 1, and add the additional elements of “database of medical terminology”, “ICD10 database”, “SNOMED database”, “NCBI database”, “OVIDTM database”, “machine-learning model/s” , “natural language processing (NLP) model.”, “extracting the relevant text”, “machine-readable file format” .
Under step 2A, prong two, the additional elements don’t integrate the exception into a practical application of the exception as merely adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f).
Regarding claim limitation “extracting the relevant text”( Adding insignificant extra-solution activity to the judicial exception - see MPEP 2106.05(g) )
extracting the relevant text “the courts have recognized the computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner (v. Electronically scanning or extracting data from a physical document, Content Extraction and Transmission, LLC v. Wells Fargo Bank, 776 F.3d 1343, 1348, 113 USPQ2d 1354, 1358 (Fed. Cir. 2014) (optical character recognition);
As discussed previously with respect to Step 2A Prong Two, the additional elements in the claim amounts to no more than mere instructions to apply the exception using a generic computer component.
The same analysis applies here in 2B, i.e., mere instructions to apply an exception using a generic computer component cannot integrate a judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Regarding claim 18, it is rejected under the same rationale as claim 1.
Dependent claims 19-20 further narrows the abstract idea described in claim 18, and add the additional element of “BioBERTTM natural language processing model.”
Under step 2A, prong two, the additional elements don’t integrate the exception into a practical application of the exception as merely adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f).
As discussed previously with respect to Step 2A Prong Two, the additional elements in the claim amounts to no more than mere instructions to apply the exception using a generic computer component.
The same analysis applies here in 2B, i.e., mere instructions to apply an exception using a generic computer component cannot integrate a judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claim 1-4, 13, and 17-18 are rejected under 35 U.S.C. 102a(2) as anticipated by SHARMA et al. ( US 20250078971 A1)
Regarding claim 1, SHARMA teaches a method for generating structured metadata from a plurality of published case reports, the method comprising:
identifying a plurality of relevant case reports from a database of published case reports (para, “[0006] The structured medical data can be provided for various applications. For example, the structured medical data can be stored in a searchable database from which the entities and their values (standardized or not) can be retrieved based on search queries. The searchable database, as well as the structured medical data, can also be made available to various applications, such as a clinical decision support application, an analytics application, etc., for processing. For example, the clinical decision support application can retrieve entities relevant to a clinical decision (e.g., diagnosis, procedure history, medication history) and their values from the database, and process the entities to generate an output to support a clinical decision. An analytics application can obtain entities related to, for example, treatment history and diagnosis from the pathology reports of a large number of patients and perform analysis to obtain insights in healthcare delivery and quality of care. In other examples, a clinical portal application can be provided to display the structured medical data, and/or to display an image of a pathology report with extracted entity information overlaid on the image.”
“[0025] …. …The baseline NLP model sub-can be trained/built based on biomedical articles from various major sources such as, for example, PubMed Central®, a free full-text archive of biomedical and life sciences journal literature at the U.S. National Institutes of Health's National Library of Medicine…..”
Note: Also, see para 0021,0030);
extracting relevant text from one or more of the relevant case reports (para “[0028] …..The pre-configured image recognition operation can be performed on images of pathology reports to extract text strings, and the text strings can be input to the NLP to extract pathology entities. The parameters of the image recognition operation can then be adjusted based on the accuracy of extraction by the NLP.”
“[0025] The NLP model can be trained to identify sequences of text strings including entities and values, and extract the entities and values based on the identification. The NLP can be trained in a two-step process. As a first step, the NLP model can be trained based on documents including common medical terminologies to build a baseline NLP sub-model. The baseline NLP sub-model can be used to provide a primary context for identifying sequences of text strings that include common medical terminologies that may (or may not) include pathology entities. The baseline NLP model sub-can be trained/built based on biomedical articles from various major sources such as, for example, PubMed Central®, a free full-text archive of biomedical and life sciences journal literature at the U.S. National Institutes of Health's National Library of Medicine. As a second step, the baseline NLP sub-model is then trained using text strings from pathology reports, to expand the sub-model to include pathology entities. ….”);
extracting a plurality of entities from the relevant text, wherein each of the entities has an entity type and corresponds to at least a part of the relevant text( para, “[0022] In some embodiments, a workflow is provided to extract pathology entities from images of pathology reports. The workflow can begin with extracting text strings from an image file of a pathology report. The extraction of the text strings from the image file can be based on an image recognition process to recognize characters and/or text strings from the image, such as optical character recognition (OCR), optical word recognition, etc. The workflow further includes recognizing, using a natural language processor (NLP), entities from the text strings, each entity including a label and a value, and determining the values of the entities from the text strings. The entities generally refer to pre-defined medical categories and classifications, such as medical diagnoses, procedures, medications, specific locations/organs in the patient's body, etc. Each entity has a label which indicates the category/classification, and a value which indicates the data being categorized/classified. In some examples, the workflow includes mapping the values of at least some of the entities to standard terminologies. The mapping can be part of an enrichment process, in which the values of at least some of the entities, which may be non-standardized representation of the categorized/classified data, are converted into standardized data, such as clinical terminologies and codes defined under the Systematized Nomenclature of Medicine (SNOMED) standard. The workflow can then generate structured medical data that associate the labels of the entities with at least one of the values of the entities or the standardized terminologies.”)
predicting a relationship between one or more pairs of the extracted entities (para, “[0060]…... As part of the training operation, a sequence of text strings with a particular label (e.g., a labeled entity, a labeled entity value, a labeled context) can be input to NLP model 328 to determine whether NLP outputs correct entity-value pairs and/or context information. If training module 340 determines that NLP model 328 does not output correct entity-value pairs and/or context information (e.g., based on comparing the labelled entity/entity value of the sequence of text strings and the entity-value pairs output by the NLP model for the sequence of text strings), the training module 340 can modify NLP model 328 by creating new nodes representing new words, adding edges between existing nodes, etc. The decision mechanism to output an entity-value pair (e.g., a parameterized equation) can be also be updated (e.g., by updating the parameters) to increase the likelihood of outputting the correct entity-pair and/or context information.”
Also, claim 6 “6. The method of claim 5 wherein the parameters of the NLP model are modified in response to determining that the NLP model incorrectly extracts entity/value pairs and/or context information, wherein the parameters are modified by at least one of adding a new terminology to the NLP model or new relationship between entities of the NLP model.” Also, see Fig. 4A);
grouping two or more of the entities into a group based on the entity types of one or more of the entities and the relationship(para, “[0060] Referring back to FIG. 3, NLP model 328 can be a machine-learning model that is trained. As shown in FIG. 3, system 300 may include a training module 340, which can train NLP model 328. Training module 340 can train NLP model 328 based on labeled general medical documents 348 and labeled pathology report 350. General medical documents 348 can include various categories of biomedical literatures, reports, etc. The training create nodes representing words of medical terminologies, as well as edges representing the sequential relationships among the words, such as those of NLP model 328 of FIG. 4A. As part of the training operation, a sequence of text strings with a particular label (e.g., a labeled entity, a labeled entity value, a labeled context) can be input to NLP model 328 to determine whether NLP outputs correct entity-value pairs and/or context information. If training module 340 determines that NLP model 328 does not output correct entity-value pairs and/or context information (e.g., based on comparing the labelled entity/entity value of the sequence of text strings and the entity-value pairs output by the NLP model for the sequence of text strings), the training module 340 can modify NLP model 328 by creating new nodes representing new words, adding edges between existing nodes, etc. The decision mechanism to output an entity-value pair (e.g., a parameterized equation) can be also be updated (e.g., by updating the parameters) to increase the likelihood of outputting the correct entity-pair and/or context information.”
Also, see Fig. 4B (para 0051, “…. nodes 440e and 440f are associated with different organs to receive the surgery, such as heart and breast.”));
and mapping one or more of one or more of the entities and the predicted relationship to a database of medical terminology (para, “[0089] In step 810, enrichment module 310 can convert, using a mapping table that maps the entities and the values to pre-determined terminologies, the values of at least some of the entities to corresponding pre-determined terminologies. The pre-determined terminologies can include standard terminologies defined based on a universal standard, such as SNOMED. The mapping table can be based on data stored in a terminology mapping database, which can include a mapping between an entity-value pair to a standard terminology, such as a SNOMED concept and a concept ID. For each entity-value pair and the associated context, enrichment module 310 can perform a search for the associated SNOMED concept and concept ID in terminology mapping database 370.”
Also, “[0051] … Baseline NLP sub-model 430 can include, for example, nodes 430a, 430b, and 430c. Nodes 430a and 430b can be associated with generic medical terms that are related to histology, such as lesion, tissue, etc., whereas node 430c is associated with generic medical terms not related to histology, such as surgery….”)
Regarding claim 2, SHARMA teaches the method according to claim 1.
SHARMA further teaches wherein the database of medical terminology comprises one or more of: the ICD10 database, the SNOMED database, and the NCBI database(para, “[0022]….. The mapping can be part of an enrichment process, in which the values of at least some of the entities, which may be non-standardized representation of the categorized/classified data, are converted into standardized data, such as clinical terminologies and codes defined under the Systematized Nomenclature of Medicine (SNOMED) standard. The workflow can then generate structured medical data that associate the labels of the entities with at least one of the values of the entities or the standardized terminologies.”)
Regarding claim 3, SHARMA teaches the method according to claim 1.
SHARMA further teaches further comprising normalizing one or more of the entities(para, “[0022]….. Each entity has a label which indicates the category/classification, and a value which indicates the data being categorized/classified. In some examples, the workflow includes mapping the values of at least some of the entities to standard terminologies. The mapping can be part of an enrichment process, in which the values of at least some of the entities, which may be non-standardized representation of the categorized/classified data, are converted into standardized data, such as clinical terminologies and codes defined under the Systematized Nomenclature of Medicine (SNOMED) standard. The workflow can then generate structured medical data that associate the labels of the entities with at least one of the values of the entities or the standardized terminologies.” Note: Also, see para 0037, 0034)
Regarding claim 4, SHARMA teaches the method according to claim 3.
SHARMA further teaches wherein normalizing one or more of the entities comprises associating two or more of the entities with a sub-category (“[0053] ….. Nodes 430a and 430b can be associated with generic medical terms that are related to histology, such as lesion, tissue, etc., whereas node 430c is associated with generic medical terms not related to histology, such as surgery. In addition, pathology NLP sub-model 440 can include nodes 440a, 440b, 440c, 440d, 440e, and 440f. Nodes 440a, 440b, 440c, and 440d can be linked by edges 442, 444, and 446 to form a sequence “lung squamous cell carcinoma.” On the other hand, nodes 440e and 440f are associated with different organs to receive the surgery, such as heart and breast.”)
Regarding claim 13, SHARMA teaches the method according to claim 1.
SHARMA further teaches wherein extracting the relevant text comprises:
converting one or more of the relevant case reports to a machine-readable file format (para, “[0085] In step 804, after receiving the image file, optical processing module 306 can perform an image recognition operation to extract input text strings from the image file. The extraction may include identifying text images from the image file, generating text data represented by the text images, and generating an intermediate text file (e.g., text file 312) including the text data. The image recognition operation may include, for example, optical character recognition (OCR) or optical word recognition. In both operations, optical processing module 306 can extract pixel patterns of characters (e.g., by identifying patterns of pixels with a dark color), compare each pixel pattern with pre-defined pixel patterns of characters, and determine which character (or which word/phrase) each pixel pattern represents based on the comparison. Optical processing module 306 can then store the character/word/phrase into text file 312. Optical processing module 306 can scan through image file 312 following a pre-determined pattern (e.g., raster scanning) to extract and process pixel patterns in a row from left to right, and repeat the scanning for each row. Based on the scanning pattern, optical processing module 306 can generate a sequence of text strings (e.g., characters, words, phrases) and store the sequence of text strings in text file 312.”);
and identifying patient data in one or more of the relevant case reports(para, “[0030]…… Moreover, by making the structured medical data accessible by other applications, such as clinical support applications, analytics applications, etc., a large scale analysis of pathology reports of a large patient population can be performed to provide insights in healthcare delivery and quality of care, to provide relevant data to support a clinical decision made by a clinician, etc. With the improvements in the overall speed of data flow and in the correctness and completeness of extraction of medical data, wider and faster access of high-quality patient data can be provided for clinical and research purposes, which can facilitate the development in treatments and medical technologies, as well as the improvement of the quality of care provided to the patients.” Note: Also, see para 0024)
Regarding claim 17, SHARMA teaches the method according to claim 1.
SHARMA further teaches further comprising categorizing the extracted entities into sub-categories using a fourth machine-learning model(para, “[0051] FIG. 4B illustrates another example of NLP model 328. As shown in FIG. 4B, NLP model 328 can include a baseline NLP sub-model 430 and a pathology NLP sub-model 440. Baseline NLP sub-model 430 can include, for example, nodes 430a, 430b, and 430c. Nodes 430a and 430b can be associated with generic medical terms that are related to histology, such as lesion, tissue, etc., whereas node 430c is associated with generic medical terms not related to histology, such as surgery. In addition, pathology NLP sub-model 440 can include nodes 440a, 440b, 440c, 440d, 440e, and 440f. Nodes 440a, 440b, 440c, and 440d can be linked by edges 442, 444, and 446 to form a sequence “lung squamous cell carcinoma.” On the other hand, nodes 440e and 440f are associated with different organs to receive the surgery, such as heart and breast.”)
Regarding claim 18, SHARMA teaches a method for training a machine-learning model for generating a patient journey from a medical case report, the method comprising:
identifying a plurality of relevant case reports from a database of published case reports( para, “[0006] The structured medical data can be provided for various applications. For example, the structured medical data can be stored in a searchable database from which the entities and their values (standardized or not) can be retrieved based on search queries. The searchable database, as well as the structured medical data, can also be made available to various applications, such as a clinical decision support application, an analytics application, etc., for processing. For example, the clinical decision support application can retrieve entities relevant to a clinical decision (e.g., diagnosis, procedure history, medication history) and their values from the database, and process the entities to generate an output to support a clinical decision. An analytics application can obtain entities related to, for example, treatment history and diagnosis from the pathology reports of a large number of patients and perform analysis to obtain insights in healthcare delivery and quality of care. In other examples, a clinical portal application can be provided to display the structured medical data, and/or to display an image of a pathology report with extracted entity information overlaid on the image.”
“[0025] …. …The baseline NLP model sub-can be trained/built based on biomedical articles from various major sources such as, for example, PubMed Central®, a free full-text archive of biomedical and life sciences journal literature at the U.S. National Institutes of Health's National Library of Medicine…..”
Note: Also, see para 0021,0030);
extracting relevant text from the relevant case reports para “[0028] …..The pre-configured image recognition operation can be performed on images of pathology reports to extract text strings, and the text strings can be input to the NLP to extract pathology entities. The parameters of the image recognition operation can then be adjusted based on the accuracy of extraction by the NLP.”
“[0025] The NLP model can be trained to identify sequences of text strings including entities and values, and extract the entities and values based on the identification. The NLP can be trained in a two-step process. As a first step, the NLP model can be trained based on documents including common medical terminologies to build a baseline NLP sub-model. The baseline NLP sub-model can be used to provide a primary context for identifying sequences of text strings that include common medical terminologies that may (or may not) include pathology entities. The baseline NLP model sub-can be trained/built based on biomedical articles from various major sources such as, for example, PubMed Central®, a free full-text archive of biomedical and life sciences journal literature at the U.S. National Institutes of Health's National Library of Medicine. As a second step, the baseline NLP sub-model is then trained using text strings from pathology reports, to expand the sub-model to include pathology entities. ….”);
generating a plurality of entities from the relevant text, wherein each of the entities has an entity type and corresponds to at least a part of the relevant text para, “[0022] In some embodiments, a workflow is provided to extract pathology entities from images of pathology reports. The workflow can begin with extracting text strings from an image file of a pathology report. The extraction of the text strings from the image file can be based on an image recognition process to recognize characters and/or text strings from the image, such as optical character recognition (OCR), optical word recognition, etc. The workflow further includes recognizing, using a natural language processor (NLP), entities from the text strings, each entity including a label and a value, and determining the values of the entities from the text strings. The entities generally refer to pre-defined medical categories and classifications, such as medical diagnoses, procedures, medications, specific locations/organs in the patient's body, etc. Each entity has a label which indicates the category/classification, and a value which indicates the data being categorized/classified. In some examples, the workflow includes mapping the values of at least some of the entities to standard terminologies. The mapping can be part of an enrichment process, in which the values of at least some of the entities, which may be non-standardized representation of the categorized/classified data, are converted into standardized data, such as clinical terminologies and codes defined under the Systematized Nomenclature of Medicine (SNOMED) standard. The workflow can then generate structured medical data that associate the labels of the entities with at least one of the values of the entities or the standardized terminologies.”)
generating a plurality of relationships between a plurality of pairs of the entities para, “[0060]…... As part of the training operation, a sequence of text strings with a particular label (e.g., a labeled entity, a labeled entity value, a labeled context) can be input to NLP model 328 to determine whether NLP outputs correct entity-value pairs and/or context information. If training module 340 determines that NLP model 328 does not output correct entity-value pairs and/or context information (e.g., based on comparing the labelled entity/entity value of the sequence of text strings and the entity-value pairs output by the NLP model for the sequence of text strings), the training module 340 can modify NLP model 328 by creating new nodes representing new words, adding edges between existing nodes, etc. The decision mechanism to output an entity-value pair (e.g., a parameterized equation) can be also be updated (e.g., by updating the parameters) to increase the likelihood of outputting the correct entity-pair and/or context information.”
Also, claim 6 “6. The method of claim 5 wherein the parameters of the NLP model are modified in response to determining that the NLP model incorrectly extracts entity/value pairs and/or context information, wherein the parameters are modified by at least one of adding a new terminology to the NLP model or new relationship between entities of the NLP model.” Also, see Fig. 4A);
grouping two or more of the entities into a group based on one or more of the entity types of one or more of the entities, and one or more of the relationships between two or more of the entities(para, “[0060] Referring back to FIG. 3, NLP model 328 can be a machine-learning model that is trained. As shown in FIG. 3, system 300 may include a training module 340, which can train NLP model 328. Training module 340 can train NLP model 328 based on labeled general medical documents 348 and labeled pathology report 350. General medical documents 348 can include various categories of biomedical literatures, reports, etc. The training create nodes representing words of medical terminologies, as well as edges representing the sequential relationships among the words, such as those of NLP model 328 of FIG. 4A. As part of the training operation, a sequence of text strings with a particular label (e.g., a labeled entity, a labeled entity value, a labeled context) can be input to NLP model 328 to determine whether NLP outputs correct entity-value pairs and/or context information. If training module 340 determines that NLP model 328 does not output correct entity-value pairs and/or context information (e.g., based on comparing the labelled entity/entity value of the sequence of text strings and the entity-value pairs output by the NLP model for the sequence of text strings), the training module 340 can modify NLP model 328 by creating new nodes representing new words, adding edges between existing nodes, etc. The decision mechanism to output an entity-value pair (e.g., a parameterized equation) can be also be updated (e.g., by updating the parameters) to increase the likelihood of outputting the correct entity-pair and/or context information.”
Also, see Fig. 4B (para 0051, “…. nodes 440e and 440f are associated with different organs to receive the surgery, such as heart and breast.”));
and training a first machine-learning model with one or more of the plurality of relevant case reports, the extracted relevant text, the entities, and the relationships(para, “[0060] Referring back to FIG. 3, NLP model 328 can be a machine-learning model that is trained. As shown in FIG. 3, system 300 may include a training module 340, which can train NLP model 328. Training module 340 can train NLP model 328 based on labeled general medical documents 348 and labeled pathology report 350. General medical documents 348 can include various categories of biomedical literatures, reports, etc. The training create nodes representing words of medical terminologies, as well as edges representing the sequential relationships among the words, such as those of NLP model 328 of FIG. 4A. As part of the training operation, a sequence of text strings with a particular label (e.g., a labeled entity, a labeled entity value, a labeled context) can be input to NLP model 328 to determine whether NLP outputs correct entity-value pairs and/or context information. If training module 340 determines that NLP model 328 does not output correct entity-value pairs and/or context information (e.g., based on comparing the labelled entity/entity value of the sequence of text strings and the entity-value pairs output by the NLP model for the sequence of text strings), the training module 340 can modify NLP model 328 by creating new nodes representing new words, adding edges between existing nodes, etc. The decision mechanism to output an entity-value pair (e.g., a parameterized equation) can be also be updated (e.g., by updating the parameters) to increase the likelihood of outputting the correct entity-pair and/or context information.”)
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.
Claims 5-8 are rejected under 35 U.S.C. 103 as being unpatentable over SHARMA in view of Masood et al. (US 20210034652 A1)
Regarding claim 5, SHARMA teaches the method according to claim 1.
SHARMA does not explicitly teach wherein grouping the two or more of the entities into the group comprises:
identifying a head entity for the group;
and identifying one or more child entities for the group.
Masood teaches :
identifying a head entity for the group(para, “[0048] In some implementations, the ontology server 102 selects parent and child nodes based on relative importance and recurrence of topics within the unstructured document. For example, the importance can be determined by frequency of occurrence of terms within the unstructured document, as well as a top level hierarchical influence the terms have on an overall corpus. For example, terms with highest frequency of occurrence can be indicated as parent nodes. Similarly, terms covering large clusters in the DSSS (having a top level hierarchical influence) can be indicated as parent nodes. Child nodes can be populated using a similar approach applied in reduced order of dependency. That is, terms with a reduced order of dependency to the other topics become corresponding child nodes. The relationship between a parent and child node is based on the domain specific sensitivity mapping.”)
and identifying one or more child entities for the group(para, “[0048] In some implementations, the ontology server 102 selects parent and child nodes based on relative importance and recurrence of topics within the unstructured document. For example, the importance can be determined by frequency of occurrence of terms within the unstructured document, as well as a top level hierarchical influence the terms have on an overall corpus. For example, terms with highest frequency of occurrence can be indicated as parent nodes. Similarly, terms covering large clusters in the DSSS (having a top level hierarchical influence) can be indicated as parent nodes. Child nodes can be populated using a similar approach applied in reduced order of dependency. That is, terms with a reduced order of dependency to the other topics become corresponding child nodes. The relationship between a parent and child node is based on the domain specific sensitivity mapping.”)
It would have been obvious for a person of ordinary skill in the art to apply head and child identifying teachings of Masood into the teachings of SHARMA at the time the application was filed in order to define more frequent/important occurrence of terms. Masood, para, (“[0048] In some implementations, the ontology server 102 selects parent and child nodes based on relative importance and recurrence of topics within the unstructured document. For example, the importance can be determined by frequency of occurrence of terms within the unstructured document, as well as a top level hierarchical influence the terms have on an overall corpus. For example, terms with highest frequency of occurrence can be indicated as parent nodes. Similarly, terms covering large clusters in the DSSS (having a top level hierarchical influence) can be indicated as parent nodes. Child nodes can be populated using a similar approach applied in reduced order of dependency. That is, terms with a reduced order of dependency to the other topics become corresponding child nodes. The relationship between a parent and child node is based on the domain specific sensitivity mapping.”)
Regarding claim 6, SHARMA as modified by Masood teaches the method according to claim 5.
SHARMA as modified by Masood does not explicitly teach wherein identifying the head entity for the group comprises identifying one of a plurality of entities with an entity type of greater priority than an entity type of a related plurality of entities.
Masood further teaches wherein identifying the head entity for the group comprises identifying one of a plurality of entities with an entity type of greater priority than an entity type of a related plurality of entities([0048] In some implementations, the ontology server 102 selects parent and child nodes based on relative importance and recurrence of topics within the unstructured document. For example, the importance can be determined by frequency of occurrence of terms within the unstructured document, as well as a top level hierarchical influence the terms have on an overall corpus. For example, terms with highest frequency of occurrence can be indicated as parent nodes. Similarly, terms covering large clusters in the DSSS (having a top level hierarchical influence) can be indicated as parent nodes. Child nodes can be populated using a similar approach applied in reduced order of dependency. That is, terms with a reduced order of dependency to the other topics become corresponding child nodes. The relationship between a parent and child node is based on the domain specific sensitivity mapping.”)
It would have been obvious for a person of ordinary skill in the art to apply head identifying teachings of Masood into the teachings of SHARMA as modified by Masood at the time the application was filed in order to define more frequent/important occurrence of terms. Masood, para, (“[0048] In some implementations, the ontology server 102 selects parent and child nodes based on relative importance and recurrence of topics within the unstructured document. For example, the importance can be determined by frequency of occurrence of terms within the unstructured document, as well as a top level hierarchical influence the terms have on an overall corpus. For example, terms with highest frequency of occurrence can be indicated as parent nodes. Similarly, terms covering large clusters in the DSSS (having a top level hierarchical influence) can be indicated as parent nodes. Child nodes can be populated using a similar approach applied in reduced order of dependency. That is, terms with a reduced order of dependency to the other topics become corresponding child nodes. The relationship between a parent and child node is based on the domain specific sensitivity mapping.”)
Regarding claim 7, SHARMA as modified by Masood teaches the method according to claim 5.
SHARMA as modified by Masood does not explicitly teach wherein identifying the child entities for the group comprises identifying one or more of the plurality of entities with an entity type of lower priority than an entity type of a related plurality of entities.
Masood further teaches wherein identifying the child entities for the group comprises identifying one or more of the plurality of entities with an entity type of lower priority than an entity type of a related plurality of entities( para, “[0048] In some implementations, the ontology server 102 selects parent and child nodes based on relative importance and recurrence of topics within the unstructured document. For example, the importance can be determined by frequency of occurrence of terms within the unstructured document, as well as a top level hierarchical influence the terms have on an overall corpus. For example, terms with highest frequency of occurrence can be indicated as parent nodes. Similarly, terms covering large clusters in the DSSS (having a top level hierarchical influence) can be indicated as parent nodes. Child nodes can be populated using a similar approach applied in reduced order of dependency. That is, terms with a reduced order of dependency to the other topics become corresponding child nodes. The relationship between a parent and child node is based on the domain specific sensitivity mapping.”)
It would have been obvious for a person of ordinary skill in the art to apply child identifying teachings of Masood into the teachings of SHARMA as modified by Masood at the time the application was filed in order to define less frequent occurrence of terms. Masood, para, (“[0048] In some implementations, the ontology server 102 selects parent and child nodes based on relative importance and recurrence of topics within the unstructured document. For example, the importance can be determined by frequency of occurrence of terms within the unstructured document, as well as a top level hierarchical influence the terms have on an overall corpus. For example, terms with highest frequency of occurrence can be indicated as parent nodes. Similarly, terms covering large clusters in the DSSS (having a top level hierarchical influence) can be indicated as parent nodes. Child nodes can be populated using a similar approach applied in reduced order of dependency. That is, terms with a reduced order of dependency to the other topics become corresponding child nodes. The relationship between a parent and child node is based on the domain specific sensitivity mapping.”)
Regarding claim 8, SHARMA as modified by Masood teaches the method according to claim 5.
SHARMA as modified by Masood does not explicitly teach wherein identifying the child entities for the group comprises identifying one or more of the plurality of entities not identified as the head entity.
Masood further teaches wherein identifying the child entities for the group comprises identifying one or more of the plurality of entities not identified as the head entity(para, “[0048] In some implementations, the ontology server 102 selects parent and child nodes based on relative importance and recurrence of topics within the unstructured document. For example, the importance can be determined by frequency of occurrence of terms within the unstructured document, as well as a top level hierarchical influence the terms have on an overall corpus. For example, terms with highest frequency of occurrence can be indicated as parent nodes. Similarly, terms covering large clusters in the DSSS (having a top level hierarchical influence) can be indicated as parent nodes. Child nodes can be populated using a similar approach applied in reduced order of dependency. That is, terms with a reduced order of dependency to the other topics become corresponding child nodes. The relationship between a parent and child node is based on the domain specific sensitivity mapping.”)
It would have been obvious for a person of ordinary skill in the art to apply child identifying teachings of Masood into the teachings of SHARMA as modified by Masood at the time the application was filed in order to define frequency of occurrence of terms. Masood, para, (“[0048] In some implementations, the ontology server 102 selects parent and child nodes based on relative importance and recurrence of topics within the unstructured document. For example, the importance can be determined by frequency of occurrence of terms within the unstructured document, as well as a top level hierarchical influence the terms have on an overall corpus. For example, terms with highest frequency of occurrence can be indicated as parent nodes. Similarly, terms covering large clusters in the DSSS (having a top level hierarchical influence) can be indicated as parent nodes. Child nodes can be populated using a similar approach applied in reduced order of dependency. That is, terms with a reduced order of dependency to the other topics become corresponding child nodes. The relationship between a parent and child node is based on the domain specific sensitivity mapping.”)
Claims 9 is rejected under 35 U.S.C. 103 as being unpatentable over SHARMA in view of Srivastava et al. (US 20130254178 A1)
Regarding claim 9, SHARMA teaches the method according to claim 1.
SHARMA further teaches wherein [the database comprises the OVIDTM database and] the published case reports comprise medical case reports(para, “[0004] Disclosed herein are techniques for automated information extraction and enrichment in pathology reports. The pathology reports can include electronic reports from various primary sources (e.g., at one or more healthcare institutions) including, for example, an EMR (electronic medical record) database, a PACS (picture archiving and communication system), a Digital Pathology (DP) system, an LIS (laboratory information system) including genomic data, an RIS (radiology information system), patient reported outcomes database, wearable and/or digital technologies, and social media. The pathology reports can also be in paper form and originate from the clinician/clinician staff. The pathology reports can be in the form of image files (e.g., Portable Document Format (pdf), bitmap image file (BMP file)) obtained by scanning the paper-form pathology reports.”)
SHARMA does not explicitly teach wherein the database comprises the OVIDTM database [and the published case reports comprise medical case reports]
Srivastava teaches wherein the database comprises the OVIDTM database [and the published case reports comprise medical case reports](para, “[0005] Physicians, researchers, and patients often analyze medical literature to learn about the efficacy and results of various patient clinical studies. For example, a physician treating a patient with breast cancer may analyze medical literature to glean best treatment practices used in studies for treating breast cancer patients with similar disease profiles (e.g., stage, type, histology, etc . . . ). Accordingly, to meet those needs, the details of many medical studies often are published as an article in a medical journal, such as the widely known New England Journal of Medicine, and available to the public from any of a number of publicly available and private data stores (e.g., Ovid, Embase, Cinahl, Medline or Pubmed).”)
It would have been obvious for a person of ordinary skill in the art to incorporate OVIDTM database teachings of Srivastava into the teachings of SHARMA at the time the application was filed to help analyze medical literature to learn about the efficacy and results of various patient clinical studies. Srivastava, para “para, “[0005] Physicians, researchers, and patients often analyze medical literature to learn about the efficacy and results of various patient clinical studies. For example, a physician treating a patient with breast cancer may analyze medical literature to glean best treatment practices used in studies for treating breast cancer patients with similar disease profiles (e.g., stage, type, histology, etc . . . ). Accordingly, to meet those needs, the details of many medical studies often are published as an article in a medical journal, such as the widely known New England Journal of Medicine, and available to the public from any of a number of publicly available and private data stores (e.g., Ovid, Embase, Cinahl, Medline or Pubmed).”)
Claims 10-12 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over SHARMA in view of Aldairy et al. (US 20190272907 A1)
Regarding claim 10, SHARMA teaches the method according to claim 1.
SHARMA does not explicitly teach:
generating a confidence score for each of the published case reports with a first machine-learning model
and identifying the published case reports with a confidence score above a threshold confidence interval.
Aldairy teaches:
generating a confidence score for each of the published case reports with a first machine-learning model(para, “[0067]….. NLP filter 44 can be configured to analyze this unstructured AE data to detect natural language characteristics of the input and determine a confidence score for the distinction (or similarity) between the subsequent unstructured reported AE data 40A and the unstructured reported AE data 40. For example, NLP filter 44 can assign a confidence score to the distinctions (or similarities) between the subsequent unstructured reported AE data 40A and the unstructured reported AE data 40 using a conventional F-score approach.”
Also, para, “[0058] NLP filter 44 is also configured to assign a confidence score in its matching of natural language phrases 52 with AE reporting codes 54. That is, according to various embodiments, NLP algorithm 56 may have scores assigned to particular relationships between natural language terms and symptoms. For example, a term such as “dragging,” could be tied with “fatigue,” but could also be tied with “drowsiness.” As such, a code match for “dragging” with the symptom Fatigue could be given a lower confidence score than a code match for “exhausted” with Fatigue. A term such as “sleepy” could have a higher confidence score for the symptom Drowsiness than would the term “dragging.” These confidence scores can be indicated in the initial reporting codes 58, and certain threshold confidence scores (e.g., below level X) can be flagged for additional or special review by healthcare professional 14. In various embodiments, NLP algorithm 56 can take the form of a machine learning algorithm, e.g., a decision tree, naïve Bayesian algorithm and/or a logit algorithm.”;
and identifying the published case reports with a confidence score above a threshold confidence interval(para, “[0058] NLP filter 44 is also configured to assign a confidence score in its matching of natural language phrases 52 with AE reporting codes 54. That is, according to various embodiments, NLP algorithm 56 may have scores assigned to particular relationships between natural language terms and symptoms. For example, a term such as “dragging,” could be tied with “fatigue,” but could also be tied with “drowsiness.” As such, a code match for “dragging” with the symptom Fatigue could be given a lower confidence score than a code match for “exhausted” with Fatigue. A term such as “sleepy” could have a higher confidence score for the symptom Drowsiness than would the term “dragging.” These confidence scores can be indicated in the initial reporting codes 58, and certain threshold confidence scores (e.g., below level X) can be flagged for additional or special review by healthcare professional 14. In various embodiments, NLP algorithm 56 can take the form of a machine learning algorithm, e.g., a decision tree, naïve Bayesian algorithm and/or a logit algorithm.”)
It would have been obvious for a person of ordinary skill in the art to apply confidence score teachings of Aldairy into the teachings of SHARMA at the time the application was filed in order to specify particular relationships between natural language terms and symptoms. Aldairy, para “[0058] NLP filter 44 is also configured to assign a confidence score in its matching of natural language phrases 52 with AE reporting codes 54. That is, according to various embodiments, NLP algorithm 56 may have scores assigned to particular relationships between natural language terms and symptoms. For example, a term such as “dragging,” could be tied with “fatigue,” but could also be tied with “drowsiness.” As such, a code match for “dragging” with the symptom Fatigue could be given a lower confidence score than a code match for “exhausted” with Fatigue. A term such as “sleepy” could have a higher confidence score for the symptom Drowsiness than would the term “dragging.” These confidence scores can be indicated in the initial reporting codes 58, and certain threshold confidence scores (e.g., below level X) can be flagged for additional or special review by healthcare professional 14. In various embodiments, NLP algorithm 56 can take the form of a machine learning algorithm, e.g., a decision tree, naïve Bayesian algorithm and/or a logit algorithm.”)
Regarding claim 11, SHARMA as modified by Aldairy teaches the method according to claim 10.
SHARMA as modified by Aldairy does not explicitly teach wherein generating the confidence interval comprises:
identifying an abstract of each of the published case reports;
and generating the confidence score based at least in part from the identified abstract of each of the published case reports.
Aldairy further teaches:
identifying an abstract of each of the published case reports(para, “[0073] FIG. 8 shows an example depiction of structured reported AE data 42, in the form of a section from a fillable severe adverse event (SAE) reporting form 800, used to report severe adverse events for particular pharmaceutical, vaccine or medical device clinical trials. As shown, the SAE reporting form 800 includes fillable sections 802 for providing information about the subject (patient), such as personal identifying information including subject, height, weight, date-of-birth, race, etc. Fillable sections 802 can also be designed to include event-specific data 804, such as Event Term (e.g., hemorrhaging in the abdomen), Onset Date, Date of Resolution, Serious Criteria, Relationship to Study Drug, Grade (e.g., Common Terminology Criteria for Adverse Events, CTCAE criteria), and Outcome. Fillable sections 802 can be organized by particular headings 806 in the AE data 42. In some cases, particular event-specific data 804 is scored or ranked according to particular reporting criteria. For example, a particular event, such as hemorrhaging in the abdomen, could be classified as “Life-threatening” (score of 2, with 1 being most severe) when it required hospitalization, but did not cause the patient to die. With reference to FIG. 6, the OCR module 46 is configured to identify the terminology in the fillable sections 802, including the event-specific data 804, and select AE reporting codes 54 for that particular event-specific data 804…”);
and generating the confidence score based at least in part from the identified abstract of each of the published case reports((para, “[0073] ….. In some cases, particular event-specific data 804 is scored or ranked according to particular reporting criteria. For example, a particular event, such as hemorrhaging in the abdomen, could be classified as “Life-threatening” (score of 2, with 1 being most severe) when it required hospitalization, but did not cause the patient to die. With reference to FIG. 6, the OCR module 46 is configured to identify the terminology in the fillable sections 802, including the event-specific data 804, and select AE reporting codes 54 for that particular event-specific data 804…” )
It would have been obvious for a person of ordinary skill in the art to apply generating score based on abstract teachings of Aldairy into the teachings of SHARMA as modified by Aldairy at the time the application was filed in order to report severe adverse events. Aldairy, para (“[0073] FIG. 8 shows an example depiction of structured reported AE data 42, in the form of a section from a fillable severe adverse event (SAE) reporting form 800, used to report severe adverse events for particular pharmaceutical, vaccine or medical device clinical trials……… In some cases, particular event-specific data 804 is scored or ranked according to particular reporting criteria. For example, a particular event, such as hemorrhaging in the abdomen, could be classified as “Life-threatening” (score of 2, with 1 being most severe) when it required hospitalization, but did not cause the patient to die…..”)
Regarding claim 12, SHARMA as modified by Aldairy teaches the method according to claim 10.
SHARMA further teaches wherein the first machine-learning model comprises a trained natural language processing (NLP) model(para, “[0060] Referring back to FIG. 3, NLP model 328 can be a machine-learning model that is trained. As shown in FIG. 3, system 300 may include a training module 340, which can train NLP model 328. Training module 340 can train NLP model 328 based on labeled general medical documents 348 and labeled pathology report 350….”)
Regarding claim 20, SHARMA teaches the method according to claim 18.
SHARMA does not explicitly teach wherein identifying the plurality of relevant case reports comprises:
generating a confidence score for each of the published case reports with a second machine-learning model;
and identifying the published case reports with a confidence score above a threshold confidence interval.
Aldairy teaches:
generating a confidence score for each of the published case reports with a second machine-learning model para, “[0067]….. NLP filter 44 can be configured to analyze this unstructured AE data to detect natural language characteristics of the input and determine a confidence score for the distinction (or similarity) between the subsequent unstructured reported AE data 40A and the unstructured reported AE data 40. For example, NLP filter 44 can assign a confidence score to the distinctions (or similarities) between the subsequent unstructured reported AE data 40A and the unstructured reported AE data 40 using a conventional F-score approach.”
Also, para, “[0058] NLP filter 44 is also configured to assign a confidence score in its matching of natural language phrases 52 with AE reporting codes 54. That is, according to various embodiments, NLP algorithm 56 may have scores assigned to particular relationships between natural language terms and symptoms. For example, a term such as “dragging,” could be tied with “fatigue,” but could also be tied with “drowsiness.” As such, a code match for “dragging” with the symptom Fatigue could be given a lower confidence score than a code match for “exhausted” with Fatigue. A term such as “sleepy” could have a higher confidence score for the symptom Drowsiness than would the term “dragging.” These confidence scores can be indicated in the initial reporting codes 58, and certain threshold confidence scores (e.g., below level X) can be flagged for additional or special review by healthcare professional 14. In various embodiments, NLP algorithm 56 can take the form of a machine learning algorithm, e.g., a decision tree, naïve Bayesian algorithm and/or a logit algorithm.”;
and identifying the published case reports with a confidence score above a threshold confidence interval(para, “[0058] NLP filter 44 is also configured to assign a confidence score in its matching of natural language phrases 52 with AE reporting codes 54. That is, according to various embodiments, NLP algorithm 56 may have scores assigned to particular relationships between natural language terms and symptoms. For example, a term such as “dragging,” could be tied with “fatigue,” but could also be tied with “drowsiness.” As such, a code match for “dragging” with the symptom Fatigue could be given a lower confidence score than a code match for “exhausted” with Fatigue. A term such as “sleepy” could have a higher confidence score for the symptom Drowsiness than would the term “dragging.” These confidence scores can be indicated in the initial reporting codes 58, and certain threshold confidence scores (e.g., below level X) can be flagged for additional or special review by healthcare professional 14. In various embodiments, NLP algorithm 56 can take the form of a machine learning algorithm, e.g., a decision tree, naïve Bayesian algorithm and/or a logit algorithm.”)
It would have been obvious for a person of ordinary skill in the art to apply confidence score teachings of Aldairy into the teachings of SHARMA at the time the application was filed in order to specify particular relationships between natural language terms and symptoms. Aldairy, para “[0058] NLP filter 44 is also configured to assign a confidence score in its matching of natural language phrases 52 with AE reporting codes 54. That is, according to various embodiments, NLP algorithm 56 may have scores assigned to particular relationships between natural language terms and symptoms. For example, a term such as “dragging,” could be tied with “fatigue,” but could also be tied with “drowsiness.” As such, a code match for “dragging” with the symptom Fatigue could be given a lower confidence score than a code match for “exhausted” with Fatigue. A term such as “sleepy” could have a higher confidence score for the symptom Drowsiness than would the term “dragging.” These confidence scores can be indicated in the initial reporting codes 58, and certain threshold confidence scores (e.g., below level X) can be flagged for additional or special review by healthcare professional 14. In various embodiments, NLP algorithm 56 can take the form of a machine learning algorithm, e.g., a decision tree, naïve Bayesian algorithm and/or a logit algorithm.”)
Claims 14-16 are rejected under 35 U.S.C. 103 as being unpatentable over SHARMA in view of Bhatia et al. (US 11556579 B1)
Regarding claim 14, SHARMA teaches the method according to claim 1.
SHARMA does not explicitly teach wherein generating the plurality of entities comprises:
generating a plurality of tokens from the relevant text
and generating an entity type for each of the tokens using a second machine-learning model.
Bhatia teaches :
generating a plurality of tokens from the relevant text(col 9, line 48 “In some embodiments, each identified segment of text and metadata identifying the tokens therein is provided, by the orchestrator 116, to multiple ML models 120A-120C according to a “scatter” type technique as reflected by circle (3). Each of the ML models 120A-120C may be trained to detect a particular entity type from within unstructured text, and in some cases, ones of the models 120 may be executed in parallel for a same segment or group of segments. In this example, the orchestrator 116 is shown as utilizing three models 120A-120C, though in other embodiments more or fewer models (e.g., via more or fewer services 118A-118C, respectively) may be used.”
Also, (col 17, line 46 “In some embodiments, the TSE 125 then tokenizes the segments to identify one or more “tokens” within each segment. A token may be a word or a grouping of characters, and the tokenization may be performed by applying another set of rules that indicate where the segment is to be split. For example, the tokenization may include identifying the existence of spaces, tabs, column delimiters (such as pipes or colons), etc., to thus identify the beginning and end of a token. Thus, the TSE 125 may generate token metadata indicating a beginning location and an ending location of a token (within a segment) or a character length.);
and generating an entity type for each of the tokens using a second machine-learning model(col 19, line 47 “In some embodiments, each identified segment of text (e.g., and metadata identifying the tokens therein) is provided, by the orchestrator 116, to medical condition ML model 120B to detect at circle (4B) a medical classification entity type (e.g., symptoms and diagnosis of medical conditions) from within unstructured text and to an anatomy ML model 120C to detect at circle (4C) an anatomy entity type (e.g., references to the anatomical parts of the body or body systems and/or the locations of those parts or systems), and in some cases, the models 120B and 120C may be executed in parallel for a same segment or group of segments, e.g., for requests sent in parallel at circle (3).”)
It would have been obvious for a person of ordinary skill in the art to incorporate token generation teachings of Bhaita into the teachings of SHARMA at the time the application was filed in order to identify particular entities and relationships. Bhatia, Abstract “Techniques for ontology linking of unstructured text as a service are described. A service may receive a request to link unstructured text to a standardized ontology, and the service may segment and tokenize the unstructured text and send the result to multiple services implementing multiple deep machine learning models trained to identify particular entities and one or more relationships between entities. The service may perform a search of the standardized ontology to identify a set of similar candidates from the standardized ontology for the detected entities and the one or more relationships, and then rank the set of similar candidates from the standardized ontology according to their similarity to the detected entities within the unstructured text…...”)
Regarding claim 15, SHARMA teaches the method according to claim 1.
SHARMA does not explicitly teach wherein predicting the relationship comprises generating the relationship using a third machine-learning model.
Bhatia teaches wherein predicting the relationship comprises generating the relationship using a third machine-learning model(col 4, line 34 “For further detail, FIG. 1 is a diagram illustrating an environment for synchronous entity and relationship detection from unstructured text according to some embodiments. In this exemplary environment, an ontology linking service 112 includes an orchestrator 116 that receives requests 130 for ontology linking of unstructured text and utilizes multiple ML models 120,124, (e.g., of a set of one or more unstructured text services 114) trained to detect particular entities and/or relationships between detected entities. In certain embodiments, the orchestrator 116 utilizes an ontology search service 126 to search for a set of similar candidates (e.g., from medical ontology 128) based on the entities and relationships. In certain embodiments, the orchestrator 116 utilizes ML model 132 to rank the set of candidates (e.g., concepts therein) of an ontology based on their similarity to entities and relationships (e.g., concepts therein).
It would have been obvious for a person of ordinary skill in the art to apply generate relationship using machine learning model teachings of Bhatia into the teachings of SHARMA at the time the application was filed in order to detect particular entities and/or relationships between detected entities. Bhatia, col 4, line 37 “an ontology linking service 112 includes an orchestrator 116 that receives requests 130 for ontology linking of unstructured text and utilizes multiple ML models 120,124, (e.g., of a set of one or more unstructured text services 114) trained to detect particular entities and/or relationships between detected entities. In certain embodiments, the orchestrator 116 utilizes an ontology search service 126 to search for a set of similar candidates (e.g., from medical ontology 128) based on the entities and relationships”)
Regarding claim 16, SHARMA as modified by Bhatia teaches the method according to claim 15.
SHARMA as modified by Bhatia does not explicitly teach wherein predicting the relationship comprises, for one or more pairs of the entities:
generating a relationship confidence score for each of the pairs of entities with the third machine-learning model;
and predicting the relationship between each of the pairs of entities based at least in part on the relationship confidence score corresponding to the pair of entities
Bhatia further teaches :
generating a relationship confidence score for each of the pairs of entities with the third machine-learning model(col 13, line 15 “For further detail, FIG. 3 is a diagram illustrating an abbreviated example result 300 including detected entities and relationships according to some embodiments. For example, in some embodiments the result includes an entry or node for each detected entity. Each Entity may include an array of Attributes extracted that relate to the entity, a BeginOffset integer that provides the 0-based character offset in the input text that shows where the entity begins, a string Category indicating what type the entity is (e.g., MEDICATION, MEDICAL_CONDITION, PROTECTED_HEALTH_INFORMATION, TEST_TREATMENT_PROCEDURE, ANATOMY, which correspond to models 118/services 118), an EndOffset integer that provides the 0-based character offset in the input text that shows where the entity ends, an Id integer that is a monotonically increasing identifier unique within this response rather than a global unique identifier, a Score float that indicates a level of confidence that the ontology linking service has in the accuracy of the detection (based on an accuracy/confidence score provided by the respective detecting model), a Text string indicating the segment of input text extracted as this entity, an array of Traits providing contextual information for the entity, a Type string describing the specific type of entity.
and predicting the relationship between each of the pairs of entities based at least in part on the relationship confidence score corresponding to the pair of entities(col 4, line 34 “For further detail, FIG. 1 is a diagram illustrating an environment for synchronous entity and relationship detection from unstructured text according to some embodiments. In this exemplary environment, an ontology linking service 112 includes an orchestrator 116 that receives requests 130 for ontology linking of unstructured text and utilizes multiple ML models 120,124, (e.g., of a set of one or more unstructured text services 114) trained to detect particular entities and/or relationships between detected entities. In certain embodiments, the orchestrator 116 utilizes an ontology search service 126 to search for a set of similar candidates (e.g., from medical ontology 128) based on the entities and relationships. In certain embodiments, the orchestrator 116 utilizes ML model 132 to rank the set of candidates (e.g., concepts therein) of an ontology based on their similarity to entities and relationships (e.g., concepts therein).)
It would have been obvious for a person of ordinary skill in the art to apply relationship confidence score teachings of Bhatia into the teachings of SHARMA as modified by Bhatia at the time the application was filed in order to indicate that the attribute is correctly related to the particular entity. (Bhatia, col 13, line 39 “ Each Attribute may similarly include a BeginOffset integer, an EndOffset integer, an Id, a RelationshipScore float indicating a level of confidence that the ontology linking service has that this attribute is correctly related to the particular entity, a Score float indicating the level of confidence that ontology linking service has that the segment of text is correctly recognized as an attribute, a Text string, an array of Traits, a Type string, etc. Each Trait may include, for example, a Name string providing a name or contextual description about the trait, a Score float indicating a level of confidence that ontology linking service 112 has in the accuracy of this trait, etc.”)
Claim 19 is rejected under 35 U.S.C. 103 as being unpatentable over SHARMA in view of MUNNURU et al. (US 20250246281 A1)
Regarding claim 19, SHARMA teaches the method according to claim 18.
SHARMA does not explicitly teach wherein the first machine-learning model comprises a BioBERTTM natural language processing model.
MUNNURU teaches wherein the first machine-learning model comprises a BioBERTTM natural language processing model(para, “[0077] FIG. 11 depicts a process of performing AI/NLP based vocabulary categorization from electronic medical record/electronic health record (EMR/EHR). In step 1101, a training data set is trained using at least one clinically relevant use case, wherein the data set is a selected set of EMR/EHR data. Using the training data set, in step 1102, one or more categories are extracted for entities. In an embodiment herein, the one or more categories can be extracted based on Observational Medical Outcomes Partnership (OMOP). In step 1103, a classifier is trained using the extracted categories. In an example herein, Google BioBert model classifier can be trained. In step 1104, the results of the classifier training can be corrected (if required). In an embodiment herein, the results of the classifier training can be corrected manually. In an embodiment herein, the corrected results can be used as ground truth (GT) for the classifier training. In step 1105, the corrected results are validated using manual ground truth. ….”)
It would have been obvious for a person of ordinary skill in the art to apply BioBERTTM natural language processing model teachings of MUNNURU into the teachings of SHARMA at the time the application was filed in order to utilize well known learning technique used by Google( MUNNURU, para “[0077]….In an example herein, Google BioBert model classifier can be trained……”)
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US 20240177818 A1 “Various methods and systems are provided for generating and displaying summaries of patient information extracted from one or more medical reports stored in an electronic medical record (EMR) of a patient. In one embodiment, a method for summarizing medical reports includes, receiving a medical report for a patient, classifying the medical report into a category of a plurality of pre-determined categories, matching the medical report with an entity recognition model from a library of entity recognition models based on the category, identifying a plurality of named entities in the medical report using the entity recognition model, refining the plurality of named entities to produce a summary of the medical report, and displaying the summary of the medical report via a display device.”
Any inquiry concerning this communication or earlier communications from the examiner should be directed to HUMA WASEEM whose telephone number is (571)272-1316. The examiner can normally be reached Monday-Friday(9:00 am - 5 pm) EST.
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, Jason B. Dunham can be reached on (571) 272-8109. 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.
/HUMA WASEEM/Examiner, Art Unit 3686
/JASON B DUNHAM/Supervisory Patent Examiner, Art Unit 3686