WO2023242879A1

WO2023242879A1 - System and method for generating prioritized recommendations using a probabilistic causation model with predicted anomaly

Info

Publication number: WO2023242879A1
Application number: PCT/IN2023/050579
Authority: WO
Inventors: Anmol SINGH; Kumar SHUBHAM; Ankush Patel; Lavish MAHESHWARI; Sanskriti A; Singamsetty Sanjeeva Krishna SAI DINESH; Abhishek Joshi; Ima RASHID; Harshit AGRAWAL; Sreevidya K V; Mayank NEGI; Chaitanya BHARADWAJ; Abdussamad G M
Original assignee: Apollo Healthco Limited
Priority date: 2022-06-17
Filing date: 2023-06-17
Publication date: 2023-12-21

Abstract

The system and a computer-implemented method for generating prioritized recommendations for a user using a probabilistic causation model with at least one predicted anomaly is provided. The method includes (a) receiving text-based input from the user comprising user symptoms and medical history, (b) performing automated adaptive query generation at the user interface recording responses, (c) predicting an anomaly using a Bayesian probabilistic causation model, wherein training for the Bayesian probabilistic causation model is performed by: (i) obtaining real-world data of patients and clinical data comprising symptoms, historical pathways and outcomes, (ii) determining probabilities associated with the historical pathways and outcomes and (iii) dynamically constructing a knowledge graph capturing relationships between symptoms and anomalies, (d) generating prioritized recommendation based on the predicted anomaly using the Bayesian probabilistic causation model, (e) transmitting, the prioritized recommendation to the user.

Description

SYSTEM AND METHOD FOR GENERATING PRIORITIZED RECOMMENDATIONS USING A PROBABILISTIC CAUSATION MODEL WITH PREDICTED ANOMALY BACKGROUND

Technical Field

[0001] The embodiments herein generally relate to artificial intelligence (Al) and Natural Language Processing (NLP), and more particularly, to a system and method for generating prioritized recommendations for a user using a probabilistic causation model with at least one predicted anomaly for decision support and providing recommendations to users and/or medical professionals.

Description of the Related Art

[0002] One of the dominant discourses in the public health domain within context of provision of Universal Health Coverage is the shortage of adequate number of qualified medical doctors and other healthcare professionals. The World Health Organization (WHO) has recommended a doctor-population ratio of 1 : 1 ,000. Despite this, over 44% of WHO Member States have less than one physician per 1,000 people. Further, these doctors are also pressed for time to offer consultations to patients. Offering consultations in person has become more challenging when the patient may be suffering from an infectious disease that can be transmitted to the doctor. Telemedicine, or remote medical support, allows a patient to consult with a doctor virtually using text, email, voice and/or video conferencing. This has helped to alleviate the problem by reducing the requirement for queuing up at a clinic or hospital, commute time, etc., but it still requires a doctor to spend significant amount of time asking questions and gathering information from the patient in order to understand the patient’s symptoms, and take decisions on next steps such as suggesting a test.

[0003] Technologies such as artificial intelligence and robotics have contributed significantly to bring about automation and reducing the requirement for skilled experts across various fields, especially when there is a shortage of skilled professionals.

[0004] Traditional solutions lack accurate recommendations and recommendations are often provided without considering the relative probabilities of different outcomes. This can lead to suboptimal guidance and ineffective decision-making. Another technical problem is the consideration of the context of the user and their medical history. The challenge lies in capturing and analyzing the relevant information from the symptoms of the user and medical history to generate contextually aware recommendations.

[0005] Accordingly, there remains a need to address the aforementioned technical problems using a system and method for generating prioritized recommendations for a user using a probabilistic causation model with at least one predicted anomaly.

SUMMARY

[0006] In view of the foregoing, according to a first aspect, there is provided a method for generating prioritized recommendations for a user using a probabilistic causation model with at least one predicted anomaly. The method comprising (a) receiving a text-based input from the user comprising user symptoms and medical history, (b) performing automated adaptive query generation for presenting queries to the user at the user interface using a text-based conversation and recording, from the user interface, a set of responses the queries, (c) predicting at least one anomaly by processing the set of responses using a Bayesian probabilistic causation model, wherein training for the Bayesian probabilistic causation model is performed by (i) obtaining at least one of real- world data of patients generated by a hospital or, and publicly available clinical data that comprises symptoms, historical pathways of patients and outcomes, (ii) determining probabilities associated with the historical pathways and outcomes, and (iii) training the Bayesian probabilistic causation model by dynamically constructing a knowledge graph that captures relationships between the symptoms and anomalies as probabilities, (d) generating at least one prioritized recommendation based on the predicted anomaly using the Bayesian probabilistic causation model, and (e) transmitting, at the user interface, the at least one prioritized recommendation test to the user.

[0007] The method is of advantage that the method improves the accuracy of recommendations by employing an order of prioritization. By utilizing a probabilistic causation model, the method assigns probabilities to different anomalies based on symptoms of the user and medical history. The method takes into account the context of the user and medical history and generates recommendations that are personalized and take into account the specific context of the user.

[0008] Additionally, the method offers computational resource optimization by utilizing an adaptive query generation approach and utilizing the probabilistic causation model. The method captures relevant information from the user in a more efficient manner. By minimizing repetitive queries and gathering meaningful data with fewer interactions, computational resources are optimized. This not only improves the user experience by reducing the burden of multiple queries but also streamlines the recommendation generation process, allowing for faster and more resource-efficient outcomes. The method increases accuracy of recommendations through prioritization, considering the context of the user and their medical history, and optimize computational resources.

[0009] In some embodiments, the prioritized recommendation includes a specialist or a category of test, including a blood test, an imaging test, or a genetic test.

[0010] In some embodiments, the method further includes generating a reasoning for the prioritized recommendation by dynamically generating a graph of a plurality of paths based on the anomaly, demographic details, medical history of the user, and historically recommended tests, and determining probabilities associated with each of the plurality of paths within the graph for generating a reasoning for the prioritized recommendation.

[0011] In some embodiments, the method includes dynamically updating the recommendation of the diagnostic test upon obtaining new information associated with the user.

[0012] In some embodiments, the method further includes a test prediction Al model, wherein the test prediction Al model is configured to analyze a plurality of diagnostic tests and an interpretation of the plurality of diagnostic tests from a health records database and generate a diagnostic test recommendation for the user based on the analysis and interpretation of the diagnostic tests.

[0013] In another aspect, there is provided a system for generating prioritized recommendations for a user using a probabilistic causation model with at least one predicted anomaly, the system comprising a distributed cloud comprising a prioritized recommendation server that comprises a processor and a memory that are configured to perform (a) receiving a text-based input from the user comprising user symptoms and medical history, (b) performing automated adaptive query generation for presenting queries to the user at the user interface using a text-based conversation and recording, from the user interface, a set of responses the queries, (c) predicting at least one anomaly by processing the set of responses using a Bayesian probabilistic causation model, wherein training for the Bayesian probabilistic causation model is performed by (i) obtaining at least one of real- world data of patients generated by a hospital or, and publicly available clinical data that comprises symptoms, historical pathways of patients and outcomes, (ii) determining probabilities associated with the historical pathways and outcomes, and (iii) training the Bayesian probabilistic causation model by dynamically constructing a knowledge graph that captures relationships between the symptoms and anomalies as probabilities, (d) generating at least one prioritized recommendation based on the predicted anomaly using the Bayesian probabilistic causation model, and (e) transmitting, at the user interface, the at least one prioritized recommendation test to the user.

[0014] The system is of advantage that the system improves the accuracy of recommendations by employing an order of prioritization. By utilizing a probabilistic causation model, the system assigns probabilities to different anomalies based on symptoms of the user and medical history. The system takes into account the context of the user and medical history and generates recommendations that are personalized and take into account the specific context of the user.

[0015] Additionally, the system offers computational resource optimization by utilizing an adaptive query generation approach and utilizing the probabilistic causation model. The system captures relevant information from the user in a more efficient manner. By minimizing repetitive queries and gathering meaningful data with fewer interactions, computational resources are optimized. This not only improves the user experience by reducing the burden of multiple queries but also streamlines the recommendation generation process, allowing for faster and more resource-efficient outcomes. The system increases accuracy of recommendations through prioritization, considering the context of the user and their medical history, and optimize computational resources.

[0016] In some embodiments, the prioritized recommendation includes a specialist or a category of test, including a blood test, an imaging test, or a genetic test.

[0017] In some embodiments, the system further includes generating a reasoning for the prioritized recommendation by dynamically generating a graph of a plurality of paths based on the anomaly, demographic details, medical history of the user, and historically recommended tests, and determining probabilities associated with each of the plurality of paths within the graph for generating a reasoning for the prioritized recommendation.

[0018] In some embodiments, the system includes dynamically updating the recommendation of the diagnostic test upon obtaining new information associated with the user.

[0019] In some embodiments, the system further includes a test prediction Al model, wherein the test prediction Al model is configured to analyze a plurality of diagnostic tests and an interpretation of the plurality of diagnostic tests from a health records database and generate a diagnostic test recommendation for the user based on the analysis and interpretation of the diagnostic tests.

[0020] These and other aspects of the embodiments herein will be better appreciated and understood when considered in conjunction with the following description and the accompanying drawings. It should be understood, however, that the following descriptions, while indicating preferred embodiments and numerous specific details thereof, are given by way of illustration and not of limitation. Many changes and modifications may be made within the scope of the embodiments herein without departing from the spirit thereof, and the embodiments herein include all such modifications.

BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The embodiments herein will be better understood from the following detailed description with reference to the drawings, in which:

[0022] FIG. 1 illustrates a system for generating prioritized recommendations for a user using a probabilistic causation model with at least one predicted anomaly according to an embodiment herein;

[0023] FIG. 2 illustrates a flow diagram of generating a recommendation of a diagnostic test for a patient according to an embodiment herein;

[0024] FIG. 3 illustrates an exemplary process flow diagram of an assistance system for providing medical recommendations according to an embodiment herein;

[0025] FIG. 4 illustrates an exemplary system view of an assistance system according to an embodiment herein;

[0026] FIG. 5A illustrates an exploded view of an assistance system for providing medical recommendations based on user inputs using an Artificial Intelligence (Al) model according to an embodiment herein;

[0027] FIG. 5B illustrates an exploded view of a data processing unit of FIG. 2A according to an embodiment herein;

[0028] FIG. 5C illustrates an exemplary view of a parsing architecture of a data processing unit of FIG. 5 A according to an embodiment herein;

[0029] FIG. 6 illustrates an exemplary block diagram of a reasoning unit of FIG. 2 according to an embodiment herein; [0030] FIGS. 7A to 7E illustrate a user interface of an assistance system for assisting a patient at a patient device according to an embodiment herein;

[0031] FIGS. 8 A to 8G illustrate a user interface of an assistance system for providing medical assistance to a patient in a doctor consultation setting according to an embodiment herein;

[0032] FIGS. 9A and 9B illustrate a user interface of an assistance system for accessing patient data by a doctor for assisting a patient according to an embodiment herein;

[0033] FIGS. 10A-B is a flow diagram illustrating a computer-implemented method of generating automated adaptive queries to automatically determine a triage level according to an embodiment herein FIGS. 10A to 10C illustrate a report generated for a patient by the assistance system for providing medical recommendations;

[0034] FIGS. 11A-B is a flow diagram illustrating a computer-implemented method of generating a recommendation of a diagnostic test based on at least one predicted condition according to an embodiment herein;

[0035] FIG. 12 is a flow diagram illustrating a computer-implemented method for generating prioritized recommendations for a user using a probabilistic causation model with at least one predicted anomaly according to an embodiment herein.

[0036] FIG. 13 illustrates an exploded view of an assistance system of FIG. 4 according to an embodiment herein; and

[0037] FIG. 14 illustrates a schematic view of a hardware configuration of device management/ computer architecture according to an embodiment herein.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

[0038] The embodiments herein and the various features and advantageous details thereof are explained more fully with reference to the non-limiting embodiments that are illustrated in the accompanying drawings and detailed in the following description. Descriptions of well-known components and processing techniques are omitted so as to not unnecessarily obscure the embodiments herein. The examples used herein are intended merely to facilitate an understanding of ways in which the embodiments herein may be practiced and to further enable those of skill in the art to practice the embodiments herein. Accordingly, the examples should not be construed as limiting the scope of the embodiments herein.

[0039] In view of the foregoing, the need for generating prioritized recommendations for a user using a probabilistic causation model with at least one predicted anomaly is fulfilled in the ongoing description by (a) receiving a text-based input from the user comprising user symptoms and medical history, (b) performing automated adaptive query generation for presenting queries to the user at the user interface using a text-based conversation and recording, from the user interface, a set of responses the queries, (c) predicting at least one anomaly by processing the set of responses using a Bayesian probabilistic causation model, wherein training for the Bayesian probabilistic causation model is performed by (i) obtaining at least one of real-world data of patients generated by a hospital or, and publicly available clinical data that comprises symptoms, historical pathways of patients and outcomes, (ii) determining probabilities associated with the historical pathways and outcomes, and (iii) training the Bayesian probabilistic causation model by dynamically constructing a knowledge graph that captures relationships between the symptoms and anomalies as probabilities, (d) generating at least one prioritized recommendation based on the predicted anomaly using the Bayesian probabilistic causation model, and (e) transmitting, at the user interface, the at least one prioritized recommendation test to the user. Referring now to the drawings, and more particularly to FIGS. 1 through 14, where similar reference characters denote corresponding features consistently throughout the figures^ there are shown preferred embodiments.

[0040] FIG. 1 illustrates a system for generating automated adaptive queries to automatically determine a triage level according to an embodiment herein. The system 100 includes one or more users 102A-N that are associated with a plurality of user devices 104A-N that are communicatively connected to a distributed cloud 108 via a data communication network 106. The data communication network 106 may be one or more of a wired network, a wireless network, a combination of the wired network and the wireless network or the Internet. The user devices 104A-N include, but are not limited to, a mobile device, a smartphone, a smart watch, a notebook, a Global Positioning System (GPS) device, a tablet, a desktop computer, a laptop or any network enabled device. The distributed cloud 108 comprises a prioritized recommendation server 110 and a Bayesian probabilistic causation model 112.

[0041] The prioritized recommendation server 110 receives a text-based input from the one or more users 102A-N comprising user symptoms and medical history, the prioritized recommendation server 110 performs automated adaptive query generation for presenting queries to the one or more users 102A-N at the user interface of the one or more user devices 104A-N using a text-based conversation and recording, from the user interface at the one or more user devices 104A-N , a set of responses the queries.

[0042] The prioritized recommendation server 110 predicts at least one anomaly by processing the set of responses using the Bayesian probabilistic causation model 112. Training for the Bayesian probabilistic causation model 112 is performed by (i) obtaining at least one of real- world data of patients generated by a hospital or, and publicly available clinical data that comprises symptoms, historical pathways of patients and outcomes, (ii) determining probabilities associated with the historical pathways and outcomes, and (iii) training the Bayesian probabilistic causation model by dynamically constructing a knowledge graph that captures relationships between the symptoms and anomalies as probabilities.

[0043] The prioritized recommendation server 110 generates at least one prioritized recommendation based on the predicted anomaly using the Bayesian probabilistic causation model 112. The prioritized recommendation server transmits, at the user interface of the one or more user devices 104A-N, the at least one prioritized recommendation test to the one or more users 102A-N.

[0044] The system 100 is of advantage that the system 100 improves the accuracy of recommendations by employing an order of prioritization. By utilizing a probabilistic causation model, the system 100 assigns probabilities to different anomalies based on symptoms of the user and medical history. The system 100 takes into account the context of the user and medical history and generates recommendations that are personalized and take into account the specific context of the user.

[0045] Additionally, the system 100 offers computational resource optimization by utilizing an adaptive query generation approach and utilizing the probabilistic causation model. The system 100 captures relevant information from the user in a more efficient manner. By minimizing repetitive queries and gathering meaningful data with fewer interactions, computational resources are optimized. This not only improves the user experience by reducing the burden of multiple queries but also streamlines the recommendation generation process, allowing for faster and more resource-efficient outcomes. The system 100 increases accuracy of recommendations through prioritization, considering the context of the user and their medical history, and optimize computational resources.

[0046] In some embodiments, the prioritized recommendation includes a specialist or a category of test, including a blood test, an imaging test, or a genetic test. [0047] In some embodiments, the system 100 further includes generating a reasoning for the prioritized recommendation by dynamically generating a graph of a plurality of paths based on the anomaly, demographic details, medical history of the user, and historically recommended tests, and determining probabilities associated with each of the plurality of paths within the graph for generating a reasoning for the prioritized recommendation.

[0048] In some embodiments, the system 100 includes dynamically updating the recommendation of the diagnostic test upon obtaining new information associated with the user.

[0049] In some embodiments, the system 100 further includes a test prediction Al model, wherein the test prediction Al model is configured to analyze a plurality of diagnostic tests and an interpretation of the plurality of diagnostic tests from a health records database and generate a diagnostic test recommendation for the user based on the analysis and interpretation of the diagnostic tests.

[0050] FIG. 2 illustrates a flow diagram of generating a recommendation of a diagnostic test for a patient according to an embodiment herein. A lab test prediction Al 202 loads the context of the user from a user database that includes history, manifestation of a condition and demographic. An assistance system 200 provides Differential diagnosis (DDx) probabilities to the lab test prediction Al 202. A diagnostic analysis module 206 obtains test results from a public health records database 204. The lab test prediction Al 202 analyses one or more diagnostic tests and an interpretation of the one or more diagnostic tests based on the public health records database 204 and generates a diagnostic test recommendation for the user.

[0051] The following table “Table 1” illustrates experimental data illustrating responses from the one or more users 102A-N, anomalies predicted and recommendations generated by the prioritized recommendation server 110.

Table 1

[0052] With reference to the FIG. 2, FIG. 3 illustrates an exemplary process flow diagram of an assistance system for providing medical recommendations according to an embodiment herein. At step 302, model inputs are provided by the Artificial Intelligence (Al) model. The model inputs include demographic details, patient history, complaints/ symptoms. In some embodiments, the demographic details include age, gender, location and social-economic backgrounds. In some embodiments, the patient history includes chronic conditions, short term conditions, medications currently taken, allergies and surgeries. In some embodiments, the complaints/ symptoms include headache, fever, onset, severity, lab data and imaging reports. At step 304, the Artificial Intelligence (Al) model performs a learning process using UMLS, ICDs, open data, Personal Health Record (PHR), predicted symptoms, consultations/ user actions and the model inputs. At step 306, the assistance system 200 provides queries to the user to obtain more medical condition details using an interaction unit 312. The user responses to the queries are constructed as the model inputs. At step 308, a triage is classified for the user using the Artificial Intelligence (Al) model based on the medical conditions of the user and the medical recommendation are provided. At step 310, a report is generated with a reasoning for the medical recommendation.

[0053] FIG. 4 illustrates an exemplary system view of an assistance system 200 according to an embodiment herein. The system view includes a user 102, an assistance system 200, a doctors/ physicians/ clinician 406, a computational cloud 408 and a data base 410. In some embodiments, the assistance system 200 enables at least one of the user 402 or the doctors/ physicians/ clinician 406 using an application programming interface. The assistance system 200 enables the user 102 to provide at least one medical conditions of the user 102 using at least one medical queries. In some embodiments, the at least one medical queries are real-time queries which are from the doctors/ physicians/ clinician 406. In some embodiments, the assistance system 200 includes a conversation chat box and a symptom checker 412. In some embodiments, the conversational chat box provides at least one of clinical inquiry, past medical history, visual cues or multi-lingual support. In some embodiments, the symptom checker 412 provides at least one of an inquiry that follows the SOCRATES method, Differential diagnosis (DDx) probabilities, reports or application programming interfaces. The SOCRATES method is a form of cooperative argumentative dialogue between individuals, based on asking and answering questions to stimulate critical thinking and to draw out ideas and underlying presuppositions. The at least one medical queries are provided to the user 102 using the conversation chat box and obtains user responses. The user responses are processed by an Artificial Intelligence (Al) model for training. In some embodiments, the Artificial Intelligence (Al) model accesses at least one (a) electronic medical reports from the database 410 and the computational cloud 408 or (b) real time medical data from smart devices associated with the user 102. The assistance system 200 analyses the user responses, the electronic medical reports and the real time data to determine medical conditions. Medical recommendations that were provided earlier for the medical conditions that are similar to the determined medical conditions may be used by the Al models of the assistance system 200 as an input for performing analysis.

[0054] The assistance system 200 provides the determined medical conditions and the identified medical recommendations to the doctors/ physicians/ clinician 406 using the application programming interface. In some embodiments, the assistance system 200 includes a condition management system 414. The condition management system 414 provides the determined medical conditions to the doctors/ physicians/ clinician 406. In some embodiment, the assistance system 200 provides the symptom checker 412 for patients and a decision intelligence system (DIS) 416 for doctors. The decision intelligence system (DIS) 416 determines medical recommendations for the determined medical conditions and provides the medical recommendations to the doctors/ physicians/ clinician 406. The condition management system 414 provides recommendations to manage chronic conditions such as diabetes, cardiovascular diseases, COPD, mental health, cancer or gastroenterology. The decision intelligence system (DIS) provides at least one of risk predictions, triaging, clinical assessments, clinical algorithms, clinical search, pattern identification or contextual organization of clinical data. In some embodiments, the assistance system 200 provides triage, Differential Diagnosis (DDx), Prognosis, Next Best Steps for both users and doctors. In some embodiments, the assistance system 200 trains the Artificial Intelligence (Al) model using both real world clinical data from an organization database and clinical knowledge base created and reviewed by specialists of the organization. In some embodiments, the organization includes hospitals, clinics, and virtual consultation applications.

[0055] In some embodiments, the assistance system 200 enables at least one of the user 102 or the doctors/ physicians/ clinician 406 to install an application at a user device. The user device includes, but not limited to, a mobile, a tablet, a desktop computer, a laptop. The assistance system 200 communicates with the user 402 using the application programming interface.

[0056] With reference to FIG. 3, FIG. 5 A illustrates an exploded view of the assistance system 200 for providing the medical recommendations based on the user inputs using the Artificial Intelligence (Al) model according to an embodiment herein. The assistance system 200 includes a database 502, an interaction unit 312, a data processing unit 506, a reasoning unit 508, a triage determination unit 510 and a recommendation unit 512. The database 502 stores at least one of real time data from the organization, a clinical knowledge base, triage and clinical pathways in a digitized form or a digitized clinical content from medical journals. The data from the organization includes at least one of de-identified, real-world hospital generated patient data, publicly available clinical data such as MIMIC and Pubmed datasets or parsed data that is constructed as knowledge graphs, where relationships are captured as probabilities. The clinical knowledge base includes at least one of data that is curated by a team of in-house doctors, peer reviewed by organization specialists, case sheets cover clinical background, care and management, treatment options, discharge advice. In some embodiments, data is ingested into Artificial Intelligence (Al) models 516 periodically to update. The interaction unit 312 includes one or more application programming interfaces that enables at least one of the end users (example: patient and doctor) to provide inputs and access the medical recommendations. [0057] The data processing unit 506 obtains the inputs from the end user and transforms into an analyzed data. The analyzed data includes one or more medical condition and one or more medical recommendations provided for similar medical conditions.

[0058] The reasoning unit 508 determines the medical recommendations along with justification to the user responses using the Artificial Intelligence (Al) models 516. In some embodiments, the reasoning unit 508 acts as a brain and connects dots together for the assistance system 200. In some embodiments, the reasoning unit 508 is aware of patient demographics, medical history - details such as chronic conditions, surgeries, family medical history etc. The reasoning unit 508 analyses every detail such as symptoms, symptom attributes, vitals, lab/imaging results, comorbidity, hospitalization history, etc. using clinical knowledge base, millions of patient records and user inputs. The reasoning unit 508 dynamically creates a graph with possible outcomes and calculates probabilities associated with each path. In some embodiments, the graph is converted into actions or recommendations such as Differential diagnosis, Lab, imaging, medication prescription, Assessment, Treatment, Management, etc.

[0059] The triage determination unit 510 determines a triage level using the Artificial Intelligence (Al) models 516. The triage determination unit 510 includes triage protocols to determine the triage level. In some embodiments, the triage protocols were selected as per body system category. In some embodiments, the triage protocols include at least one of Cardiac, Respiratory, Neurology, Gastrointestinal or Mental health flagging signs. In some embodiments, Pediatrics and Pregnancy Related emergency protocols are also captured for body systems. The system covers 5000+ Triage protocols. The triage case sheet is a user- friendly triage dashboard that is built to create and edit the triage protocols.

The triage level is briefed, and the triage case sheets are reviewed by the doctors of the organization.

[0060] With reference to the FIG. 5A, FIG. 5B illustrates an exploded view of the data processing unit 506 of FIG. 5A according to an embodiment herein. The data processing unit 506 includes a Natural Language Processing (NLP) system 514, the Artificial Intelligence (Al) models 516 and a computation system 518. In some embodiments, the computation system 514 performs computations on billions of data points. The computation system 514 includes a spark based scalable cluster computational framework 520. The computation system parses medical notes and builds knowledge graph. The Artificial Intelligence (Al) models 516 include a symptom checker module 522 to predict diagnosis from symptoms. The symptom checker module 522 capture symptoms, diagnosis and relationships between them as knowledge. In some embodiments, the Artificial Intelligence (Al) models 516 include a Bayesian probabilistic causation model 524 for predicting disease probability from symptoms. In some embodiments, the Artificial Intelligence (Al) models 516 include a symptom inquiry model 526 that suggests unreported symptoms. In some embodiments, the Artificial Intelligence (Al) models 516 include a neural network model 528 that captures relationships between medical entities and recommendations provided by the medical entities. The Artificial Intelligence (Al) models 516 capture graphs Patient demography, history and its impact on knowledge graph. [0061] In some embodiments, the data processing unit 506 generates the Natural Language Processing (NLP) system 514 based on open-source technologies (example: Apache C-takes). In some embodiments, the Natural Language Processing (NLP) system 514 includes various stages of processing that includes (a) pre-processing, (b) Natural Language Processing (NLP) parsing, (c) unified medical language system (UMLS) lookup system for normalizing & standardizing clinical terms, (d) entity-entity relationship identification and (e) knowledge graph extraction and medical language modeling. The Natural Language Processing (NLP) system 514 classifies medical terms such as symptoms, diagnosis, medication, history etc. The Natural Language Processing (NLP) system 514 identifies assertions and relationships between entities such as onset, severity, duration, location etc. In some embodiments, the preprocessing includes spell correct, abbreviation expansion, substitution etc. In some embodiments, the Unified Medical Language System (UMLS) is a compendium of many controlled vocabularies in biomedical sciences. The unified medical language system provides a mapping structure among vocabularies that allows one to translate among a various terminology system for interoperability. The Unified Medical Language System (UMLS) provides facilities for Natural Language Processing (NLP) and that is intended to be used mainly by developers of medical informatics systems. In some embodiments, the Metathesaurus forms are the base of the Unified Medical Language System (UMLS). The Metathesaurus construction understands the intended meaning of each name in each source vocabulary and to link all the names that mean the same thing (the synonyms). In some embodiments, the examples of the incorporated controlled vocabularies include at least one of CPT, ICD-10, MeSH, SNOMED CT, LOINC, WHO Adverse Drug Reaction Terminology or RxNorm. A Concept Unique Identifier (CUI) is assigned to a meaning or a term which includes many different names.

[0062] The Natural Language Processing (NLP) system 514 performs at least one action of (a) normalizing clinical text, (b) extracting medical concepts out of free text, (c) identifying associations and relevance of medical concepts and (d) understanding context and user intent for Natural language understanding (NLU) systems. In some embodiments, the data processing unit 506 utilizes the database 502 as a knowledge base database to store at least one of digitized clinical contents from journals, digitized publicly available data sources, physician/expert knowledge from surveys, medical domain knowledge repository, triage and clinical pathways in a digitized form or parsed clinical information.

[0063] With reference to the FIG. 5A, FIG. 5C illustrates an exemplary view of a parsing architecture 530 of the data processing unit 506 of FIG. 5 A according to an embodiment herein. In some embodiments, the Natural Language Processing (NLP) parsing includes stemming, lemmatizing, sentence segmentation, word tokenization, stop words removal, part-of-speech recognition, dependency passing, nouns and named entity recognition etc. The Natural Language Processing (NLP) parsing includes named entity recognition, relation extraction, entity linking and negation detection. The named entity recognition detects words and phrases mentioned in unstructured text that is associated with one or more semantic types, such as diagnosis, medication name, symptom/sign, or age. The relation extraction identifies meaningful connections between concepts mentioned in text. For example, a "time of condition" relation is found by associating a condition name with a time. The entity linking disambiguates distinct entities by associating named entities mentioned in text to concepts found in a predefined database of concepts. For example, the Unified Medical Language System (UMLS). In some embodiments, the meaning of medical content is highly affected by modifiers such as negation, which includes critical implication if misdiagnosed. Text Analytics for health supports negation detection for the different entities mentioned in the text.

[0064] FIG. 6 illustrates an exemplary block diagram of a reasoning unit 508 of FIG. 2 according to an embodiment herein. The block diagram includes a computational cloud 408, a clinical knowledge base 604, a real-world data 606, a reinforcement learning model 608, a clinical Natural Language Processing (NLP)/ Natural Language Understanding (NLU) system 610 and conversational A [/(Natural Language Generation) NLG module 612. In some embodiments, the clinical NLP/NLU system 610 is communicatively connected with the computational cloud 408, the clinical knowledge 604, the real-world data 606, the reinforcement 608 and the reasoning unit 508. The computational cloud 408 provides previous case history to the clinical NLP/NLU system 610. The clinical knowledge 604 captures at least one of latest disease specific assessment, diagnosis and management details from specialists or journals. The clinical Natural Language Processing (NLP)/ Natural Language Understanding (NLU) 610 identifies clinical concepts, phrases, context, relationships. The reasoning unit 508 determines the medical recommendations along with justification to the user responses using the Artificial Intelligence (Al) model. In some embodiments, real-world data includes at least one of de-identified, real world hospital generated patient data, publicly available clinical data such as MIMIC and Pubmed datasets or parsed data that is constructed as knowledge graphs where relationships are captured as probabilities. The conversational AI/NLG module 612 enables inquiry with the user by providing an intelligent query to the user based on a response of the user to a previous query.

[0065] FIGS. 7A to 7E illustrate a user interface of an assistance system for assisting a patient at a patient device according to an embodiment herein. The assistance system 200 provides a symptom checker for patients to obtain the user responses. The user interface of the symptom checker includes a guide 702, a disclaimer 704, symptom assessment 706, a symptom assessment report 708, medical recommendations 710. In some embodiments, the guide 702 includes one or more instructions that include “tell us your symptoms, your answers will be carefully analyzed, review possible causes etc.”. In some embodiments, the symptom assessment 706 includes one or more multiple choice questions and an option to describe the symptoms. In some embodiments, the symptom assessment report 708 includes one or more medical conditions related to the symptoms. In some embodiments, the medical recommendations 710 includes suggested lab, imaging tests and/or specialty for doctor consultation.

[0066] FIGS. 8 A to 8G illustrate a user interface of providing medical assistance to a patient in a doctor consultation setting according to an embodiment herein. The virtual consultation room enables the user to provide the medical conditions of the user using one or more queries 802 and briefs the medical conditions 804 for the doctor. The one or more queries 802 are generated based on the user response to a query. The user is enabled to report more medical conditions on any other symptoms option 806. The one or more queries 802 includes multiple choices and the option 806 to describe the symptoms. The choices may include, for example, sore throat, nasal congestion, coryza, ear pain etc. In some embodiments, the medical conditions include a summary that is determined based on the user inputs. The summary may include at least one of age, height, weight, medicine allergies, food allergies, a type of medical condition or a duration of medical condition and associated details. In some embodiments, the summary is provided to the user.

[0067] FIGS. 9 A and 9B illustrate a user interface of accessing patient data by a doctor for assisting a patient according to an embodiment herein. The assistance system 200 enables the doctor to access the medical conditions 804 as a summary 902 and a reported symptom 904. Clinical recommendations 906 are provided to the doctor for providing medical recommendations. In some embodiments, the clinical recommendations 906 include lab, imaging tests (example: X-ray, RT-PCR test), possible conditions (example: flu, migraine) or medical history and previous medications (example: diabetes, hypertension).

[0068] FIGS. 10A to 10C illustrate a report generated for a patient by the assistance system 200 for providing medical recommendations. As an example, a male patient having age of 50 years, height of 5.5 feet and weight 70 kilograms is a patient being assisted by the assistance system. The following is summary for available to a doctor when the doctor uses the user interface for assisting the patient: “A 50 year old male with a history of smoking presents with moderate chest pain. He noticed sudden onset of chest pain. He describes his chest pain staying the same over period of time, located at left side, without radiation, aggravated by stress, associated with difficulty swallowing. There is no history of shortness of breath/ difficulty breathing.” Further, following are the reported symptoms”.

[0069] The report in FIGs 10B to 10C includes diagnostic test recommendation generated for the user.

[0070] FIGS. 11A-B are a flow diagram illustrating a computer-implemented method of generating automated adaptive queries to automatically determine a triage level according to an embodiment herein. At step 1102, the method includes receiving, from a user interface, unstructured text input from the user, wherein the unstructured text is provided by the user in response to a query and comprises current or past state of the user. At step 1104, the method includes disambiguating at least one of distinct entities and distinct concepts from the unstructured text using a natural language processing (NLP) model to obtain a context of the user. At step 1106, the method includes performing relation extraction between the distinct entities and distinct concepts using the NLP model to detect a manifestation of at least one anomaly. At step 1108, the method includes automatically generating an updated query based on the manifestation of the at least one anomaly using an adaptive query generation model that personalizes the updated query based on a medical history and the context of the user, wherein the updated query inquires the user for additional information associated with the manifestation of the at least one anomaly. At step 1110, the method includes presenting the updated query to the user using a text-based conversation at the user interface and recording, from the user interface, a set of responses of the updated query. At step 1112, the method includes determining at least one flagging sign based on the set of responses by comparing the set of responses with a predetermined threshold or a reference data. At step 1114, the method includes automatically generating a second updated query by using the adaptive query generation model with the flagging sign, wherein the second updated query inquires the user for additional information associated with the flagging sign. At step 1116, the method includes automatically determining a triage level of the user by using a triage determination model with a response to the second updated query.

[0071] FIG. 12 is a flow diagram illustrating a computer-implemented method for generating prioritized recommendations for a user using a probabilistic causation model with at least one predicted anomaly according to an embodiment herein. At step 1202, the method includes receiving a text-based input from the user comprising user symptoms and medical history. At step 1204, the method includes performing automated adaptive query generation for presenting queries to the user at the user interface using a text -based conversation and recording, from the user interface, a set of responses the queries. At step 1206, the method includes predicting at least one anomaly by processing the set of responses using a Bayesian probabilistic causation model, wherein training for the Bayesian probabilistic causation model is performed by (i) obtaining at least one of real-world data of patients generated by a hospital or, and publicly available clinical data that comprises symptoms, historical pathways of patients and outcomes, (ii) determining probabilities associated with the historical pathways and outcomes, and (iii) training the Bayesian probabilistic causation model by dynamically constructing a knowledge graph that captures relationships between the symptoms and anomalies as probabilities. At step 1208, the method includes generating at least one prioritized recommendation based on the predicted anomaly using the Bayesian probabilistic causation model. At step 1210, the method includes transmitting, at the user interface, the at least one prioritized recommendation test to the user.

[0072] The method is of advantage that the method improves the accuracy of recommendations by employing an order of prioritization. By utilizing a probabilistic causation model, the method assigns probabilities to different anomalies based on symptoms of the user and medical history. The method takes into account the context of the user and medical history and generates recommendations that are personalized and take into account the specific context of the user.

[0073] Additionally, the method offers computational resource optimization by utilizing an adaptive query generation approach and utilizing the probabilistic causation model. The method captures relevant information from the user in a more efficient manner. By minimizing repetitive queries and gathering meaningful data with fewer interactions, computational resources are optimized. This not only improves the user experience by reducing the burden of multiple queries but also streamlines the recommendation generation process, allowing for faster and more resource-efficient outcomes. The method increases accuracy of recommendations through prioritization, considering the context of the user and their medical history, and optimize computational resources.

[0074] FIG. 13 illustrates an exploded view of the assistance system 200 of FIG.4 having a memory 1302 having a set of instructions, a bus 1304, a display 1306, a speaker 1308, and a processor 1310 capable of processing the set of instructions to perform any one or more of the methodologies herein, according to an embodiment herein. The processor 1310 may also enable digital content to be consumed in the form of a video for output via one or more displays 1306 or audio for output via speaker and/or earphones 1308. The processor 1310 may also carry out the methods described herein and in accordance with the embodiments herein.

[0075] Digital content may also be stored in the memory 1302 for future processing or consumption. The memory 1302 may also store program-specific information and/or service information (PSI/SI), including information about digital content (e.g., the detected information bits) available in the future or stored from the past. A user of the receiver 1300 may view this stored information on display 1306 and select an item for viewing, listening, or other uses via input, which may take the form of a keypad, scroll, or another input device (s) or combinations thereof. When digital content is selected, the processor 1310 may pass information. The content and PS1/SI may be passed among functions within the receiver using the bus 1304.

[0076] The embodiments herein can take the form of, an entire hardware embodiment, an entire software embodiment or an embodiment including both hardware and software elements. The embodiments that are implemented in software include but are not limited to, firmware, resident software, microcode, etc. Furthermore, the embodiments herein can take the form of a computer program product accessible from a computer-usable or computer-readable medium providing program code for use by or in connection with a computer or any instruction execution system. For the purposes of this description, a computer-usable or computer readable medium can be any apparatus that can comprise, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.

[0077] The medium can be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system (or apparatus or device) or a propagation medium. Examples of a computer-readable medium include a semiconductor or solid-state memory, magnetic tape, a removable computer diskette, a random-access memory (RAM), a read-only memory (ROM), a rigid magnetic disk and an optical disk. Current examples of optical disks include compact disk - read-only memory (CD-ROM), compact disk - read/write (CD-R/W) and DVD.

[0078] A data processing system suitable for storing and/or executing program code will include at least one processor coupled directly or indirectly to memory elements through a system bus. The memory elements can include local memory employed during actual execution of the program code, bulk storage, and cache memories which provide temporary storage of at least some program code in order to reduce the number of times code must be retrieved from bulk storage during execution.

[0079] Input/output (RO) devices (including but not limited to keyboards, displays, pointing devices, remote controls, etc.) can be coupled to the system either directly or through intervening I/O controllers. Network adapters may also be coupled to the system to enable the data processing system to become coupled to other data processing systems or remote printers or storage devices through intervening private or public networks. Modems, cable modem, and Ethernet cards are just a few of the currently available types of network adapters.

[0080] A representative hardware environment for practicing the embodiments herein is depicted in FIG. 14. This schematic drawing illustrates a hardware configuration of an information handling/computer system in accordance with the embodiments herein. The system comprises at least one processor or central processing unit (CPU) 10. The CPUs 10 are interconnected via system bus 12 to various devices such as a random-access memory (RAM) 14, read-only memory (ROM) 16, and an input/output (I/O) adapter 18. The RO adapter 18 can connect to peripheral devices, such as disk units 11 and tape drives 13, or other program storage devices that are readable by the system. The system can read the inventive instructions on the program storage devices and follow these instructions to execute the methodology of the embodiments herein.

[0081] The system further includes a user interface adapter 19 that connects a keyboard 15, mouse 17, speaker 24, microphone 22, and/or other user interface devices such as a touch screen device (not shown) or remote control to the bus 12 to gather user input. Additionally, a communication adapter 20 connects the bus 12 to a data processing network 25, and a display adapter 21 connects the bus 12 to a display device 23 which may be embodied as an output device such as a monitor, printer, or transmitter, for example.

[0082] The foregoing description of the specific embodiments will so fully reveal the general nature of the embodiments herein that others can, by applying current knowledge, readily modify and/or adapt for various applications such specific embodiments without departing from the generic concept, and, therefore, such adaptations and modifications should and are intended to be comprehended within the meaning and range of equivalents of the disclosed embodiments. It is to be understood that the phraseology or terminology employed herein is for the purpose of description and not of limitation. Therefore, while the embodiments herein have been described in terms of preferred embodiments, those skilled in the art will recognize that the embodiments herein can be practiced with modification within the spirit and scope.

Claims

CLAIMS I/We Claim:

1. A method for generating prioritized recommendations for a user using a probabilistic causation model with at least one predicted anomaly, comprising: receiving, at the automated adaptive query server (110), a text-based input from at least one user devices (104A-N) comprising user symptoms and medical history; performing, at the automated adaptive query server (110), automated adaptive query generation for presenting queries to the user at the user interface of the one or more user devices (104A-N) using a text-based conversation and recording, from the user interface of the one or more user devices (104A-N), a set of responses the queries; predicting, at the automated adaptive query server (110), at least one anomaly by processing the set of responses using a Bayesian probabilistic causation model (112), wherein training for the Bayesian probabilistic causation model is performed by: obtaining at least one of (a) real-world data of patients generated by a hospital or, and (b) publicly available clinical data that comprises symptoms, historical pathways of patients and outcomes; determining probabilities associated with the historical pathways and outcomes; and training the Bayesian probabilistic causation model (112) by dynamically constructing a knowledge graph that captures relationships between the symptoms and anomalies as probabilities; generating, at the automated adaptive query server (110), at least one prioritized recommendation based on the predicted anomaly using the Bayesian probabilistic causation model (112); transmitting, at the user interface of the at least one user devices (104A-N), the at least one prioritized recommendation test to the user.

2. The method as claimed in claim 1 , wherein the prioritized recommendation includes a specialist or a category of test, including a blood test, an imaging test, or a genetic test.

3. The method as claimed in claim 1, further comprising generating a reasoning for the prioritized recommendation by: dynamically generating a graph of a plurality of paths based on the anomaly, demographic details, medical history of the user, and historically recommended tests; and determining probabilities associated with each of the plurality of paths within the graph for generating a reasoning for the prioritized recommendation.

4. The method as claimed in claim 1, further comprising dynamically updating the recommendation of the diagnostic test upon obtaining new information associated with the user.

5. The method as claimed in claim 1, further comprising a test prediction Al model, wherein the test prediction Al model is configured to: analyze a plurality of diagnostic tests and an interpretation of the plurality of diagnostic tests from a health records database; and generate a diagnostic test recommendation for the user based on the analysis and interpretation of the diagnostic tests.

6. A system for generating prioritized recommendations for a user using a probabilistic causation model with at least one predicted anomaly, wherein the system comprises of: a distributed cloud comprising a prioritized recommendation server (110) that comprises a processor and a memory that are configured to perform: receiving a text-based input from the user comprising user symptoms and medical history; performing automated adaptive query generation for presenting queries to the user at the user interface of the at least one user devices (104A-N) using a text-based conversation and recording, from the user interface of the at least one user devices (104A-N), a set of responses the queries; predicting at least one anomaly by processing the set of responses using a Bayesian probabilistic causation model (112), wherein training for the Bayesian probabilistic causation model (112) is performed by: obtaining at least one of (a) real-world data of patients generated by a hospital or, and (b) publicly available clinical data that comprises symptoms, historical pathways of patients and outcomes; determining probabilities associated with the historical pathways and outcomes; and training the Bayesian probabilistic causation model (112) by dynamically constructing a knowledge graph that captures relationships between the symptoms and anomalies as probabilities; generating at least one prioritized recommendation based on the predicted anomaly using the Bayesian probabilistic causation model (112); transmitting, at the user interface of the at least one user devices (104A-N), the at least one prioritized recommendation test to the user.

7. The system as claimed in claim 6, wherein the prioritized recommendation includes a specialist or a category of test, including a blood test, an imaging test, or a genetic test.

8. The system as claimed in claim 6, further comprising generating a reasoning for the prioritized recommendation by: dynamically generating a graph of a plurality of paths based on the anomaly, demographic details, medical history of the user, and historically recommended tests; and determining probabilities associated with each of the plurality of paths within the graph for generating a reasoning for the prioritized recommendation.

9. The system as claimed in claim 6, further comprising dynamically updating the recommendation of the diagnostic test upon obtaining new information associated with the user.

10. The system as claimed in claim 6, further comprising a test prediction Al model, wherein the test prediction Al model is configured to: analyze a plurality of diagnostic tests and an interpretation of the plurality of diagnostic tests from a health records database; and generate a diagnostic test recommendation for the user based on the analysis and interpretation of the diagnostic tests.