Abstract
Background: Artificial intelligence is rapidly influencing medical education, clinical decision-making, research, pharmaceutical development, and public health. Ayurveda generates complex multidimensional information through Prakriti assessment, Dosha-Dushya analysis, Rogi-Roga Pariksha, dietary evaluation, and individualized treatment planning. These features make Ayurveda potentially suitable for carefully designed artificial intelligence applications. Recent competency-based curricula of the National Commission for Indian System of Medicine explicitly introduce artificial intelligence, digital health, diagnostic software, research databases, and development of technology-assisted diagnostic tools. Objective: To examine the potential applications, limitations, ethical concerns, and educational requirements of artificial intelligence in Ayurveda and propose an NCISM-aligned framework for responsible implementation. Methods: A narrative review was conducted using classical Ayurvedic literature, NCISM curriculum documents, official reports of the Ministry of Ayush and World Health Organization, and contemporary publications concerning artificial intelligence in traditional medicine, Ayurgenomics, machine learning, clinical prediction, and research governance. Results: Artificial intelligence may support Ayurveda education through adaptive learning, multilingual knowledge retrieval, simulation, assessment, and research training. Potential clinical applications include structured documentation, Prakriti classification, diagnostic support, risk stratification, image analysis, treatment monitoring, and remote follow-up. Machine-learning studies have demonstrated the feasibility of classifying Prakriti phenotypes, while recent research has explored artificial intelligence-assisted evaluation of Panchakarma procedures. Artificial intelligence may also assist literature synthesis, formulation research, pharmacovigilance, medicinal-plant identification, and analysis of complex clinical datasets. Major limitations include inadequate standardized datasets, inconsistent terminology, algorithmic bias, hallucinated classical references, poor external validation, privacy risks, opacity, and uncertain accountability. Conclusion: Artificial intelligence should function as a supervised clinical, educational, and research support system rather than an autonomous substitute for the Ayurvedic physician. Responsible adoption requires standardized terminology, validated datasets, interdisciplinary collaboration, human oversight, data governance, transparent reporting, and curriculum-based digital literacy. An NCISM-aligned roadmap can enable Ayurveda institutions to use artificial intelligence while preserving classical reasoning, patient safety, and professional accountability.
Keywords: Artificial Intelligence, Ayurveda, NCISM Curriculum, Digital Health, Clinical Decision Support, Ayurgenomics, Medical Education, Research Methodology
Introduction
Artificial intelligence (AI) refers to computational systems capable of performing tasks that ordinarily require human cognitive functions, including pattern recognition, language processing, prediction, classification, and decision support. Machine learning, deep learning, natural language processing, computer vision, and generative AI are important components of the contemporary AI ecosystem. Their applications in health care now extend from medical imaging and clinical documentation to drug discovery, remote monitoring, professional education, and evidence synthesis.⁷–¹¹
Ayurveda is characterized by individualized assessment, multidimensional clinical reasoning, and extensive textual knowledge. The physician evaluates not only the disease but also Prakriti, Vikriti, Dosha, Dushya, Agni, Bala, Satmya, Ahara, Vihara, Desha, Kala, age, and psychological status. Classical clinical reasoning is based on Pramana, Yukti, observation, inference, authoritative knowledge, and individualized application.¹ This complexity makes Ayurveda unsuitable for simplistic automation but highly relevant for carefully designed computational support.
The integration of AI into Ayurveda has moved beyond theoretical discussion. The competency-based curricula of the National Commission for Indian System of Medicine introduce artificial intelligence in research methodology, diagnostic education, Roganidana, digital health, and the development and validation of diagnostic tools.²–⁵ The Ministry of Ayush has similarly emphasized digital health, evidence-based Ayurveda, innovation, and predictive, preventive, personalized, participatory, and precision-oriented health care.⁶ Internationally, the World Health Organization and International Telecommunication Union have mapped emerging applications of AI in traditional medicine while emphasizing governance, safety, cultural protection, and data sovereignty.⁷–⁹
Artificial intelligence therefore represents both an opportunity and a responsibility for Ayurveda. Its value will depend on whether it improves education, documentation, research quality, and patient care without replacing classical reasoning or weakening professional accountability.
Aim
To critically review the applications, limitations, ethical risks, and educational implications of artificial intelligence in Ayurveda and to propose an NCISM-aligned roadmap for responsible implementation.
Objectives
- To examine the relevance of artificial intelligence to Ayurveda education, diagnosis, research, and pharmaceutical development.
- To assess emerging evidence concerning machine learning and digital tools in traditional medicine.
- To identify risks related to data quality, bias, privacy, hallucination, and clinical accountability.
- To propose a practical framework for Ayurveda institutions, researchers, teachers, and clinicians.
Materials and Methods
A narrative review methodology was adopted. Classical concepts relevant to reasoning, examination, individualization, and therapeutic planning were examined from the Charaka Samhita. Official NCISM curriculum documents were reviewed to identify competency requirements related to AI, digital health, research portals, diagnostics, and technology-assisted learning. Reports and guidance documents from the Ministry of Ayush, World Health Organization, and International Telecommunication Union were also reviewed.
Contemporary peer-reviewed literature was examined for applications of AI and machine learning in traditional medicine, Prakriti classification, Ayurgenomics, Panchakarma monitoring, clinical prediction, research reporting, and AI governance. The literature was synthesized under the domains of education, clinical practice, research, pharmaceutical development, public health, ethics, and institutional implementation.
Classical Reasoning and the Role of Artificial Intelligence
Ayurvedic decision-making is not based on a single symptom, laboratory parameter, or diagnostic label. It is an integrated process involving evaluation of the patient, disease, causative factors, stage, strength, digestive capacity, constitution, season, locality, and therapeutic suitability. The concepts of Aptopadesha, Pratyaksha, Anumana, and Yukti provide a structured epistemological foundation for acquiring and applying knowledge.¹
Artificial intelligence can assist in organizing observations, identifying patterns, comparing previous records, and generating probabilities. It cannot independently reproduce the full clinical wisdom represented by Yukti. Yukti includes contextual interpretation, prioritization, timing, therapeutic sequencing, and adaptation to changing patient conditions. These functions require clinical experience, ethical judgment, and responsibility.
The appropriate model is therefore physician-supervised augmented intelligence. The AI system may retrieve, classify, calculate, summarize, or suggest, while the qualified physician verifies the information and retains authority over diagnosis and treatment.
NCISM Curriculum and the Emergence of AI Competency
The NCISM curriculum for Research Methodology and Medical Statistics requires students to demonstrate the use of research portals, databases, and artificial intelligence in Ayurveda. It includes exposure to literature-search databases, clinical-trial registries, AI-assisted diagnostics, and digital tools relevant to research.²
The undergraduate curriculum for Roga Nidana evam Vikriti Vigyan introduces digital health and AI in the context of diagnosis and prognosis.³ At the postgraduate level, Applied Basics of Roganidana-Vikritivijnana includes competencies related to AI, diagnostic software, NAMASTE coding, Prakriti and Agni assessment tools, ICD-11, quality assurance, and development and validation of diagnostic instruments.⁴
The Ayurpraveshika transitional curriculum also introduces the role of AI in academics, research, learning, and data mining from classical literature.⁵ This curricular shift indicates that AI literacy is becoming a professional competency rather than an optional technological interest.
Training in artificial intelligence should include more than demonstrations of chatbots. Students must learn evidence verification, data quality, algorithmic limitations, research ethics, privacy, clinical accountability, and appropriate disclosure of AI assistance.
Applications of AI in Ayurveda Education
Adaptive and Personalized Learning
AI-enabled learning systems can adjust educational content according to the learner’s performance, pace, and identified knowledge gaps. A student struggling with Dosha-Dushya Samurchhana or pharmaceutical calculations may receive additional examples, formative questions, and remedial content. Advanced learners may be directed toward clinical scenarios and critical appraisal.
Such systems may be particularly useful in competency-based education, where students must demonstrate progressive achievement rather than merely attend lectures.
Classical-Text Retrieval
Natural language processing may assist in searching large collections of Samhita literature, commentaries, Nighantus, pharmacopoeias, and research publications. A structured retrieval system could identify references related to a disease, formulation, Dravya, Prakriti, or therapeutic principle.
However, generative systems may fabricate verses, chapter numbers, commentaries, or interpretations. Classical references generated by AI must therefore be verified against an authenticated printed edition before academic or clinical use.
Multilingual Education
Ayurveda is taught through Sanskrit, Hindi, English, and regional languages. AI-assisted translation, terminology mapping, pronunciation support, and multilingual summarization could improve access. Translation systems must preserve technical meaning and should not replace expert review, particularly for terms that have no exact biomedical equivalent.
Simulation and Clinical Reasoning
Virtual patients can present evolving symptoms, examination findings, and investigation reports. Students may be asked to perform Nidana Panchaka analysis, assess Rogi Bala and Roga Bala, select investigations, and formulate an integrated management plan. AI can provide structured feedback while faculty members supervise clinical reasoning.
Academic Writing and Research Training
AI can assist students in identifying search terms, organizing literature, generating data-analysis code, preparing tables, and improving language. It should not be listed as an author and should not replace independent scientific reasoning. Any substantial AI assistance should be transparently disclosed according to journal policy.
Applications in Diagnosis and Clinical Decision Support
Structured Clinical Documentation
Ayurvedic case records frequently contain extensive narrative information. AI-supported documentation can convert consultation notes into structured fields such as Prakriti, Vikriti, Agni, Koshtha, Satmya, Bala, Dosha, Dushya, Srotas, Nidana, Lakshana, investigations, and treatment response.
Structured data can improve continuity of care, research quality, audit, and multicentric collaboration. The physician must verify every extracted field before it becomes part of the medical record.
Prakriti Assessment
Prakriti assessment involves multiple physical, physiological, and psychological characteristics. Machine-learning research has demonstrated that phenotypic features can form computationally identifiable clusters corresponding to major Prakriti categories.¹⁴ Earlier genomic studies also identified biological differences associated with constitutional phenotypes.¹⁵,¹⁶
These findings support the feasibility of digital Prakriti-assessment tools, but they do not justify replacing physician assessment. Tools trained only on extreme Vata, Pitta, and Kapha types may perform poorly in mixed constitutions or diverse populations. Age, sex, ethnicity, geography, culture, and current Vikriti may influence observable characteristics.
Rogi-Roga Pariksha and Risk Stratification
AI may help combine clinical history, examination findings, laboratory values, imaging, lifestyle, and Ayurveda-specific variables. Potential uses include predicting disease progression, identifying red flags, estimating treatment response, and prioritizing referrals.
Such tools must be clinically validated. A model that performs well in one college hospital may fail in another population. External validation and prospective evaluation are essential before clinical deployment.
Image and Signal Analysis
Computer vision may support analysis of tongue images, skin lesions, wounds, medicinal plants, radiological images, and procedural recordings. Comparable traditional-medicine research has used AI for tongue-image classification, although variations in lighting, camera quality, colour calibration, and annotation remain major concerns.¹⁷
Pulse and voice analysis have also been proposed, but reliable clinical deployment requires standardized acquisition devices, expert-labelled datasets, and transparent validation.
Panchakarma Monitoring
A recent research protocol applied AI and machine learning to the objective evaluation of Vamana Karma. The proposed system uses video and image analysis to monitor vomiting events, content characteristics, patient gestures, and procedural outcomes.¹³ This represents an important transition from general discussion to procedure-specific AI research in Ayurveda.
Such technology could improve documentation and interobserver reliability. It must remain supportive because Panchakarma safety depends on continuous clinical assessment, patient Bala, complications, and physician judgment.
Clinical Decision Support
AI may suggest differential diagnoses, relevant investigations, classical references, possible formulations, contraindications, and follow-up parameters. These suggestions should be treated as prompts rather than prescriptions.
An AI system must never autonomously prescribe Panchakarma, herbo-mineral medicines, or emergency treatment. Clinical responsibility remains with the registered practitioner.
AI in Ayurveda Research
Literature Search and Evidence Synthesis
Artificial intelligence can accelerate literature screening, duplicate removal, keyword generation, citation organization, and preliminary evidence mapping. This may help researchers manage literature distributed across biomedical databases, Ayurveda journals, institutional repositories, and classical texts.
Automated summaries can omit methodological limitations or misrepresent conclusions. Final evidence appraisal must therefore be performed by trained researchers.
Data Cleaning and Statistical Analysis
Clinical studies frequently encounter missing data, inconsistent terminology, and heterogeneous outcome measures. AI-assisted tools can identify anomalies, standardize fields, and support statistical modelling. Researchers must preserve the original dataset, document preprocessing, and distinguish exploratory modelling from confirmatory analysis.
Prediction Models
Machine learning may identify patterns related to prognosis, treatment response, adverse effects, or disease susceptibility. Prediction-model studies should follow established reporting standards such as TRIPOD+AI.¹⁸ AI-based clinical trials should follow CONSORT-AI and SPIRIT-AI guidance.¹⁹,²⁰
Reporting model accuracy alone is insufficient. Researchers should report sensitivity, specificity, calibration, discrimination, missing-data handling, class imbalance, external validation, error analysis, and clinical utility.
Knowledge Graphs and Classical Literature
Ayurvedic knowledge can be represented through relationships among Dravya, Rasa, Guna, Virya, Vipaka, Prabhava, Dosha, Dhatu, Srotas, disease, formulation, dose, Anupana, and contraindication. Knowledge graphs may support research hypothesis generation and systematic retrieval.
The source of every relationship should remain traceable to an authenticated text, pharmacopoeia, monograph, or research publication. Untraceable AI-generated connections should not be treated as established knowledge.
AI in Ayurvedic Pharmaceutical Research
Artificial intelligence may support medicinal-plant identification, detection of adulterants, prediction of phytochemical activity, network pharmacology, formulation optimization, stability studies, and pharmacovigilance. The Ayurinformatics Laboratory model demonstrates growing institutional interest in integrating Ayurveda with bioinformatics, computational biology, and AI.¹²
Potential applications include:
- classification of raw drugs through images;
- prediction of herb-compound-target relationships;
- analysis of chromatographic fingerprints;
- detection of manufacturing deviations;
- identification of adverse-drug-event patterns;
- prioritization of formulations for experimental research;
- prediction of herb-drug interactions.
Computational predictions cannot replace botanical authentication, pharmaceutical standardization, toxicological evaluation, or clinical trials. They should be used to prioritize and refine laboratory research.
AI in Public Health and Digital Ayurveda
AI-enabled teleconsultation systems may assist preliminary screening, follow-up reminders, lifestyle monitoring, and multilingual health education. They may improve access in areas with limited specialist availability.
Remote systems must clearly distinguish education, screening, and medical consultation. High-risk symptoms, pregnancy, pediatric emergencies, suspected malignancy, acute neurological deficits, severe infection, and potential toxicity require direct clinical assessment.
Population-level datasets may support epidemiological surveillance and analysis of dietary, lifestyle, environmental, and constitutional patterns. Such systems require equitable representation so that rural, tribal, linguistically diverse, and economically disadvantaged populations are not excluded.
Major Risks and Limitations
Inadequate Data Standardization
Ayurvedic terminology varies across institutions, languages, teachers, and documentation systems. The same term may be interpreted differently, while different terms may describe related concepts. AI trained on inconsistent data will reproduce inconsistency.
A national minimum dataset should define mandatory fields, terminology, units, coding rules, and outcome measures without eliminating clinically meaningful narrative information.
Algorithmic Bias
If training data predominantly represent one institution, region, age group, sex, or constitutional category, the model may not generalize. Bias can also arise from expert disagreement during data labelling.
Datasets should be multicentric, diverse, and independently reviewed. Performance should be reported separately across relevant demographic and clinical groups.
Hallucinated Information
Generative AI may produce plausible but false references, Sanskrit verses, ingredient lists, doses, or clinical recommendations. This is particularly dangerous in Ayurveda because authenticity often depends on exact textual formulation and processing.
No classical quotation, formulation, dose, or therapeutic claim generated by AI should be accepted without verification against the original printed source.
Loss of Clinical Context
AI may identify statistical patterns but fail to understand why a treatment is inappropriate for a particular patient. Matra, Kala, Anupana, Agni, Bala, Satmya, Koshtha, age, pregnancy, comorbidity, and ongoing medication can materially alter treatment decisions.
Privacy and Data Governance
Ayurveda datasets may include medical history, photographs, voice recordings, pulse recordings, genomic data, reproductive information, and lifestyle details. These require informed consent, secure storage, controlled access, anonymization, and defined retention policies.
Explainability
Clinicians must understand the basis of high-risk recommendations. A system that cannot explain which variables influenced its output should not independently guide therapy.
Accountability
Responsibility may be unclear when an AI recommendation causes harm. Institutions must define the roles of developers, hospitals, faculty, researchers, and clinicians. Final clinical accountability must remain with the treating practitioner.
Traditional Knowledge and Intellectual Property
Digitization can expose community knowledge, classical formulations, and local medicinal practices to unauthorized commercial extraction. AI governance must include attribution, cultural protection, benefit sharing, and respect for traditional-knowledge rights.⁷,¹¹
An NCISM-Aligned Roadmap for Responsible Implementation
1. Establish Foundational AI Literacy
Students should understand basic AI terminology, capabilities, limitations, bias, privacy, research integrity, and clinical accountability. Training should be adapted to the level of undergraduate students, postgraduate scholars, faculty, and clinicians.
2. Preserve Classical Foundations
AI instruction should be integrated with Pramana, Yukti, Rogi-Roga Pariksha, Nidana Panchaka, Prakriti, Agni, and treatment planning. Technology should reinforce rather than bypass classical reasoning.
3. Create Standardized Datasets
Institutions should adopt uniform case-record fields, terminologies, coding systems, data dictionaries, and outcome measures. Data quality should be audited before model development.
4. Use Human-in-the-Loop Systems
All clinical AI tools should require review and approval by a qualified professional. Systems should display uncertainty and allow clinicians to reject or modify outputs.
5. Validate Before Deployment
AI models should undergo internal validation, external validation, prospective testing, usability assessment, and safety evaluation. Performance in real clinical settings should be monitored after deployment.
6. Establish Institutional Governance
Every institution using clinical or research AI should have:
- an approved AI-use policy;
- ethics committee oversight;
- privacy and cybersecurity procedures;
- defined accountability;
- audit trails;
- incident-reporting mechanisms;
- periodic performance review.
7. Promote Interdisciplinary Collaboration
Ayurveda experts, data scientists, statisticians, software engineers, ethicists, pharmacologists, and patient representatives should collaborate from project conception. Technology developed without adequate Ayurveda expertise is likely to misclassify or oversimplify core concepts.
8. Follow Reporting Standards
Prediction models should follow TRIPOD+AI. AI clinical trials should follow CONSORT-AI and SPIRIT-AI. Authors must disclose the use of generative AI in manuscript preparation according to journal policy.
9. Begin with Low-Risk Applications
Institutions should initially prioritize:
- literature search;
- administrative documentation;
- student assessment;
- medicinal-plant image libraries;
- research-data cleaning;
- appointment and follow-up reminders.
High-risk diagnostic and treatment applications should be introduced only after validation and governance systems are mature.
Discussion
Artificial intelligence has substantial potential to improve Ayurveda education, documentation, research efficiency, and diagnostic standardization. Its ability to process multidimensional information aligns with Ayurveda’s individualized and systems-oriented approach. Machine-learning research on Prakriti and emerging AI-assisted Panchakarma assessment illustrate that computational methods can address genuine problems of classification, objectivity, and interobserver variation.¹³–¹⁶
Nevertheless, the same technology may generate inaccurate classical references, unsafe recommendations, biased classifications, and misleading certainty. The central challenge is not whether AI can generate an answer, but whether the answer is valid, contextual, explainable, and clinically safe.
The future of AI in Ayurveda should therefore be based on augmentation rather than replacement. AI can organize information, reveal patterns, and reduce repetitive workload. The physician must interpret the output through classical knowledge, clinical examination, patient preferences, and ethical responsibility.
NCISM’s inclusion of AI and digital health creates an opportunity to develop a nationally coherent educational framework. Without structured training, students may use public generative systems without understanding privacy, hallucination, bias, or evidence quality. Formal curriculum implementation can convert uncontrolled use into responsible professional competence.
Limitations
This is a narrative review and does not provide a quantitative meta-analysis. Ayurveda-specific AI research remains limited, and many proposed applications have not undergone prospective clinical validation. Rapid technological development may also make specific platforms or capabilities obsolete. The conclusions should therefore be interpreted as a framework for responsible development rather than evidence for routine autonomous clinical use.
Conclusion
Artificial intelligence represents an important emerging development in Ayurveda education, diagnosis, research, pharmaceutical science, and public health. NCISM curricula and recent international initiatives indicate that digital and AI competencies will become increasingly relevant to Ayurveda professionals.
Artificial intelligence may improve access to knowledge, standardize documentation, support Prakriti assessment, assist diagnostic research, monitor therapeutic procedures, and accelerate pharmaceutical investigation. These benefits will be meaningful only when supported by reliable data, external validation, ethical governance, privacy safeguards, and qualified human oversight.
The Ayurvedic physician must remain the final clinical decision-maker. Artificial intelligence should serve as a transparent and accountable assistant that strengthens Yukti and evidence-informed practice rather than replacing them. An NCISM-aligned approach based on classical integrity, digital literacy, interdisciplinary collaboration, and patient safety can help Ayurveda benefit from AI without compromising its foundational principles.
References
1. Agnivesha. Charaka Samhita. Revised by Charaka and Dridhabala. Ayurveda Dipika commentary of Chakrapanidatta. Acharya YT, editor. Varanasi: Chaukhambha Surbharati Prakashan; 2017. Sutra Sthana 11 and Vimana Sthana 4, 8. 2. National Commission for Indian System of Medicine. Competency-Based Dynamic Curriculum for Third Professional BAMS: Research Methodology and Medical Statistics (AyUG-RM). New Delhi: NCISM; 2025. 3. National Commission for Indian System of Medicine. Competency-Based Dynamic Curriculum for Second Professional BAMS: Roga Nidana evam Vikriti Vigyan (AyUG-RN). New Delhi: NCISM; 2025. 4. National Commission for Indian System of Medicine. Applied Basics of Roganidana-Vikritivijnana (AYPG-AB-RN), Semester II. New Delhi: NCISM; 2025. 5. National Commission for Indian System of Medicine. Ayurpraveshika: Transitional Curriculum for BAMS. New Delhi: NCISM; 2025. 6. Ministry of Ayush, Government of India. 11 Years AYUSH Report 2025. New Delhi: Ministry of Ayush; 2025. 7. World Health Organization, International Telecommunication Union. Mapping the Application of Artificial Intelligence in Traditional Medicine: Technical Brief. Geneva: World Health Organization; 2025. ISBN 978-92-4-010766-3. 8. World Health Organization. Ethics and Governance of Artificial Intelligence for Health: Guidance on Large Multi-Modal Models. Geneva: World Health Organization; 2025. ISBN 978-92-4-008475-9. 9. World Health Organization. Ethics and Governance of Artificial Intelligence for Health: WHO Guidance. Geneva: World Health Organization; 2021. ISBN 978-92-4-002920-0. 10. Jongjiamdee K, Pornwonglert P, Na Bangchang N, Akarasereenont P. Artificial intelligence in traditional medicine: evidence, barriers, and a research roadmap for personalized care. Front Artif Intell. 2025;8:1659338. doi:10.3389/frai.2025.1659338. 11. Pujari S, Singh R, Soon GC, Nesari T, Ghelman R, Zhao Y, et al. Artificial intelligence in traditional medicine: policy and governance strategies. Bull World Health Organ. 2025;103(11):738-740. doi:10.2471/BLT.24.292888. 12. Gupta PK, Nesari TM. Ayurinformatics Laboratory: a synergy platform for Ayurveda and technology. J Ayurveda Integr Med. 2024;15(5):101019. doi:10.1016/j.jaim.2024.101019. 13. Rani P, Kalra S, Singh S, David R, Gupta AR, Anandaraman PV. Using AI algorithms and machine learning in the analysis of a bio-purification method (therapeutic emesis, known as Vamana Karma): protocol for a mixed methods study. JMIR Res Protoc. 2026;15:e79875. doi:10.2196/79875. 14. Tiwari P, Kutum R, Sethi T, Shrivastava A, Girase B, Aggarwal S, et al. Recapitulation of Ayurveda constitution types by machine learning of phenotypic traits. PLoS One. 2017;12(10):e0185380. doi:10.1371/journal.pone.0185380. 15. Prasher B, Negi S, Aggarwal S, Mandal AK, Sethi TP, Deshmukh SR, et al. Whole genome expression and biochemical correlates of extreme constitutional types defined in Ayurveda. J Transl Med. 2008;6:48. doi:10.1186/1479-5876-6-48. 16. Govindaraj P, Nizamuddin S, Sharath A, Jyothi V, Rotti H, Raval R, et al. Genome-wide analysis correlates Ayurveda Prakriti. Sci Rep. 2015;5:15786. doi:10.1038/srep15786. 17. Liu Q, Li Y, Yang P, Liu X, Jin Y, Liu Y. A survey of artificial intelligence in tongue image for disease diagnosis and syndrome differentiation. Digit Health. 2023;9:20552076231191044. doi:10.1177/20552076231191044. 18. Collins GS, Moons KGM, Dhiman P, Riley RD, Beam AL, Van Calster B, et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ. 2024;385:e078378. doi:10.1136/bmj-2023-078378. 19. Liu X, Cruz Rivera S, Moher D, Calvert MJ, Denniston AK; SPIRIT-AI and CONSORT-AI Working Group. Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension. Nat Med. 2020;26(9):1364-1374. doi:10.1038/s41591-020-1034-x. 20. Cruz Rivera S, Liu X, Chan AW, Denniston AK, Calvert MJ; SPIRIT-AI and CONSORT-AI Working Group. Guidelines for clinical trial protocols for interventions involving artificial intelligence: the SPIRIT-AI extension. Nat Med. 2020;26(9):1351-1363. doi:10.1038/s41591-020-1037-7.