Quality Assurance Ecosystems for AI-Driven Ingestible Capsules: A Systematic Review in Gastrointestinal Diagnostics

Abstract

Based on this, the review proposes a holistic quality assurance ecosystem necessary for the safe and trustworthy deployment of artificial intelligence capsule systems from manufacturing to real-world use. AI-enabled ingestible capsules transform non-invasive diagnostics of the GI tract by realizing 90-99?curacy in lesion detection; yet, at the same time, they present unprecedented challenges related to quality assurance, involving hardware reliability, algorithmic non-determinism, data security, and clinical integration. Synthesis is supported by recent literature on hardware in capsule endoscopy, AI validation protocols, regulatory frameworks such as FDA SaMD, ISO 13485, and IMDRF GMLP, risk management, and emerging multi-sensor technologies. The lack of unified standards for edge AI, continuous learning, and therapeutic autonomy keeps fragmented today's QA approaches. Several preclinical phantoms, multi-center trials, and post-market registries have demonstrated feasibility but underlined the lack of harmonization across jurisdictions. It advocates for the implementation of a unified QA ecosystem: from hardware controls down to the trustworthiness of AI, ethical governance, and adaptive regulation, a critically required multi-stakeholder collaboration lays the foundation for establishing global benchmarks for explainable multimodal models and human-AI symbiosis paradigms that realize full potential for autonomous GI diagnostics.

Keywords

AI capsule endoscopy, quality assurance ecosystem, ingestible medical devices, Software as a Medical Device (SaMD), regulatory harmonization, edge AI, risk management, explainable AI, post-market surveillance, gastrointestinal diagnostics

Introduction

OVERVIEW OF AI ENABLED INGESTIBLE CAPSULE

"The following section provides a summary on the technological and clinical foundations for AI-powered, ingestible capsules that would be the foundation for future quality-related conversations."

The AI-assisted ingestion capsules are a fusion of miniaturized imaging equipment and sophisticated software algorithms that help in non-invasive imaging and analysis of the human gastrointestinal tract. These are based upon traditional wireless capsules but now involve artificial intelligence that increases sensitivity and quickens reading times.[1,5]

Capsule Hardware Components and Functionality

Most capsules are about 11×26 mm and commonly contain an optical lens, a CMOS image sensor, white LEDs for illumination, button batteries, a dedicated ASIC, RF transmitter, and an antenna on board within a biocompatible case. This ASIC controls, among others, the image acquisition, on-capsule processing, power management, and wireless data transmission to receivers outside the human body. It has a typical battery life in an order of approximately 8 hours. Frame rates typically vary between 2 and 6 frames per second depending on capsule motion. More recent “intelligent” capsule designs integrate various sensors like pH, pressure, and gas, among others, and feature advanced functionalities enabling active locomotion by magnetism or robotics, energy harvesting by wireless transfer, and hybrid localization for enhanced navigation and diagnostic coverage.[5,6,7]

AI Algorithms applied in Gastrointestinal Diagnostics

In the case of capsule endoscopy, the state of the art in AI solutions is currently dominated by deep learning, especially convolutional neural networks for the detection of small-bowed lesions like bleeding, ulcers, erosions, vascular malformations, polyps, and tumors, with more complex designs involving object detection networks and the newest vision transformation models capable of multiple lesion detection, bowel preparation evaluation, and the prioritization of the most important frames, with sensitivity and specificity values above 90% in research studies being reported. These systems can serve as Computer-Aided Detection (CADe) software for pointing out suspicious frames or Computer-Aided Detection (CADx) software for lesion typing and severity, with current efforts focusing on the application of Explainable AI for better transparency of the predictive process for the clinician.[8,9]

Current Clinical Applications and Impact

In clinical contexts, capsule endoscopy is mainly employed in the setting of suspected small bowel bleeding and in Crohn’s disease. As a reference standard in mucosal examination in these conditions, AI-assisted capsule endoscopy has already proved beneficial in increasing the detection rates while shortening the time required for reading the data. For the other parts of the alimentary tract, the utilization of AI in colon capsule endoscopy examination, panorama endoscopic examination, and quality scoring is being assessed. This has wide implications in the screening and management of GI disorders as AI-based capsule endoscopies become mainstream.[3,11]

These are the technological and clinical attributes that determine the quality assurance needs which will be dealt with in the next sections of this review.[2]

QUALITY ASSURANCE FUNDAMENTALS

The quality assurance for medical devices includes the systematic approach towards ensuring the performance, safety, and reliability of the various devices at all stages, from design to the surveillance stages.[2]

Principles of QA in Medical Devices

Core principles include risk-based thinking, process validation, traceability, and continuous improvement, as outlined in ISO 13485 which mandates documented procedures for design controls, supplier qualification, and adverse event reporting in device manufacturing. These principles ensure traceability from capsule manufacturing to clinical outcomes, critical for post-market surveillance of AI performance drift. They adapt traditional manufacturing QA to diagnostic devices by emphasizing clinical utility metrics like diagnostic accuracy (>90% sensitivity reported for AI capsules) alongside physical durability.[2]

Unique QA Challenges for AI-Enabled Ingestible Capsules

AI introduces non-deterministic behavior, algorithmic bias (up to 15% variation across demographics), and "black box" opacity, exacerbated by physiological variables including GI transit unpredictability, battery failure risks (2-5% retention rates), and variable image quality from bowel preparation. Unlike static devices, capsules require real-time QA for wireless data integrity and locomotion reliability, with edge AI processing demanding onboard validation absent in conventional endoscopy.[1]

Relevant Frameworks and Standards

• ISO 13485:2016: Establishes the requirements of quality management systems for medical devices, which has been applied to include risk management (ISO 14971) into the process of ensuring capsule biocompatibility and sterility.[2]
• FDA Software as a Medical Device (SaMD) Guidance (2023 update): This categorizes AI capsules in the category of Class II/III SaMD, which involves clinical performance demonstration; cybersecurity controls, FDA 2023 draft; and lifecycle management on account of algorithm changes.[11]
• IEC 62304: Software lifecycle processes adapted to AI validation; unit testing for CNN models would present specificity ranging from 97 to 100% for the detection of lesions.[12]
• IEC 61508 / ISO/IEC 42001: New AI safety standards, which cover explainability (SHAP analysis) and bias mitigation for autonomous diagnostics.[13]

These frameworks provide the foundation for a cohesive QA ecosystem, bridging hardware reliability with AI trustworthiness essential for clinical adoption.[2]

KEY QUALITY ASSURANCE COMPONENTS IN AI CAPSULE SYSTEMS

The essential QA elements must be considered from a hardware perspective, a perspective of artificial intelligence model development, a perspective of data flow, and from a perspective of clinical practice, where they must be viewed as an entire system to ensure proper capsule function.[1]

1. Hardware Quality Controls

Hardware QA ensures the capsule is mechanically safe, reliable, and capable of capturing diagnostic?grade data throughout GI transit.

• Consistent manufacturing: A consistent manufactured product, edge smooth, seal tight, reducing leakage, tested during batch tests. [14]
• Biocompatibility: The biocompatibility of materials to be biologically inert, non-toxins, and to meet ISO-10993 criteria - biocompatibility testing, such as cytotoxicity, sensitivity, and irritation. [14]
• Battery performance: Silver oxide or lithium-based cells typically offer 8–12 hours of operation; QA requires verification of operating time under worst-case imaging and transmission load, thermal safety, and protection against leakage or swelling.[5,6,15]

2. AI Algorithm Validation

AI validation confirms that capsule-generated interpretations are accurate, robust, and clinically trustworthy.

• Benchmarking with annotated datasets: Training and testing on well-labeled video/image datasets (e.g., small-bowel bleeding, ulcers, Crohn’s lesions) with expert consensus labels; external validation is mandatory to avoid overfitting.[3]

• Performance metrics: Sensitivity and specificity for key targets such as P2 bleeding lesions commonly reach 90–99% and 94–100%, respectively, with diagnostic accuracy above 90%; F1 score, AUC, and NPV are also critical in QA documentation.[3,16]
• Explainability: Use of attention maps or saliency overlays to highlight image regions influencing AI decisions, improving clinician verification and satisfying emerging expectations around explainable AI in capsule pathways.[17]

3. Data Management

Data QA secures the full pipeline from image capture to reporting, ensuring integrity, security, and regulatory compliance.

• Image/data quality control: Automated checks for blur, underexposure, and incomplete views; assessment of bowel cleansing quality and frame-rate sufficiency to guarantee lesion visibility throughout the small bowel.[15,18]

• Transmission integrity and security: Reliable wireless links (e.g., RF/BLE) with error detection, retransmission policies, and end?to?end encryption to avoid packet loss and unauthorized interception during capsule–receiver–cloud communication.[17,18]
• Privacy and compliance: Storage and processing architectures must meet GDPR and HIPAA requirements—access control, audit logging, breach notification, and consent management—especially when capsule data are reviewed remotely or stored in the cloud.[10,18]

4. Clinical Workflow Integration

QA also concerns how the AI capsule fits into daily practice, from reading to remote follow?up.

• Training and interpretation: Structured training for gastroenterologists and nurses on AI-assisted reading, including recognition of AI limitations, override procedures, and standardized reporting templates to maintain consistency and reduce inter?reader variability.[1]

• Error reporting and feedback: Formal processes for flagging false positives/negatives, correlating with follow?up endoscopy or surgery, and feeding this information back into QA reviews and potential model updates.[1]
• Telemedicine and remote monitoring: Validated remote reading platforms with encryption, robust authentication, and documented performance equivalence to on?site reading; these systems enable home?based capsule studies while maintaining data protection and diagnostic quality.[10,18]

VALIDATION AND VERIFICATION STRATEGIES

Validation and verification ensure that AI-enabled capsule systems are safe, effective, and maintain performance across their lifecycle, from preclinical experiments to real world use.[5]

Preclinical Testing Methods :- Preclinical validation typically starts with benchtop and phantom studies, where GI tract mimicking models (straight and curved paths, variable friction, and fluids) are used to test capsule locomotion, orientation stability, and image quality under controlled conditions. Animal models, especially porcine GI tracts, are employed to evaluate in vivo safety, transit behavior, retention risk, and feasibility of imaging or sensing, including specialized capsules such as ultrasound or thermometric devices.[19,20,21]

Clinical Trials Design for Efficacy and Safety :- Prospective clinical trials usually compare AI-assisted capsule reading with standard expert reading as a reference, focusing on endpoints such as lesion detection rate, per?patient sensitivity and specificity, reading time, and adverse event rates (retention, aspiration, obstruction). Multicenter designs and adequately powered sample sizes enable assessment of generalizability across different populations and centers, while subanalyses examine performance across bowel preparation quality, lesion types, and indications such as suspected small?bowel bleeding or Crohn’s disease.[3,22]

Post-Market Performance Monitoring and Real-World Evidence :- After approval, real?world evidence is gathered through registries, routine clinical data, and post?market surveillance systems that track diagnostic yield, complication rates, and device or AI malfunctions over time. Real?world monitoring also supports detection of performance drift, rare failure modes, and biases that may not appear in trials, prompting periodic re?validation or updates to labeling, user training, or technical controls.[23,24,25]

Continual Learning and Software Updates Under QA Frameworks :- Adaptive or continually learning AI models pose regulatory challenges because performance can change after deployment, leading agencies such as the FDA and MHRA to propose “predetermined change control plans” and lifecycle oversight for AI?SaMD. Within a QA ecosystem, any model updates should follow a controlled process: versioning, pre?release validation on locked datasets, cybersecure deployment, and post?update monitoring with rollback capability if predefined safety or accuracy thresholds are not met.[23,24,25]

RISK MANAGEMENT AND ETHICAL CONSIDERATIONS

Risk management in AI capsule systems must address technical failure modes and broader ethical, legal, and social implications to ensure safe, fair, and trustworthy use in GI diagnostics.[5]

Failure Mode Effects Analysis (FMEA) for AI Capsule Systems

FMEA can be applied across the capsule pathway (prescription, ingestion, recording, AI reading, reporting) to systematically list potential failures, such as capsule retention, battery depletion, wireless link loss, mislabeling of critical lesions, or incorrect report generation, and score each by severity, occurrence, and detectability to derive a risk priority number (RPN). High?RPN items (e.g., missed active bleeding due to AI misclassification or unencrypted cloud upload failure) are then targeted with controls such as redundant alarms, human over?read policies, standardized escalation pathways, and equipment maintenance or recall procedures adapted from endoscopy FMEA case studies.[26,27]

Cybersecurity Risk Assessments

AI capsule systems create an Internet of Medical Things chain—from ingestible device to body?worn recorder, hospital network, and cloud platform—that is vulnerable to eavesdropping, spoofing, ransomware, and unauthorized access if not properly secured. Cybersecurity risk assessment requires threat modeling (assets, attack vectors, attacker motives), enforcement of end?to?end encryption, strong authentication, role based access control, patch management, and incident response plans consistent with guidance for connected medical devices and cloud based capsule platforms.[18,28,29]

Addressing AI Bias, Transparency, and Trustworthiness

Capsule endoscopy AI models may perpetuate disparity when trained on unrepresentative data, resulting in variances in performance based on gender, age, or ethnic origin, which have been earmarked as areas that demand priority attention in current consensus statements in GI endoscopy. Trustworthy AI development necessitates diverse training sets, the provision of performance data for subgroups, the use of probability output scales, and the application of transparency aids such as graphical and textbased descriptions of data sets and limitations, facilitating the practitioner in deciding when to follow or override AI algorithm recommendations and preventing automation bias and algorithm dependence.[30,31]

Patient Consent, Data Ethics, and Regulatory Compliance

The ethical use of AI capsules requires specific, informed consent regarding not only the capsule process but also AI engagement, cloud storage, and secondary use of de-identified data to enhance models that patients consider complex, as found in recent studies of consent. AI use of data also requires consistency with GDPR, HIPAA, and developing regulations regarding AI, which relate to data minimization, limiting purposes of processing, conducting Data Protection Impact Assessments (DPIA), and recognizing patient rights to access, correct, and erase data, as found to present substantial legal and reputational risks to trust in AI-assisted diagnoses.[32,33]

REGULATORY LANDSCAPE AND STANDARDIZATION EFFORTS

AI-enabled capsule endoscopy is at the crossroads between the regulation of ingestible medical devices and Software as a Medical Device, and most countries are still using existing device and software regulations, rather than developing specific regulations concerning capsules. The performed narrative reviews of SaMD in digestive healthcare underline that AI analysis software for standalone capsule images is considered to be a medical device in its own right, requiring evidence of clinical validity, robust data protection, and integration into existing endoscopy workflows. Recent reviews on capsule endoscopy and ingestive devices note that even though electromagnetic safety, radiofrequency transmission, and interoperability (e.g., with hospital PACS/EHR systems) are covered by traditional device and telecom standards, no global unified standard exists that specifically addresses AI-driven, wireless ingestible imaging platforms.[34,35,36]

Harmonization of quality assurance standards for AI?enabled capsules is therefore difficult, because regulators and professional societies use different performance metrics, validation thresholds, and expectations for external testing. Some reviews highlight that, unlike drugs, AI models for capsule endoscopy do not undergo a fully standardized training and validation framework; instead, manufacturers rely on heterogeneous datasets, endpoints, and study designs, which complicates comparisons and international homologation of devices. Authors argue that without common benchmarks and agreed definitions of clinically acceptable sensitivity, specificity, and miss rates for capsule AI, it is challenging to define universal QA indicators or to transfer approvals between regions.[1,37]

Emerging guidance is starting to bridge this gap: papers on SaMD in digestive healthcare and on AI in capsule endoscopy propose clear AI governance, with model documentation that is transparent, continuous post?market evaluation, and standardized reporting of algorithm updates within a regulated change?control framework. Current analyses of the regulatory environment for ingestible devices also point to adaptive processes and integration efforts from the perspective of regulatory bodies, industry, and practicing clinicians with a view to afford learning AI systems, performance data from the real world, and a new crop of versatile ingestible sensors new regulatory frameworks in the future while continuing safety and quality at the same high level. [8, 35, 38]

FUTURE DIRECTIONS AND RESEARCH NEEDS

Next-generation QA environments should be developed in concert with next-generation ingestive technologies, as well as guidelines that actively strive to ensure that efficacy and safety are maintained.[1,5]

Emerging Technologies

Edge AI enables onboard processing within capsules, reducing data transmission latency to milliseconds and enabling real-time lesion detection without cloud dependency, though this demands miniaturized computing with robust thermal and power management. Multi-sensor capsules integrating optical imaging with pH, pressure, temperature, and gas sensors promise comprehensive GI phenotyping, but require unified data fusion validation protocols to ensure sensor-AI interoperability. Autonomous therapeutic capsules with biopsy needles, drug delivery micropumps, or hemostasis mechanisms represent the horizon, necessitating QA frameworks that verify not only diagnostic accuracy but also interventional precision and sterility under uncontrolled GI conditions.[5,14,17,34]

Quality Assurance Challenges Posed by Evolving Capabilities

Edge and multi-modal AI introduce verification complexity, as models must perform reliably across heterogeneous sensor inputs, variable computational constraints, and physiological extremes without human oversight. Therapeutic autonomy amplifies risk profiles, demanding real-time safety interlocks, fail-safe expulsion mechanisms, and traceability for adverse events that span diagnosis-to-treatment sequences. Continual learning exacerbates regulatory uncertainty, as post-deployment adaptation challenges the "locked algorithm" paradigm central to current SaMD approvals.[5,23,38,39]

Recommendations for Standardized QA Frameworks and Collaborative Research

Create international standards benchmarks (e.g., multi-sensor GI atlas maps with expert consensus annotations), performance levels graded according to clinical risk to support replicable validation across all devices. Develop intelligent QA systems based on Predetermined Change Control Plans (PCCPs), with autonomous drift detection and rollback strategies. Mobilize global consortia involving all stakeholders—regulators (IMDRF), industry, healthcare professionals, and patients—to collaboratively develop standards for ingestible AI solutions, with high priority to real-world evidence registries stratified according to device “generation” and patient characteristics. Finally, high priority should be assigned to advancing research on explainable fusion approaches and paradigms involving collaboration between humans and AI agents to retain clinician control with autonomous capsule systems.[31,40]

CONCLUSION

This review demonstrates that while AI-enabled ingestible capsules achieve impressive diagnostic performance (90-99% sensitivity/specificity), their safe clinical translation demands a comprehensive quality assurance ecosystem spanning hardware reliability, algorithm validation, data integrity, clinical workflows, risk management, and adaptive regulatory compliance.[1]

An integrated QA framework is essential—not merely additive quality checks, but a interconnected system ensuring traceability from manufacturing through real-world deployment, addressing AI-specific challenges like non-determinism, bias propagation, and continual learning under physiological variability. Fragmented approaches risk eroding clinician trust, amplifying safety gaps, and stalling innovation in multi-sensor therapeutic capsules.[35,38]

It is a grave necessity to have a multidisciplinary collaborative effort between regulators (IMDRF/FDA), industries, gastroenterologists, ethicists, and patients to set a standard and benchmark dataset and post-market registry. Further research should focus on having a comprehensive understanding of artificial intelligence, human-artificial intelligence synergy, and adaptive QA models to harness the full potential of autonomous GI diagnostic systems to ensure successful outcomes for patients.

REFERENCES

1. Giordano A, Romero-Mascarell C, González-Suárez B, Guarner-Argente C. Integration of Artificial Intelligence-Enhanced Capsule Endoscopy in Clinical Practice: A Review of Market-Available Tools for Clinical Practice. Dig Dis Sci. 2025 Sep;70(9):2966-2976. doi: 10.1007/s10620-025-09099-4.
2. Daniele Salvi, Stefania Piccirelli, Maria Parmigiani, Paola Cesaro, Cristiano Spada. How to measure quality in capsule endoscopy. Best Practice & Research Clinical Gastroenterology, Volume 76, 2025, 102012, ISSN 1521-6918, https://doi.org/10.1016/j.bpg.2025.102012.
3. Bin Y, Peng R, Lee Y, Lee Z and Liu Y (2025). Artificial intelligence-assisted capsule endoscopy for detecting lesions in Crohn’s disease: a systematic review and meta-analysis. Front. Artif. Intell. 8:1531362. Doi: 10.3389/frai.2025.1531362.
4. Raju Vaishya, Mahesh Kumar Goenka, Anupam Sibal, Sujoy Kar and Sangita Reddy. National Board of Examinations - Journal of Medical Sciences. Volume 3, Issue 6, Pages 729 -745, June 2025. DOI 10.61770/NBEJMS.2025.v03.i06.011.
5. Cao, Q., Deng, R., Pan, Y. et al. Robotic wireless capsule endoscopy: recent advances and upcoming technologies. Nat Commun 15, 4597 (2024). https://doi.org/10.1038/s41467-024-49019-0.
6. Manolescu VD, AlZu’bi H, Secco EL. Review on endoscopic capsules. Explor Digit Health Technol. 2024;2:346–59. https://doi.org/10.37349/edht.2024.00033.
7. Mehedi, I.M, Rao, K.P, Alotaibi, F.M.Alkanfery, H.M. Intelligent Wireless Capsule Endoscopy for the Diagnosis of Gastrointestinal Diseases. Diagnostics 2023, 13, 1445. https://doi.org/10.3390/diagnostics13081445.
8. George AA, Tan JL, Kovoor JG, Lee A, Stretton B, Gupta AK, Bacchi S, George B, Singh R. Artificial intelligence in capsule endoscopy: development status and future expectations. Mini-invasive Surg. 2024;8:4. http://dx.doi.org/10.20517/2574-1225.2023.102.
9. Jiaqi Bo, Zishuai Liu, Yiran wang, Qingfeng Luo. Research progress on the application of artificial intelligence in colonoscopy, Gastroenterology & Endoscopy, Volume 3, Issue 4, 2025, Pages 263-270, ISSN 2949-7523, https://doi.org/10.1016/j.gande.2025.10.003.
10. Lei, I. I., Valdivia, P. C., Arasaradnam, R. P., & Koulaouzidis, A. (2025). The future of colon capsule endoscopy: an AI-enabled model linking telemedicine, cloud infrastructure, and colonoscopy. Colorectal Cancer, 14(1). https://doi.org/10.1080/1758194X.2025.2533110.
11. Helmut Messmann and Raf Bisschops. Expected value of artificial intelligence in gastrointestinal endoscopy: European Society of Gastrointestinal Endoscopy (ESGE) Position Statement. ISSN 0013-726X. DOI 10.1055/a-1950-5694.
12. Song Y-q, Mao X-l, Zhou X-b, He S-q, Chen Y-h, Zhang L-h, Xu S-w, Yan L-l, Tang S-p, Ye L-p and Li S-w (2021). Use of Artificial Intelligence to Improve the Quality Control of Gastrointestinal Endoscopy. Front. Med. 8:709347. doi:10.3389/fmed.2021.709347.
13. Romain Leenhardt, Ervin Toth, Xavier Dray and Uri Kopylov. Key research questions for implementation of artificial intelligence in capsule endoscopy. 2022, Vol. 15: 1–8 DOI:10.1177/17562848221132683.
14. Xiangyu Wei, Peipei Xi, Minjie Chen, Ya Wen, Hao Wu, Li Wang, Yujuan Zhu, Yile Ren, Zhifeng Gu. Capsule robots for the monitoring, diagnosis, and treatment of intestinal diseases, Materials Today Bio, Volume 29, 2024, 101294, ISSN 2590-0064, https://doi.org/10.1016/j.mtbio.2024.101294.
15. Su CC, Chou CK, Mukundan A, Karmakar R, Sanbatcha BF, Huang CW, Weng WC, Wang HC. Capsule Endoscopy: Current Trends, Technological Advancements, and Future Perspectives in Gastrointestinal Diagnostics. Bioengineering (Basel). 2025 Jun 4;12(6):613. doi: 10.3390/bioengineering12060613.
16. Giordano A, Escapa M, Urpí-Ferreruela M, Casanova G, Fernández-Esparrach G, Ginès À, Llach J, González-Suárez B. Diagnostic accuracy of artificial intelligence-aided capsule endoscopy (TOP100) in overt small bowel bleeding. Surg Endosc. 2023 Oct;37(10):7658-7666. doi: 10.1007/s00464-023-10273-w.
17. Sahafi A, Wang Y, Rasmussen CLM, Bollen P, Baatrup G, Blanes-Vidal V, Herp J, Nadimi ES. Edge artificial intelligence wireless video capsule endoscopy. Sci Rep. 2022 Aug 12;12(1):13723. doi: 10.1038/s41598-022-17502-7.
18. Costigan C, Walker C, O'Connell J, Omallao E, Manoharan T, Eagle N, Bailey Y, O'Hara F, Mc Namara D. Cloud technology and capsule endoscopy: A single-center users' experience of remote online video analysis and reporting. Endosc Int Open. 2024 Feb 15;12(2):E227-E230. doi: 10.1055/a-2226-0755.
19. Peker, F., & Ferhano?lu, O. (2021). Multi-Capsule Endoscopy: An initial study on modeling and phantom experimentation of a magnetic capsule train. Journal of Medical and Biological Engineering, 41(3), 315-321. https://doi.org/10.1007/s40846-021-00610-6.
20. Kopácová M, Tachecí I, Kvetina J, Bures J, Kunes M, Spelda S, Tycová V, Svoboda Z, Rejchrt S. Wireless video capsule enteroscopy in preclinical studies: methodical design of its applicability in experimental pigs. Dig Dis Sci. 2010 Mar;55(3):626-30. doi: 10.1007/s10620-009-0779-3.
21. Fakhr Abdollahi, A., Shaheed, M. H., Thaha, M. A., & Vepa, R. (2024). A review of modeling and control of remote-controlled capsule endoscopes. Expert Review of Medical Devices, 21(4), 293–306. https://doi.org/10.1080/17434440.2024.2336135.
22. O'Hara FJ, Mc Namara D. Capsule endoscopy with artificial intelligence-assisted technology: Real-world usage of a validated AI model for capsule image review. Endosc Int Open. 2023 Oct 11;11(10):E970-E975. Doi: 10.1055/a-2161-1816.
23. Santra S, Kukreja P, Saxena K, Gandhi S, Singh OV. Navigating regulatory and policy challenges for AI enabled combination devices. Front Med Technol. 2024 Nov 28;6:1473350. doi: 10.3389/fmedt.2024.1473350.
24. Ahmed El-Sayed, Laurence B. Lovat, Omer F. Ahmad. Clinical Implementation of Artificial Intelligence in Gastroenterology: Current Landscape, Regulatory Challenges, and Ethical Issues, Gastroenterology, Volume 169, Issue 3, 2025, Pages 518-530, ISSN 0016-5085, https://doi.org/10.1053/j.gastro.2025.01.254.
25. Palaniappan K, Lin EYT, Vogel S. Global Regulatory Frameworks for the Use of Artificial Intelligence (AI) in the Healthcare Services Sector. Healthcare (Basel). 2024 Feb 28;12(5):562. doi: 10.3390/healthcare12050562.
26. He H, Wang F, Zhou C, Liu X. Optimization of the emergency endoscopy process for patients with esophagogastric variceal bleeding using failure mode and effect analysis. Am J Transl Res. 2023 May 15;15(5):3365-3374.
27. Zuber Mujeeb Shaikh. (2020). A Case Study on Recall of used Scopes in the Endoscopy Department by using a Failure Mode & Effect Analysis (FMEA) Proactive Risk Management. International journal of health sciences and pharmacy (IJHSP), 4(1), 13–24. https://doi.org/10.5281/zenodo.3766843.
28. Shravani Amar , Shaik Hussain Shaik Ibrahim , P. Anitha. A Comprehensive Survey on IoMT-Based Gastrointestinal Disease Diagnosis Combining Data Security and Automated Classification, NeuroQuantology|June 2024| Volume22 |Issue 3 | Page 340-354|doi: 10.48047/nq.2024.22.3.NQ24037.
29. Vikas Hassija, Vinay Chamola, Balindam Chandra Bajpai,  Naren, Sherali Zeadally. Security issues in implantable medical devices: Fact or fiction?, Sustainable Cities and Society, Volume 66, 2021, 102552, ISSN 2210-6707, https://doi.org/10.1016/j.scs.2020.102552.
30. Ramoni D, Scuricini A, Carbone F, Liberale L, Montecucco F. Artificial intelligence in gastroenterology: Ethical and diagnostic challenges in clinical practice. World J Gastroenterol. 2025 Mar 14;31(10):102725. doi: 10.3748/wjg.v31.i10.102725.
31. Campion JR, O'Connor DB, Lahiff C. Human-artificial intelligence interaction in gastrointestinal endoscopy. World J Gastrointest Endosc. 2024 Mar 16;16(3):126-135. Doi: 10.4253/wjge.v16.i3.126.
32. Khadijeh Moulaei, Saeed Akhlaghpour, Farhad Fatehi. Patient consent for the secondary use of health data in artificial intelligence (AI) models: A scoping review, International Journal of Medical Informatics, Volume 198, 2025, 105872, ISSN 1386-5056, https://doi.org/10.1016/j.ijmedinf.2025.105872.
33. Pham T. Ethical and legal considerations in healthcare AI: innovation and policy for safe and fair use. R Soc Open Sci. 2025 May 14;12(5):241873. Doi: 10.1098/rsos.241873.
34. Manolescu VD, AlZu’bi H, Secco EL. Review on endoscopic capsules. Explor Digit Health Technol. 2024;2:346–59. https://doi.org/10.37349/edht.2024.00033.
35. Mascarenhas M, Martins M, Ribeiro T, Afonso J, Cardoso P, Mendes F, Cardoso H, Almeida R, Ferreira J, Fonseca J, Macedo G. Software as a Medical Device (SaMD) in Digestive Healthcare: Regulatory Challenges and Ethical Implications. Diagnostics (Basel). 2024 Sep 23;14(18):2100. doi: 10.3390/diagnostics14182100.
36. Mandsberg NK, Christfort JF, Kamguyan K, Boisen A, Srivastava SK. Orally ingestible medical devices for gut engineering. Adv Drug Deliv Rev. 2020;165-166:142-154. Doi: 10.1016/j.addr.2020.05.004.
37. Piccirelli S, Salvi D, Pugliano CL, Tettoni E, Facciorusso A, Rondonotti E, Mussetto A, Fuccio L, Cesaro P, Spada C. Unmet Needs of Artificial Intelligence in Small Bowel Capsule Endoscopy. Diagnostics (Basel). 2025 Apr 25;15(9):1092. Doi: 10.3390/diagnostics15091092.
38. Babic B, Glenn Cohen I, Stern AD, Li Y, Ouellet M. A general framework for governing marketed AI/ML medical devices. NPJ Digit Med. 2025 May 31;8(1):328. Doi: 10.1038/s41746-025-01717-9.
39. Liron Pantanowitz, Matthew Hanna, Joshua Pantanowitz, Joe Lennerz, Walter H. Henricks, Peter Shen, Bruce Quinn, Shannon Bennet, Hooman H. Rashidi, Regulatory Aspects of Artificial Intelligence and Machine Learning, Modern Pathology, Volume 37, Issue 12, 2024, 100609, ISSN 0893-3952, https://doi.org/10.1016/j.modpat.2024.100609.
40. Sami Elamin, Shreya Johri, Pranav Rajpurkar, Enrik Geisler, Tyler M Berzin, From data to artificial intelligence: evaluating the readiness of gastrointestinal endoscopy datasets, Journal of the Canadian Association of Gastroenterology, Volume 8, Issue Supplement_2, March 2025, Pages S81–S86, https://doi.org/10.1093/jcag/gwae041.