Artificial intelligence has promised to remake medicine for decades, but a sweeping new systematic review suggests the field is at a decisive turning point: the technology is advancing faster than its ability to reach the clinic. Researchers Deepika Yadav, Pooja Yadav and Hemant Yadav, publishing in the journal Discover Informatics, have compiled one of the most comprehensive maps to date of how machine learning, deep learning, the Internet of Things and blockchain are being deployed across healthcare, and where the gaps remain. Their conclusion is striking. Most healthcare AI systems described in the recent literature are still trapped at intermediate stages of technological maturity, validated in laboratories or pilot settings rather than in routine clinical practice.
The team applied the PRISMA framework, the gold-standard protocol for systematic reviews, to sift through the scientific record. Beginning with 1,860 records drawn from major databases including PubMed, IEEE Xplore, Scopus, ScienceDirect, ACM, Springer and Wiley, they removed 657 duplicates and screened 1,203 titles and abstracts. After full-text eligibility assessment of 331 articles, 108 studies were ultimately included in the qualitative and quantitative synthesis. The review focused on literature published between 2021 and 2026, capturing the most recent wave of AI-driven healthcare innovation. Each selected study was then evaluated using the Technology Readiness Level framework, a nine-stage maturity scale originally developed by NASA, to determine how close each technology actually is to real-world deployment.
The TRL analysis produced one of the review’s most consequential findings. The majority of healthcare AI technologies cluster between TRL 3 and TRL 5, meaning they exist as conceptual frameworks, proof-of-concept demonstrations, laboratory prototypes or early tests in relevant environments. Very few studies demonstrated large-scale clinical implementation or fully operational deployment, which would correspond to TRL 7 through 9. In practical terms, the review shows that healthcare AI remains largely in transition from research-based proof-of-concept systems to genuine clinical applications, with TRL 4 and TRL 5 being the most common maturity levels observed across the literature.
Technically, the review organizes the field into a hierarchical taxonomy of AI methods and their healthcare applications. Machine learning, the older sibling of the AI family, develops data-analysis algorithms that extract features from data and improve with exposure to more examples. Techniques such as Support Vector Machines and Naïve Bayes classifiers are already being used to classify facial expressions for patient monitoring and disease diagnosis. Deep learning goes further, employing artificial neural networks with multiple hidden layers and millions or even billions of parameters. These architectures have proven so powerful in medical imaging that systems such as Google’s DeepMind and IBM’s Watson have demonstrated performance on malignant tumor detection that rivals or exceeds human radiologists, according to studies cited in the review.
The application landscape the authors map is remarkably broad. Medical image analysis, transformed by deep learning, underpins modern diagnosis, treatment and monitoring across radiology, pathology, dermatology and ophthalmology. Disease prediction and risk assessment models help clinicians anticipate dangers, identify lesion locations and reduce medical errors. Automated screening systems are accelerating diagnostics through predictive analytics, medical imaging and clinical decision support. Beyond diagnosis, AI is reshaping treatment planning and patient care, detecting healthcare insurance fraud through blockchain-empowered analytics, powering patient engagement tools that generate personalized insights, and enabling preventive care through predictive models that flag disease risks before symptoms appear. Subdomains such as robotic surgery, virtual health aides, drug discovery and remote patient monitoring round out the picture of a technology touching nearly every corner of medicine.
The review also highlights how AI is converging with other emerging technologies. The Internet of Things connects wearable devices and sensors that stream continuous patient data, enabling real-time telehealth, remote diagnosis and even remote surgery when paired with 5G networks. Blockchain, the decentralized and immutable ledger technology originally conceived for Bitcoin, offers integrity, traceability and non-repudiation for electronic medical records, securing data sharing across institutions through hash chains, digital signatures and consensus mechanisms. The authors point to studies combining AI and blockchain for secure health record management and patient identity systems, as well as digital twin technologies that create patient-specific computational models for personalized medicine, including neurosymbolic digital twins for cardiovascular disease prediction.
Yet the challenges catalogued in the review are formidable. Data collection remains a fundamental bottleneck: patient confidentiality concerns limit the availability of relevant information, and privacy regulations such as GDPR, while essential for protecting personal data, complicate research collaboration. Data quality problems, including inconsistent records, directly degrade algorithm performance. On the algorithmic side, bias in training data can distort AI outcomes, and overfitting causes models to latch onto irrelevant correlations. The notorious black-box problem, in which deep learning systems reach conclusions that even their creators cannot fully explain, undermines clinical trust and accountability. When a physician cannot understand why an AI system recommends a treatment, the reliability of medical advice itself comes into question.
Ethical and social concerns compound the technical ones. Accountability for AI errors is difficult to assign when decision-making is opaque, and universal ethical standards for healthcare AI have yet to be established, although regulatory bodies such as the FDA are developing assessment frameworks. Fear of job displacement fuels skepticism among healthcare workers, and the authors argue that roles must be transformed rather than eliminated to allow AI advancement. Clinical implementation poses its own barriers: most AI research has not been developed within actual clinical settings, generalization to diverse patient populations is complicated by small or biased training datasets, and successful adoption requires stakeholder engagement, workflow integration that does not disrupt care, and training for healthcare personnel. The review also flags that mental health, chronic disease and elder care remain understudied areas, and that integration of AI with IoT and blockchain is still restricted, leaving fertile ground for future research.
The authors chart a forward agenda that reads like a roadmap for the next decade of medical AI. Clinical validation and real-world deployment emerge as the most critical research needs, given that most current studies sit at intermediate readiness levels. Future work should prioritize explainable AI models that enhance transparency and clinician trust, federated learning that enables privacy-preserving collaboration across institutions, and large language models and generative AI for clinical decision support, medical knowledge extraction and synthetic data generation. Digital twin technologies could personalize care at the level of the individual patient, while stronger regulatory frameworks and extensive clinical validation trials are needed to guarantee safe, effective and sustainable integration. The review’s ultimate message is one of measured optimism: AI-based systems have already delivered marked improvements in diagnostic precision, tailored treatment strategies and healthcare efficiency, but building trustworthy, scalable and clinically applicable systems will require the field to close the gap between what works in the laboratory and what works at the bedside.
Subject of Research: A PRISMA-based systematic review of artificial intelligence applications, techniques and challenges in healthcare
Article Title: A PRISMA-based Systematic Review of Artificial Intelligence in Healthcare, its Applications and Challenges
Article References: Yadav, D., Yadav, P., & Yadav, H. (2026). A PRISMA-based Systematic Review of Artificial Intelligence in Healthcare, its Applications and Challenges. Discover Informatics, 1(1), Article 12. https://doi.org/10.1007/s44564-026-00009-y
Image Credits: AI Generated
DOI: 10.1007/s44564-026-00009-y
Keywords: Artificial Intelligence, Healthcare, Machine Learning, Deep Learning, IoT, Blockchain, Technology Readiness Level, PRISMA, Medical Imaging, Explainable AI, Data Privacy, Algorithmic Bias
Cite Scienmag News
APA MLA Chicago
Blake Davidson. (September 12, 2026). AI in Healthcare Poised to Transform Medicine, But Most Tools Still Stuck in the Lab. Scienmag. https://scienmag.com/ai-in-healthcare-poised-to-transform-medicine-but-most-tools-still-stuck-in-the-lab/
Blake Davidson. “AI in Healthcare Poised to Transform Medicine, But Most Tools Still Stuck in the Lab.” Scienmag, 12 September 2026, https://scienmag.com/ai-in-healthcare-poised-to-transform-medicine-but-most-tools-still-stuck-in-the-lab/. Accessed 12 September 2026.
Blake Davidson. “AI in Healthcare Poised to Transform Medicine, But Most Tools Still Stuck in the Lab.” Scienmag. September 12, 2026. https://scienmag.com/ai-in-healthcare-poised-to-transform-medicine-but-most-tools-still-stuck-in-the-lab/
Copy citation Download RIS
Tags: AI deployment in clinical practiceAI in healthcarealgorithmic biasArtificial Intelligenceblockchainblockchain healthcare solutionschallenges in translating healthcare AIData Privacydeep learningdeep learning for clinical applicationsexplainable AIgaps in AI healthcare implementationHealthcarehealthcare AI technological maturityInternet of Things in healthcareIoTMachine learningmachine learning in medicinemedical AI validation stagesMedical ImagingPRISMArecent advancements in healthcare AIsystematic review of AI healthcare toolsTechnology Readiness Level
