For years, artificial intelligence (AI) has been anticipated to revolutionize the field of medicine, yet a recent thorough review indicates that we may be at a pivotal moment where technological progress significantly outpaces its clinical application. Researchers Deepika Yadav, Pooja Yadav, and Hemant Yadav conducted an extensive analysis, published in Discover Informatics, aiming to chart how various AI technologies, including machine learning, deep learning, the Internet of Things (IoT), and blockchain, are utilized within healthcare and to identify prevalent gaps.
Their findings reveal a sobering reality: many healthcare AI systems currently outlined in literature seem to be stuck at intermediate stages of development, often only validated in labs or pilot projects rather than being routinely implemented in clinical settings.
Employing the PRISMA framework—deemed the gold standard for systematic reviews—the researchers began with a pool of 1,860 scholarly articles sourced from leading databases like PubMed and IEEE Xplore. After carefully eliminating duplicates and scrutinizing titles and abstracts, they narrowed it down to 331 articles for a detailed review, ultimately including 108 studies that focused on literature published between 2021 and 2026 to highlight the latest in AI healthcare advancements. Each study was assessed using the Technology Readiness Level (TRL) framework, originally created by NASA, which categorizes technologies based on their maturity and readiness for deployment.
The TRL analysis shed light on one of the review's key conclusions: a significant majority of healthcare AI tools are in the TRL 3 to TRL 5 range. This indicates that they remain at the conceptual phase or as early prototypes and have not transitioned to large-scale clinical implementation, which aligns with TRL 7 through 9. This suggests that AI in healthcare is still largely in a state of flux, transitioning from theoretical applications to tangible clinical use, with TRLs 4 and 5 being the most prevalent maturity stages found.
The review categorizes AI applications in the healthcare sector into an organized hierarchy. It details how machine learning, which develops algorithms for data analysis, is already employed in tasks such as monitoring patient facial expressions and diagnosing illnesses via techniques like Support Vector Machines. On the other hand, deep learning—characterized by its use of advanced neural networks—has made significant strides, particularly in medical imaging. For example, systems like Google's DeepMind and IBM's Watson have achieved results in identifying malignant tumors that rival those of experienced radiologists.
The scope of applications described in the review is extensive. Deep learning has transformed medical image analysis, which is crucial across various fields such as radiology and dermatology, while predictive models assist in risk assessments to improve patient safety and reduce errors. Additionally, AI facilitates faster diagnostics via automated screening and enhances treatment plans while addressing healthcare fraud through blockchain technology, thereby improving patient engagement and preemptive care models.
Moreover, the review discusses the fusion of AI with other advanced technologies. The IoT enables real-time data sharing from wearable devices, paving the way for remote monitoring and telehealth innovations, especially when combined with 5G networks. Blockchain, primarily known for its role in cryptocurrencies, secures electronic medical records, enhancing integrity and interoperability across systems. There are studies outlined that explore the integration of AI and blockchain for managing health records along with technologies that create digital twins tailored for individual patients.
Nevertheless, the report underscores significant challenges facing the field. Data collection issues impede progress, as patient confidentiality concerns and stringent regulations such as GDPR complicate data access and research collaboration. Consequently, the quality of data remains inconsistent, adversely affecting AI performance. Algorithm-related challenges, including bias in training data and the black-box phenomenon—which renders AI outputs incomprehensible even to developers—raise serious questions about clinical accountability and the reliability of AI-assisted recommendations.
Ethical dilemmas further complicate the technological landscape. Ascribing liability for AI-driven errors is challenging due to the opaque nature of AI decision-making. Additionally, while regulatory bodies like the FDA work toward establishing frameworks for evaluating healthcare AI, concerns regarding job security among healthcare professionals persist, emphasizing the need for role evolution rather than job loss in the face of AI advancements. The transition from AI research to real-world application is fraught with barriers, such as limited representation in study populations and the necessity for integration that seamlessly fits into existing healthcare workflows. There is also a notable gap in research focused on mental health, chronic conditions, and elder care, and the synergy between AI, IoT, and blockchain remains largely unexplored.
In light of these findings, the authors lay out a clear agenda for the future of medical AI. They highlight the urgent need for clinical validation and effective deployment of AI tools, advocating for research that focuses on explainable AI to foster clinician trust, federated learning for secure data sharing, and the use of large language models for decision support and data synthesis. They also propose advancements in digital twin technologies to enhance individualized patient care while recommending stricter regulatory measures and extensive clinical trials to ensure safe integration into healthcare practices. The overarching message conveys a cautious optimism: while AI systems have already shown considerable promise in improving diagnostics and healthcare delivery, significant efforts are needed to bridge the gap between research efficacy and practical clinical application.
