AI has come a long way in healthcare, still has a long way to go: Research recap

As AI continues its march through healthcare organizations around the world, the notion that it will replace human workers fades but does not disappear. Still, the replacement scenario isn’t becoming any likelier. 

Two researchers in the U.K. observe as much in a new literature review emphasizing current trends, theoretical insights and future directions.

“From early rule-based systems to advanced deep learning algorithms, AI has consistently demonstrated capabilities that rival … human expertise—particularly in imaging, predictive analytics and drug discovery,” write Dinesh Deckker of Wrexham University and Subhashini Sumanasekara of the University of Gloucestershire.

However, they stress, by advancing AI within a framework of ethics, inclusivity and evidence-based practice, stakeholders “can ensure that this transformative technology delivers on its promise to enhance, rather than replace, human-centered care.” 

The paper went up this month in World Journal of Advanced Research and Reviews. The authors present their discussion section as a Q&A:  

1. How has AI evolved in the field of medicine, and what are the key milestones in its development?

AI in medicine has evolved significantly from early rule-based systems like Mycin in the 1970s to contemporary models powered by deep learning and big data analytics, Deckker and Sumanasekara note. “Early systems were limited by rigid logic and lack of adaptability,” they add. “Still, recent advancements—such as convolutional neural networks (CNNs), natural language processing (NLP), and generative AI—have enabled the automation of complex clinical tasks.” More:  

‘Major milestones include FDA-approved AI diagnostics, real-time predictive tools and the integration of AI into electronic health record (EHR) systems.’

2. In what ways is AI currently applied in medical diagnostics, imaging and clinical decision-making?

Artificial intelligence is widely utilized in medical diagnostics to aid in disease detection and classification, particularly in radiology, dermatology and pathology, the authors point out, adding that AI-enhanced imaging systems precisely detect anomalies such as tumors or lesions. “In clinical decision-making,” they remind, AI-driven Clinical Decision Support Systems (CDSS) offer real-time recommendations, flag adverse drug interactions and guide clinicians toward evidence-based decisions.” Meanwhile:  

‘NLP models extract insights from unstructured data, making patient information more accessible and actionable.’ 

3. How does AI contribute to personalized and predictive medicine, and what are the implications for patient outcomes?

AI facilitates personalized medicine by analyzing genetic, behavioral and clinical data to recommend individualized treatments. In predictive medicine, machine learning algorithms anticipate disease progression, readmission risk and patient outcomes. “This enables proactive care planning and resource optimization,” The implications are profound: improved therapeutic accuracy, reduced complications, and enhanced quality of life. 

‘The role of AI in biomarker discovery and pharmacogenomics further supports the development of tailored therapeutic interventions.’ 

4. What are the ethical, legal, and regulatory challenges associated with integrating AI into healthcare systems?

The ethical and legal dimensions of AI are central to its responsible deployment, Deckker and Sumanasekara emphasize. “Key issues include data privacy breaches, algorithmic bias, lack of transparency and ambiguity in liability,” they write. “Ethical principles, such as autonomy and justice, are tested when AI outputs are unexplainable or yield unequal outcomes across populations.” Further:  

‘Current legal frameworks often lag behind AI innovations, necessitating adaptive regulatory models that balance innovation and patient safety and rights.’

5. How can AI be effectively implemented in low-resource and global health settings to address disparities in care delivery?

“AI can play a vital role in addressing healthcare inequities by enabling mobile diagnostics, telehealth and automated triage in underserved regions,” the authors maintain. “AI applications with minimal requirements enable the early detection of diseases and the monitoring of disease occurrence through systems that operate with limited facilities or no internet connections.”

‘The success of AI programs depends on adapting systems to local contexts while developing inclusive databases alongside partnerships with local community members to prevent the perpetuation of existing social inequalities.’ 

6. What are the current limitations of AI in medicine, and what future directions should research and policy focus on to ensure responsible and effective use?

Current limitations include data silos, interoperability challenges, algorithmic opacity and clinician distrust, Deckker and Sumanasekara point out. “There is a lack of longitudinal and intervention studies assessing the sustained impact of AI,” they add. “Future efforts should focus on developing explainable AI, refining ethical frameworks and expanding interdisciplinary education and training for healthcare providers.” More: 

‘Policymakers must also develop adaptive regulations that define accountability, standardize validation and promote the equitable integration of AI.’

The paper is available in full for free.