Prashanti Eachempati,1,2 Avinash Supe,3 Sumanth Kumbargere Nagraj,1,2 Alex Cresswell-Boyes,1 Safiya Robinson,1 Samata Yalamanchili4

Introduction

‘Artificial intelligence (AI) is a tool, the choice of how it gets deployed is ours.'1

The rapid evolution of AI is reshaping various industries, with healthcare standing at the forefront of this transformation.2 AI's potential to enhance diagnostic accuracy, streamline treatment planning, and improve patient outcomes holds immense promise in fields like medicine and dentistry.2,3 As AI systems grow increasingly capable, the ethical considerations surrounding their use in clinical practice, particularly in maintaining patient autonomy and human oversight, become critical areas of focus.4 In healthcare, where person-focused care and individualised decision-making are paramount, the unrestrained application of AI may lead to unintended consequences, risking the replacement of nuanced human judgment with rigid algorithms.4,5,6

The Human-In-The-Loop (HITL) model,7 widely used in sectors such as finance and autonomous systems, has traditionally addressed the critical need for human oversight. By embedding human intervention within automated processes, HITL ensures that technology remains a tool guided by human reasoning.

However, this model requires adaptation for healthcare, where decisions frequently demand empathy, contextual understanding, and clinical expertise beyond what algorithms can offer. This paper is structured in two parts. Part 1 introduces the envisioned Health Care Professional-In-The-Loop (HCP-ITL) framework, explaining its fundamental principles and how it adapts the HITL model to meet the unique challenges of healthcare. In Part 2, we explore the application of the HCP-ITL framework across diverse medical and dental specialties, outlining the potential roles AI could play within this model to enhance clinical care while ensuring ethical and professional oversight.

Introducing the Health Care Professional-In-The-Loop (HCP-ITL) Model

The Health Care Professional-In-The-Loop (HCP-ITL) model which we propose, as shown in Figure 1, is a framework tailored to meet healthcare's complex unique demands. Unlike the conventional HITL model, HCP-ITL actively integrates healthcare professionals, such as doctors, dentists, nurses, and other specialists, into each stage of AI-assisted decision-making. In this model, AI functions as a supportive partner, enhancing efficiency and precision without displacing the indispensable human elements of empathy and ethical judgement. The HCP-ITL model integrates healthcare professionals (HCPs) at each stage of care, from data analysis to diagnosis and treatment, ensuring that AI enhances rather than dictates healthcare practice.

Fig. 1
figure 1

HCP- ITL model

Full size image

The concept of integrating AI in healthcare with the oversight of HCPs, similar to the HCP-ITL model, is not entirely new. Many healthcare settings are already using AI to support HCPs, particularly in fields like radiology, pathology, and diagnostics.8,9,10,11 AI systems assist in analysing medical images, predicting disease risks, and even suggesting treatment plans based on large datasets.11,12,13 Nevertheless, most current applications function in a limited, task-specific capacity rather than as a fully integrated model spanning the entire patient care process as proposed in the HCP-ITL framework. This model introduces a novel, holistic framework, coining a unique approach where AI systems are designed to support HCPs and, in the future, aspire to learn from their expertise in real time. Currently, the system relies on periodic updates and retraining to incorporate insights and adjustments from clinicians. However, our vision is to advance this framework towards a real-time learning capability, enabling dynamic adaptation to clinical practices while maintaining alignment with real-world needs and ensuring person-focused care. The HCP-ITL model pictures a continuous, bidirectional learning process between HCPs and AI, a fully integrated workflow across patient journey stages, enhanced person-focused care, and dynamic, real-time quality assurance systems. Through the HCP-ITL model, we aim to shift AI from being an isolated tool to being part of a seamless adaptive ecosystem that strengthens healthcare delivery while preserving the critical, irreplaceable role of HCPs. Table 1 highlights the differences between current AI models and the envisioned future model under HCP-ITL.

Table 1 Current models and the envisioned models
Full size table

Steps in the HCP-ITL Model

Training and adaptive integration

The HCP-ITL model begins with a vision of HCPs actively training AI systems to interpret patient data in alignment with evidence-based practices, including clinicians' expertise and patient-specific factors. Unlike traditional AI systems trained on static datasets, the HCP-ITL model aspires to integrate a continuous, bidirectional feedback loop between HCPs and AI. In practice, while real-time adaptation remains a goal requiring technological advancements, the model prioritises periodic updates based on new clinical insights, evolving standards, and patient needs. This approach is designed to ensure AI systems are dynamically validated and aligned with clinical best practices.

To achieve this, the model emphasises strict adherence to GDPR and patient data privacy regulations, embedding a robust data governance framework that anonymises patient data and enforces strict access controls. These safeguards minimise risks and uphold confidentiality during AI learning and feedback cycles. Ethical oversight remains a core component, promoting transparency, fairness, and trust in AI-assisted healthcare. By incorporating these principles, the HCP-ITL model foresees AI as a responsibly managed tool, continuously evolving alongside human expertise.

Adaptive AI-driven diagnosis and treatment planning

In the proposed HCP-ITL model, following initial training, the AI system analyses clinical information to provide preliminary diagnoses and treatment plans. However, rather than presenting this as a current capability, the model aspires to establish AI as an adaptive collaborator that evolves through real-time HCP feedback. This forward-looking capability would require further technological validation and development to ensure feasibility.

Each diagnosis and treatment recommendation are treated as an opportunity for the AI to refine its decision-making processes based on clinician input. This bidirectional exchange accelerates initial assessments while fostering consistency and precision in treatment planning. The aspiration is to create a dynamic, clinician-guided partnership where AI evolves to meet the changing demands of patient care, even though achieving this remains a work in progress.

HCP validation, customisation, and real-time learning

In the HCP-ITL model, preliminary diagnoses and treatment plans generated by the AI are rigorously reviewed by HCPs, who confirm their accuracy and make necessary adjustments to meet patient-specific needs. This step ensures that AI-generated insights are validated by human expertise, maintaining a focus on individual patient care.

While the model envisions AI systems learning from HCP feedback in real time, it acknowledges that current technology might not fully support this capability. Instead, the focus is on embedding periodic updates based on HCP input into the AI's learning processes, which serves as a stepping stone towards the ultimate goal of real-time adaptability. This iterative feedback ensures AI aligns closely with human judgement, capturing nuances that static systems might miss.

Person-focused communication and shared decision-,making

The HCP-ITL model aspires to foster person-focused care by emphasising comprehensive discussions between HCPs and patients regarding diagnoses and personalised care plans. Here, AI is imagined as a tool that facilitates clear communication and empowers patients in shared decision-making. Although real-time adaptation of AI systems to patient feedback is an aspirational goal, the current focus is on ensuring AI systems can flexibly support HCPs in incorporating patient values and preferences into care pathways. This collaborative approach aims to embed patient voices into AI-supported decision-making, with future technological advancements expected to further enhance this integration.

Implementation of the customised personalised care plan

Once the personalised care plan is finalised, the HCP-ITL model foresees AI supporting HCPs in implementing care through routine tasks like managing records, tracking progress, and optimising workflows. The framework treats this stage as a collaborative, adaptive process where AI assists in real-time adjustments based on HCP and patient feedback.

While real-time collaboration remains an aspirational goal, the model currently emphasises the importance of flexibility in AI-supported processes to allow for iterative updates. This approach enables HCPs to focus on high-quality patient care while AI manages administrative tasks efficiently, evolving gradually as technology advances.

Enhanced care and predictable outcomes through continuous learning

The HCP-ITL model is designed to enhance healthcare quality and predictability by combining HCP expertise with AI's analytical capabilities. Unlike traditional AI implementations, this model hopes to create a system where updated outcome data and patient responses are used to refine predictive capabilities continually.

Although achieving a fully dynamic feedback loop requires further technological innovation, the current focus is on building iterative learning mechanisms into AI systems. These mechanisms aim to reduce variability in care and improve patient satisfaction, laying the groundwork for more consistent and accurate outcomes in the future.

Real-time quality assurance and outcome monitoring

A distinctive feature of the HCP-ITL model is its goal to implement AI-driven, real-time quality assurance systems that monitor processes, outcomes, and adherence to living guidelines. While the model acknowledges that real-time quality checks may not yet be fully realised, it aspires to move beyond periodic evaluations to create a continuously learning healthcare ecosystem.

Current efforts focus on establishing systems that use cumulative clinical insights from HCPs, patients, and process data to flag deviations and enable timely adjustments. This approach fosters a responsive, resilient healthcare environment while paving the way for real-time quality assurance in the future.

Conclusion

The Health Care Professional-In-The-Loop (HCP-ITL) model offers an envisioned framework for integrating AI into healthcare, emphasising a collaborative approach that safeguards the critical role of healthcare professionals. By proposing dynamic feedback loops and person-focused care, this model aspires to ensure that AI enhances, rather than replaces, human expertise.

In Part 2, we will explore how this envisioned model can be applied across various medical and dental specialties, outlining its potential to adapt AI roles to specific clinical contexts while preserving ethical standards and human oversight. ◆