How laboratory medicine will change in the near future: integrating artificial intelligence, automation, and human expertise in the era of Industry 5.0

Introduction

Since its beginning, laboratory medicine has aimed to provide a tool for swift and accurate patient diagnosis. Some decades ago, it was the laboratory specialists (LS) task to guarantee the uninterrupted service, improving intra-laboratory processes and to medically validate analytical results, based solely on the lab-produced data. Today, aiding in the process of selection and interpretation of lab tests has—or at least should have—become the main task. It has to be done in context of the patient’s clinical presentation, medical history, and other relevant data, guiding clinical decision-making and transforming the role of LS from a mainly technical and organizational to a more medical profession. However, while test portfolios expanded, turn-around-times (TAT) shrunk and laboratory services became available 24/7, due to the incorporation of automated instruments and computerized data management, the number of LS did not increase in a similar manner (1-3). Figure 1 shows a timeline of the number of LS and technicians at the University Hospital Salzburg in comparison to the provided test results over the past years, making it clear that only by automating processes, it became possible to handle the ever-increasing workload.

Figure 1 Timeline of the number of laboratory employees at the Department of Laboratory Medicine at the University Hospital Salzburg compared to the number of test results provided from 2014 to 2023. The drop in performed tests in 2020 is attributable to the COVID pandemic. COVID, coronavirus disease.

To keep up with the demand and the renewed task of LS, similar disruptive technologies need to be applied. The convergence of artificial intelligence (AI), laboratory automation, and human expertise in the context of the so-called Industry 5.0, holds the promise of revolutionizing the field of laboratory medicine, enhancing efficiency, accuracy, and patient-centric care.

This review aims to provide a balanced view on the transformative potential of AI in laboratory medicine, including opportunities, addressing challenges and pitfalls, and patient centered diagnostics. In contrast to other reviews on this topic, we aimed for a holistic and integrated approach, discussing technological, ethical, emotional and practical considerations as well as a possible future of AI-assisted diagnostics, including a proposed practical roadmap for successful integration in the context of Industry 5.0.

From the first to the fifth industrial revolution (Figure 2)

Figure 2 From the first industrial revolution to human-AI collaboration in Industry 5.0. AI, artificial intelligence. This figure was designed using Canva Pro subscription graphics (Canva Pty Ltd., Sydney, Australia).

The first industrial revolution, beginning in the late 18^th century, fundamentally transformed manufacturing processes through the introduction of steam power and water. This marked a significant shift in the way products were produced, moving away from traditional manual techniques (4). The transition to the second industrial revolution made mass production possible, through the introduction of electricity, leading to unprecedented economic growth. Switching to the late 20^th century, the third industrial revolution rises with the upcoming of electronics, information technology (IT) and early automation, further optimizing production and productivity (4,5). In particular, the introduction of electronics and digital technology in Industry 3.0 made modern high speed/high quality in laboratory medicine possible by transitioning from manual techniques to automated analyzers, leading to increased efficiency and accuracy in testing (5).

Currently, we are experiencing the evolution of the fourth industrial revolution, also known as Industry 4.0, that is characterized by the integration of digital, biological, and physical technologies, like cyber-physical systems, the Internet of Things (IoT), cloud computing, cybersecurity and AI getting integrated into industrial processes. It can be described as the connection between manufacturing and digital processes. Put in other words, this revolution is not just about automation, but mainly about the connection of intelligent systems that are capable of making autonomous decisions in industrial processes (5). In laboratory medicine, Industry 4.0 made the development of fully automated, networked laboratories with integrated information systems as a disruptive innovation possible. This connectivity allows data exchange, remote monitoring, and advanced analytics, with the result to improve workflow efficiency and patient outcomes (6,7). However, these advancements also present challenges. The vast amount of data generated by clinical laboratories on a daily basis requires analysis and interpretation and may contribute to the so-called information overload (8).

What we can see right now is the unfolding of Industry 5.0, which focusses on a more collaborative and human-centric approach, trying to create a symbiotic relationship between human and machines, aiming to create a more balanced and sustainable industrial ecosystem (4). In this concept, technology does not replace humans, but rather assist them to improve efficiency.

Several innovations such as collaborative robots (cobots), human-machine interfaces or digital twins are driving this vision. Cobots are designed to work alongside humans in a shared workspace (5,9). Unlike traditional industrial robots that only can follow a specific operating procedure, this human-robot collaboration combines the strengths of both, the precision, consistency, and tireless nature of robots with the flexibility, problem-solving skills, and adaptability of humans (5). To avoid having to interact with keyboards and screens, Industry 5.0 focusses on developing intuitive and natural user interfaces, such as gesture or voice controls, to facilitate seamless communication between humans and machines (10). Digital twins, a cornerstone of Industry 5.0, are virtual representations of individuals, designed to reflect the physical person accurately, in order to simulate real world scenarios without harming the “human twin” (5,11).

Applied correctly, Industry 5.0 approaches may lead to (I) increased efficiency, by combining human intuition with machine precision; (II) improved decision-making, based on real-time data and insights provided by AI and machine learning (ML) models; (III) enhanced safety, by cobots taking on dangerous or repetitive tasks, reducing the risk of injury to human workers; (IV) innovation and creativity, by freeing humans from mundane tasks; and (V) sustainable production, focusing on resource efficiency and environmental responsibility.

Future developments within Industry 5.0 are technologies such as quantum computing, nanotechnology, and brain-computer interfaces, bringing humans, machines, and software systems even more closely together (12).

Opportunities for AI in laboratory medicine

AI, as a cornerstone of Industry 4.0, is inevitably entangled with the future of laboratory medicine (13). ML models will provide significant benefits, especially for the extra-analytical phases of the total laboratory process (14). In the preanalytical processes, AI models may be used to detect unsuitable samples or those of minor quality, while in the postanalytical phase patient identification errors may be detected (15,16), based on the patient’s previous results. Additionally, critical value reporting could be automated or implausible results could be highlighted. The most complex but also most rewarding task in the postanalytical phase would be guidance in test result interpretation in a synoptic manner, taking all laboratory and clinical data into account, including suggestions for follow-up testing. Currently, commercial Large Language Models (LLM) such as ChatGPT are incapable of such tasks, mainly due to a lack of context understanding (17), an issue non-existent in diagnostic disciplines which are based solely on image recognition such as radiology or pathology. However, in the near future, LLMs will be released that are specifically trained on medical data, such as Google’s Med-PaLM2 model (18).

In the analytical phase, all results based on image recognition (e.g., hematology, urine sediment, etc.) are the primary target of AI implementation (19). However, studies like the one from Alcazer et al., proposing an extreme gradient boosting model to predict leukaemia subtypes on the basis of routine laboratory parameters, show that also non-image-based ML models already contribute to an improved analytical outcome (20). Additionally, AI will help detecting patterns in large amounts of analytical data (e.g., from the omics disciplines) (21) or aid in detecting assay deterioration, based on quality control by designated reagents or the floating mean of the local patient cohort. Spectra from spectroscopic/-metric, densitometric or similar measurements could be translated into discrete results. Instrument setting optimization in fields like flow cytometry, toxicology or therapeutic drug monitoring by mass spectrometry could benefit from such models or they could be used to improve workflows and productivity, ultimately improving diagnostic accuracy and precision (6,22,23).

Even outside the total laboratory process, AI may contribute to medical diagnosis, by predicting lab test results, based on other patient information. For example, as shown by Mannino et al., the estimation of hemoglobin values from images of the patients’ hands and fingernails seems feasible (24). Likewise, the estimation of potassium levels based on the patients’ electrocardiograms, as demonstrated by Yasin et al. and Corsi et al., appears to be a promising approach (25,26).

AI-powered process improvements by predictive analytics, anticipating workflow bottlenecks, optimizing resource allocation, and enhancing quality control, can lead to faster turnaround times, reduced errors, and more informed clinical decision-making. Integration of AI with laboratory automation enables seamless sample processing, real-time data analysis, and intelligent result interpretation. This convergence empowers LS to focus on the most complex cases, provide deeper clinical insights, and deliver personalized, evidence-based recommendations to healthcare providers, ultimately improving patient outcomes and transforming the laboratory from a number-producing factory to a real medical facility (6).

AI in a broader diagnostic context

In the next years, AI-supported services will become standard in our daily lives, similar to the evolution of the internet. Until now we interact with AI services via interfaces of any kind to harness its capabilities. However, we believe that in the near future these services will become seamlessly integrated into devices without us even knowing that an AI model is running in the background. This convergence of AI and the IoT is called the Artificial Intelligence of Things (AIoT) (27). With AIoT services, not only data collection and processing but also evaluation, interpretation and decision making is possible. The major segments where AIoT is or will be applied are wearables, smart homes, smart cities, and smart industries/hospitals (28). Specifically in healthcare, AIoT will most probably be used in wearables (29), also called the Internet of Medical Things (IoMT) (30), in diagnostic devices (e.g., stethoscope, ultrasound, etc.) (31), for surgical robots or advanced prosthetics (32), in documentation (33), in endless mobile health (mHealth) applications on our smartphone (34) and many more.

AI-agents

So-called AI-Agents will most likely be at the center of dedicated AI services (35). These agents are explicitly instructed and trained for a specific task and are able to integrate with electronic health record (EHR) systems, laboratory information systems (LIS), or any other health application. AI agents take the brainpower of the underlying model and turn it into actions. You may see it as the AI model being the one who knows how to do it and the AI agent being the one who is actually doing it. AI agents can have diverse capabilities beyond just processing natural language, such as making decisions, solving problems, carrying out actions, conversational assistance, automating processes, coding assistance, handling to-do lists, inspirational guidance and many more. Healthcare professionals will interact with these agents via different interfaces such as screens, gestures or their voice. Predictions are that in the near future such AI agents will take over many routine tasks, especially the ones which are repetitive or predictable (36). They will be regarded as co-workers, but without the need of sleep or food and with lower error rates. Even Multi-Agent-Systems of several AI-agents working towards a common goal are anticipated. Some even call AI agentic operations “the next big thing” in AI development (37) especially in combination with the use of model-context-protocol (MCP) servers, which represent a paradigm shift from traditional Application Programming Interfaces (APIs) by enabling direct, contextual communication between AI models and data sources. Unlike standard APIs that require predefined endpoints and structured requests, MCP servers allow AI agents to dynamically choose the right tool for the individual task at hand, thereby interacting with multiple data sources in a more intuitive, human-like manner. Especially in today’s very demanding time of workforce shortage, the use of AI agents will be the backbone of being able to provide better healthcare with the same number of workers. While AI agents will perform all routine tasks, healthcare workers will have more time to spend with the patient, being able to focus on the more demanding and complex cases.

Integrated diagnostics and the role of AI

Currently, medical diagnostics are split into laboratory testing, imaging, pathology, genomics, microbiology and other disciplines, depending on the country and local setting. This approach is historically based and far from the concept of patient-centered personalized medicine. In order to return to a more medical approach, many authors suggest to aim for the so-called integrated diagnostics strategy (3,38). In this strategy, all diagnostic disciplines are merged and clinicians are no longer forced to know exactly what test is needed for which illness or symptom, or how to interpret its result, but would be able to order diagnostics simply by asking medical questions, triggering diagnostic pathways and retrieving an actionable answer rather than numbers.

For this approach a seamless integration of various data sources will be crucial, including laboratory results, imaging, genomics, current symptoms, medication and, most importantly, the clinical history and presentation of the individual patient, a data source that currently is lacking in most medical laboratories. Within such merged data, AI models could then identify patterns, detect anomalies, suggest follow-up diagnostics and provide comprehensive, personalized diagnostic insights to healthcare providers, empowering them to make more informed decisions and deliver tailored treatments (3).

As an example, Berikol et al. trained an ML model on clinical, laboratory, and imaging data of 228 patients, presenting to the emergency department with chest pain and potential acute coronary syndrome, aiming to decide on hospitalization or discharge, resulting in a 99.13% classification success (39).

AI-based diagnostic demand management

An important aspect of integrated diagnostics is the optimization of diagnostic demand, not only to reduce costs in times of budget cuts but primarily to ensure patient safety. The topic of inappropriate laboratory testing has become increasingly relevant over the past decade, most probably because of the ever-increasing availability (40,41). It is estimated that up to 70% of ordered high-throughput laboratory tests are overused, meaning they are potentially inappropriate or of doubtful clinically importance (42). On the other hand, up to 40% of patients experience diagnostic underuse (43). Both of these circumstances may result in serious patient harm, either by unnecessary follow-up diagnostics or treatments or by missed or delayed diagnosis.

There are several strategies to counteract overuse such as gate-keeping strategies like re-testing intervals, educational interventions, harmonization of test panels and request form design and others (1,44). However, there is only one strategy applicable to overcome underuse: laboratory diagnostic pathways (40,45). Such pathways are based on clinical symptoms, a suspected diagnosis or a laboratory finding (e.g., anemia or isolated prolonged activated partial thromboplastin time), triggering a series of tests, follow-up tests and interpretation. Since in most circumstances a clear yes/no answer is not possible in a rule-based decision tree, AI models are mandatory for these approaches, to be able to weigh possibilities rather than having to decide dichotomously.

In summary, AI as the driving force behind diagnostic demand management strategies as a central aspect of the principle of integrated diagnostic may, will or must be the future of laboratory medicine (3).

Challenges in implementing AI in laboratory medicine

Despite all the euphoria about the possibilities of AI applications, we must consider the challenges and prerequisites for their development in order not to jeopardize the safety of our patients with incorrect AI models.

Data quality

The primary challenge in developing effective AI models for laboratory medicine is acquiring high-quality, structured, and labelled data. While data from LIS and EHR can be utilized, this data is often incomplete, inconsistent, and lacks the necessary annotations. Establishing a well-curated dataset, with detailed meta- and peridata is a crucial first step in developing robust AI models for laboratory medicine (46).

Even when structured data is available, as in most LIS, the challenge of interoperability with data from other sources (e.g., the EHR, national registers, other healthcare providers, or the European Health Data Space) may be the next significant hurdle on the path to achieving so-called FAIR (Findable, Accessible, Interoperable, and Reusable) data (47).

Additional challenges, such as data standardization, measurement variability, and external validation, complicate the situation even more. In a review, performed by Agnello et al., the authors found a high heterogeneity in patients’ selection, clinical and laboratory features, type and number of ML algorithms, ML model validation and ML performance evaluation (48). Consequently, discordant results were observed, even when the same online published database was used. Carobene et al. found similar heterogeneities when reviewing ML models for coronavirus disease 2019 (COVID-19) patients, additionally addressing reporting and replicability issues due to incomplete information on analytical procedures as well as poor or lacking external validation of the published ML models (49). As laboratory tests and subsequent criteria for categorization of patients may change rapidly during this infection, ground truth data must be selected by a medical LS and information on data retrieval and interpretation must be stated, accordingly. These data must not only be FAIR, but also fit for purpose, a combination of data relevance, a characteristic only medical laboratory professionals can decide on, and data quality. Therefore, the authors suggest that medical laboratory professionals and data scientists join forces for an improved outcome.

Furthermore, if models are being trained on data that might underrepresent certain social, ethnic, or demographic groups, or reflect historical inequalities, the resulting model will provide misleading assumptions, potentially harmful to the patient. Colacci et al. examined instances of such algorithmic biases in clinical ML models and found that out of the 760 identified studies completing bias evaluation, 75% indeed reported such a bias, mostly regarding race, gender and age (50).

In order to improve the quality of ML models for the use in clinical laboratories, several authors propose recommendations for different stages of ML development such as problem formulation, data collection and preparation including handling missing data, standardized coding, model validation across multiple sites and model selection, model explainability and interpretability, the reproducible workflow and monitoring after deployment (51-53). The authors summarized these recommendations into five key stages: defining the problem, preparing the data, building the model, validating it, and monitoring it continuously. Following this structure not only improves quality but also helps meet the requirements of European in-vitro-diagnostics regulation (IVDR)/medical devices regulation (MDR) regulations.

To assess the diagnostic quality of AI/ML implementations in the laboratory, Lennerz et al. presented a practical approach at the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) 2023 Strategic conference, named the “Diagnostic Quality Model” (54). This model defines diagnostic quality as the sum of quality measures of the diagnostic test [analytical method or assay, e.g., polymerase chain reaction (PCR), imaging algorithm], the diagnostic procedure (internal laboratory processes, e.g., sample preparation, standard operating procedures) and the diagnostic service (external services, e.g., reporting of findings, IT integration services, billing processes) within a healthcare ecosystem (55).

These quality metrics can mostly only be determined by laboratory professionals, reaffirming their crucial part in the study design, data collection and selection and interpretation of performance metrics (56).

Additionally, medical data privacy and security are of utmost importance as these data are far more sensitive than typical consumer data. Ensuring the confidentiality, integrity, and availability of patient health records is crucial, as unauthorized access or misuse of this information could have severe consequences. Data protection measures, such as anonymization, encryption, access controls, and audit trails, must be implemented in this context. However, anonymization may be quite challenging since patient information may be “hidden” within handwritten text or because the patient’s rare condition, in context with the date and location of treatment may be sufficient for identification (57).

Healthcare providers have accumulated vast amounts of data over the past decades, which are now considered as the new gold of the AI era. If commercially available or start-up models are to be implemented or tested, any unauthorized transfer of this data to these companies for training or other purposes must be prohibited. However, this is often not easily achievable, as seen with other LLMs, which may have used publicly available data for training, regardless of licensing restrictions or potentially infringing intellectual property (58,59). Retrospectively tracing the origins of this data for liability purposes is extremely difficult, if not impossible.

Missing the context

Unlike humans, ML models are currently struggling with medical reasoning where contextual data have to be acknowledged. This is particularly true for laboratory diagnostic tests where results may be interpreted differently, depending on the medical and even preanalytical context (17). But even in image recognition models, like in radiology, the context is not being acknowledged for the most part and in instances in which the variability of structures like cells increases, common LLMs are still struggling with the correct classification (60).

Regulatory and ethical considerations

The rapid development of AI model capabilities has outpaced the ability of regulatory bodies to respond. As a result, current regulations such as the MDR (61), the IVDR (62), the Conformité Européenne (CE) marking (63) and others, are insufficient to address the complex considerations required for medical AI applications. For example, the CE marking of medical devices can only be applied for a specific, defined state of development. If changes are made to the device, the company must reapply for the CE marking. In the case of AI models, this is not feasible, because there is no such thing as a defined, unchanging state, as AI models are constantly evolving and being updated, making it impractical to obtain a CE marking in the same way as for traditional medical devices.

As a consequence, the European Union (EU) has issued the AI Act which went into force in August of 2024 (64). This law provides a risk-based classification system for AI applications. Medical AI applications are considered as high-risk and must therefore comply with strict requirements, including human oversight [also known as the human-in-the-loop principle (65)], robustness, accuracy, transparency, and accountability.

One major aspect when integrating AI into medical decision-making are ethical concerns, including the major question who to hold accountable for the ultimate medical decision (66). When referring to one of the World Health Organization’s definitions of ethics in healthcare, namely to avoid harming others, this issue is yet to be dealt with (67).

Another very important issue is the adherence to the General Data Protection Regulation (GDPR) (68), or similar local data protection regulations. As discussed, medical data is highly sensitive, and patients must be informed about the use of their data, including the possibility to opt-out. Overly restrictive data protection regulations, however, can also hinder the development of AI models that could greatly benefit healthcare. Therefore, edge computing, meaning data being processed by the analytical device itself or local networks, rather than cloud services, have become a prerequisite in many healthcare environments.

Investments

The transition to new standards in digital transformation necessitates substantial investments in new technologies and infrastructure. This financial burden can be particularly challenging for small, privately owned laboratories or those in primary care facilities or even in state-funded tertiary care hospitals, especially in countries with a tight healthcare budget. Additionally, the rapid pace of technological advancement requires laboratories to continuously adapt and upgrade their systems, which can be both costly and complex (69). In state-funded healthcare environments the AI development will most probably outpace the time it takes to pass all bureaucratic hurdles.

Skilled workforce

Even if all of the above is available, it still needs personnel, capable of operating and managing these advanced technologies. The shift towards more digitalized and automated processes, requires employees with new skill sets, particularly in areas such as data analytics, AI, and cybersecurity (70,71). Due to their high demand in the near future, these employees’ compensation demands are likely to be matched by larger medical cooperation, than by the resources of the public sector.

Trust

The implementation of AI, especially in processes like prediction and interpretation, requires a high level of trust, as erroneous results or incorrect interpretation thereof can have severe consequences for patient care. A careful validation process is therefore essential to ensure the reliability and robustness of AI-powered analytical tools before their deployment.

So-called black boxes, AI models which do not allow for insights into their decision-making processes, are currently prohibited in healthcare, according to the “Guidance for Industry Part 11” from the US Government (72). Only so-called explainable AI (XAI) models, where it is clear how the model came to a diagnosis or recommendation, may be used in patient care (73). While the rationale behind these regulations is understandable, it is a pity, as especially these multi-layered deep-learning models often have the highest accuracy.

Additionally, as mentioned above, AI still needs human supervision and interpretation (human-in-the-loop) to deal with the subtle, situation-specific factors that are crucial for making medical decisions, not least because many AI models can also be prone to hallucinations (made up, non-existing information) (74).

Innovation shortage by missing errors

A challenge that may be underestimated, is the possibility of younger doctors relying more and more on AI technologies. As errors are crucial for learning and building knowledge and intuition, and for innovations as a result thereof, a reduction of errors may lead to less skilled healthcare workers and less innovation (75). Self-learning AI systems will then be reaffirmed of their suggestions and actions, as they are no longer challenged by skilled medical personnel. This could result in a self-fulfilling prophecy, where the AI’s recommendations are accepted without question, limiting opportunities for growth and improvement.

Overcoming these challenges will be key to unlocking the full potential of AI in revolutionizing laboratory medicine.

Knowledge and attitude towards AI among laboratory professionals

In an attempt to get an overview of the current landscape of knowledge and practical use of AI in European laboratories, The EFLM committee on Digitalization and Artificial Intelligence (C-AI) has issued an according survey (76). Not very surprising, they found significant gaps in necessary digital infrastructure and training. Major barriers included inadequate digital tools (e.g., no permission to install third-party software or provision of programming environments), restricted access to comprehensive data either due to restrictions or fragmentation of data, and a lack of AI-related skills among personnel. However, nearly all 195 participating laboratories expressed their interest in AI training, indicating a demand for educational initiatives.

Another survey conducted by Bellini et al. distributed among members of the Italian Society of Clinical Biochemistry and Clinical Molecular Biology (SIBioC), came to similar results, revealing the lack of hardware, software and corporate Wi-Fi, as well as limited access to local health data from sources other than laboratories as the major reasons for AI illiteracy (77).

Adler et al. showed that 80% of young European laboratory professionals did not learn digital skills in their academic education but 96% felt they needed to (78).

In 2021 Paranjape et al. conducted a survey among stakeholders in laboratory medicine. Most of which had an unsure attitude towards the future of AI and its area of application within the field of laboratory medicine (79). As obstacles on the way to AI implementation, investment costs, lack of proven clinical benefits, number of decision makers, and privacy concerns were identified.

Limitations

The field of AI is moving so rapidly that any attempt of a written summarization or overview has to fail in terms of completeness or reflection of the current status. Therefore, some of the information presented here may be superseded by new developments. In addition, the number of studies, reviews and articles published increases exponentially on a day-to-day basis, which is why we may have missed some references some readers may regard as essential for a review like this one.

Current AI applications in the laboratory

In contrast to disciplines such as radiology or pathology, which focus mainly on image recognition, the interpretation of laboratory results depends on several contextual variables, such as the patient’s medication, medical history, current symptoms, physical examination, anamnesis, pre-analytical conditions and many others. While for image-base disciplines, the AI models are already very advanced, the level of complexity of lab test interpretation is currently not achievable for most AI models (17).

Therefore, the vast majority of current Food and Drug Administration (FDA)-approved medical AI applications are to be found in the field of radiology, while only few models focusing on laboratory medicine are currently approved (n=9 at the time of writing this manuscript). These AI applications are mostly based on image recognition, either by evaluating cells or the colors of test strips (80). Nevertheless, there are FDA-approved models heavily relying on laboratory results which are not listed under the laboratory medicine category, like models predicting sepsis based on a combination of basic laboratory parameters (81).

Despite the low number of currently FDA-approved models, several publications highlight the potential of AI use in laboratory medicine, predicting a data driven patient care (82).

The patient in the mix

In contrast to times when the doctor was seen as the highest medical authority and was never to be questioned, today, patients increasingly involve themselves in their healthcare decisions. This patient empowerment might be triggered and/or reinforced by the ubiquitous information possibilities. Laboratories are now aiming to restructure lab reports in order to make them more understandable for patients as they have oftentimes become the primary recipient of these data (83,84). This inclusion of patients into their healthcare journey is also reflected by emerging technologies such as blood self-sampling (85) or interconnected wearables (29). The availability of continuous monitoring through wearable devices, such as smartwatches and fitness trackers, allows the collection of physiological data such as heart rate, blood pressure and activity levels. This real-time data empowers patients to obtain greater control and oversight over their own health, supporting their personal empowerment. Engagement of stakeholders and patients in health research will be critical for AI acceptance (86).

On the other hand, this technology is raising data privacy concerns and the question if the produced data can be correctly analyzed and interpreted from the patient him-/herself. In this context AI could serve as a supporting tool, helping to bridge the shift from a professional-centric approach to one where the patient becomes the primary actor in their own care pathway, as suggested by Al Kuwaiti et al., and Briganti and Le Moine (87,88). However, currently, laboratory information in the hands of lay people, using LLMs for data interpretation, may lead to severe misinterpretations, as current models are incapable of synoptic evaluation of all laboratory data with consideration of all other contextual medical variables (17).

However, upcoming LLMs, such as Med-PaLM and Med-PaLM 2 are trained with a great amount of health-related data, and are designed to assist and support medical professionals in decision-making processes, by generating health-related information in real time (18,89). This can be helpful not only for health care professionals, but also for patients, who can use such LLM’s to get a second opinion on their case (90). Aydin et al. reviewed 201 studies on LLMs in patient education and found the following groups of application (91): generating patient education materials, interpreting medical information, providing lifestyle recommendations, supporting customized medication use, offering perioperative care instructions, and optimizing doctor-patient interaction. They also reported challenges such as readability, accuracy, and potential biases.

In addition, Remote Patient Monitoring (RPM) will allow for sharing data between the patient monitoring device and the healthcare professionals (92). This could potentially avoid human errors, and increase patient safety, allowing doctors to continuously monitor a patient’s vital signs, medication adherence, and other health metrics, potentially identifying issues early and intervening before deterioration. Furthermore, robotics in medicine—such as carebots aiding the elderly or surgical robots—is augmenting human capabilities, thus opening new frontiers in patient care (93,94).

Healthcare professionals as well as patients are becoming increasingly active participants in this technological transformation and are confronted with the question how AI and humans could merge together in a beneficial manner. An interesting approach to this issue comes from China, where AI is being integrated into diagnostic centers since 2018. The “One Minute Clinic” is a medical terminal that offers consultation, testing, prescription, and medication purchase (95,96). Equipped with basic testing tools, users can self-test or assist doctors during consultations. Users can connect with one of the 200 doctors quickly and select medication through the smart screen. The medicine cabinet dispenses medication on a self-service basis. The underlying AI model has been trained with a reported 300 million previous consultations.

Will healthcare professionals be replaced?

In 2021, Kiela et al. showed that AI models have surpassed human performance in several tasks such as handwriting recognition, language understanding, image and speech recognition and reading comprehension (97). Consequently, there are significant concerns that the advancement of AI and automation in healthcare will lead to the displacement of healthcare professionals (98).

So-called “doomers” emerged, experts predicting AI taking over most of human labor or even the end of the human race, much like in some Hollywood movies. In 2016, Geoffrey Hinton, some call him the “Godfather of AI”, predicted that radiologists would no longer be needed in 5 years’ time (99). Today, in 2024, there is even a shortage of radiologists (100). Another “Godfather of AI”, Eric Topol, a US-American cardiologist, predicted the opposite, namely the conditional automation, meaning automated systems that drive and monitor the environment but rely on a human driver for backup (101). Similarly, numerous other studies suggest that AI is more likely to augment and enhance human capabilities, rather than replace them entirely (98,102,103).

AI offers promising advancements that could revolutionize diagnostics and patient care. However, the fear of replacement of human professionals in the laboratory remains. Generally speaking, we believe that tasks that are repetitive and predictable are more likely to be automated, while tasks requiring creativity, empathy, complex decision-making, manual labor or human-human interaction are much harder if not impossible to automate. Therefore, laboratory professionals need to embrace the medical part of their profession, rather than the technical one. The expertise of accurate test selection and result interpretation in the individual patient needs to be emphasized, ultimately making the laboratory more medical. Focusing solely on organizational and analytical tasks will eventually result in the displacement of laboratory professionals in the near future (104).

It is becoming increasingly clear that the real value of AI models is to increase medical efficiency, a fact that may effectively counteract the looming shortage of doctors and nurses. Even if AI surpasses the capabilities of humans, a collaborative approach between the two will be key to unlock the full potential of these advanced technologies and ensure the best patient outcomes and a balance between humans and machines (105). The Gestalt principle, which suggests that the whole is greater than the sum of its parts (106), is essential in understanding how AI and human expertise can complement each other. When Excel was introduced, the fear that kids won’t learn math in school anymore was quite real. Similar fears were voiced when the personal computer was introduced. We believe that most of these fears are based in the wrongful assumption that all of these inventions provide a final product, once applied. If recognized as tools to boost efficiency, these fears might vanish eventually. In this context, Xu and Gao defined terms human-AI teaming (HAT), human-AI joint cognitive systems (HAIJCS) and human-centered AI (HCAI), all of which underlining the added effect of AI and human skills (107).

However, this potential future of an AI-human symbiosis is still a bit away from being reality, as some complex question around ethics and liability need to be addressed first. Additionally, we need to keep in mind that many laboratories are still far away from having an IT-infrastructure allowing for medical AI adaptations (76).

Roadmap for successful AI integration in laboratory medicine

The integration of AI into laboratory medicine offers immense potential to enhance diagnostics, workflows, and improve patient outcomes. However, successful implementation requires a strategic approach that weighs opportunities, challenges and prerequisites.

Strategies & plan

A structured implementation plan involves several key steps: Evaluation of needs and goals: first, be clear about what you want to achieve by implementing AI into your healthcare processes. The implementation of Industry 4.0 technologies in healthcare must be balanced against patient needs and outcomes (54,57). These considerations need to be backed by the local healthcare management and all stakeholders.

Data preparation and management: ensure high-quality, comprehensive datasets (FAIR data) for training or integrating AI models into laboratory medicine. Data accuracy, consistency, interoperability and security are essential. In particular, the preparation and interpretation of big data can be challenging (57).

Education and training: train all involved personnel to understand the inner workings, drawbacks, capabilities and limitations of AI and ML in order for them to work collaboratively with AI systems. Following their survey, the EFLM C-AI plans to provide this kind of educational material on a European level (76).

Refine and deploy existing (commercial) AI models: use the collected data to refine the models you aim to implement and conduct controlled pilot studies to evaluate the performance and impact of its integration, allowing for iterative improvements. Starting with smaller, well-defined use cases before scaling up is a prerequisite for successful AI implementation strategies. These pilots provide an opportunity to refine workflows and gather feedback (14). Developing your own models is far more difficult and is outside the scope of this article.

Continuous monitoring and adaptation: regularly evaluate the performance of the deployed models, making adjustments as needed to maintain optimal outcomes. For these quality assessments a standardized system would be preferable. Currently, single approaches have been proposed (54).

Multidisciplinary collaboration

Involve clinicians, LS, data scientists, the legal team and IT experts in both the design and deployment phases of AI tools. This collaborative approach ensures the model’s clinical relevance and technical robustness.

Change management

To avoid a general rejection of the staff and other stakeholders, it is important to support or simplify existing processes instead of completely redefining them. This gives employees the opportunity to adapt to the changes step by step. Additionally, collecting real-time feedback from users to refine the AI system and address any performance gaps or usability concerns is imperative (14). As in every major disruptive innovation, all principles of change management should be applied (108).

Conclusions

The future of laboratory medicine and healthcare as a whole is about to change substantially. All of the Industry 4.0 as well as the currently emerging Industry 5.0 technologies and innovations will disrupt the way we treat patients in the near future. AI will provide the opportunity to enhance efficiency and accuracy in laboratory medicine and will allow for the implementation of laboratory diagnostic pathways and maybe even for merging all diagnostic disciplines, as proposed by the principle of integrated diagnostics.

When recognized as a tool, improving the efficiency of healthcare and laboratory processes, rather than a system slowly replacing all human professionals, the real value of AI becomes evident, namely the possibility for experts to focus on medical cases requiring creative problem-solving skills, empathy, complex decision-making, manual labor or human-human interaction, while repetitive and bureaucratic tasks will be taken care of by ML models. It is necessary for laboratory professionals to embrace these new technologies and work collaboratively with them in order to yield the highest possible benefit for our patients.

The convergence of AI and human expertise will be the foundation of personalized, patient-centered medicine, hopefully transforming healthcare as a whole from a disease-oriented to a health-preserving system (109).

Acknowledgments

Lex.page was used to rephrase passages of this manuscript, making it more captivating and readable.

Footnote

Peer Review File: Available at https://jlpm.amegroups.org/article/view/10.21037/jlpm-25-6/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jlpm.amegroups.org/article/view/10.21037/jlpm-25-6/coif). J.C. serves as an unpaid editorial board member of Journal of Laboratory and Precision Medicine from August 2024 to July 2026. J.C. is a shareholder of LabMed Alliance Inc. and received payment from Siemens Healthineers and Abbott for lecture honoraria. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Cadamuro J, Ibarz M, Cornes M, et al. Managing inappropriate utilization of laboratory resources. Diagnosis (Berl) 2019;6:5-13. [Crossref] [PubMed]
2. Naugler C, Church DL. Automation and artificial intelligence in the clinical laboratory. Crit Rev Clin Lab Sci 2019;56:98-110. [Crossref] [PubMed]
3. Cadamuro J. Disruption vs. evolution in laboratory medicine. Current challenges and possible strategies, making laboratories and the laboratory specialist profession fit for the future. Clin Chem Lab Med 2023;61:558-66. [Crossref] [PubMed]
4. Adel A. Future of industry 5.0 in society: human-centric solutions, challenges and prospective research areas. J Cloud Comput (Heidelb) 2022;11:40. [Crossref] [PubMed]
5. Raja Santhi A, Muthuswamy P. Industry 5.0 or industry 4.0S? Introduction to industry 4.0 and a peek into the prospective industry 5.0 technologies. Int J Interact Des Manuf 2023;17:947-79. [Crossref] [PubMed]
6. Çubukçu HC, Topcu Dİ, Yenice S. Machine learning-based clinical decision support using laboratory data. Clin Chem Lab Med 2024;62:793-823. [Crossref] [PubMed]
7. Neumaier M. Diagnostics 4.0: the medical laboratory in digital health. Clin Chem Lab Med 2019;57:343-8. [Crossref] [PubMed]
8. Cadamuro J, Hillarp A, Unger A, et al. Presentation and formatting of laboratory results: a narrative review on behalf of the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) Working Group "postanalytical phase" (WG-POST). Crit Rev Clin Lab Sci 2021;58:329-53. [Crossref] [PubMed]
9. Rahman MM, Khatun F, Jahan I, et al. Cobotics: The Evolving Roles and Prospects of Next-Generation Collaborative Robots in Industry 5.0. Journal of Robotics 2024;2024:2918089.
10. Grabowska S, Saniuk S, Gajdzik B. Industry 5.0: improving humanization and sustainability of Industry 4.0. Scientometrics 2022;127:3117-44. [Crossref] [PubMed]
11. Modoni GE, Sacco M. A Human Digital-Twin-Based Framework Driving Human Centricity towards Industry 5.0. Sensors (Basel) 2023;23:6054. [Crossref] [PubMed]
12. Liao K, Yang Z, Tao D, et al. Exploring the Intersection of Brain–Computer Interfaces and Quantum Sensing: A Review of Research Progress and Future Trends. Advanced Quantum Technologies 2023;7:2300185.
13. Cadamuro J. Rise of the Machines: The Inevitable Evolution of Medicine and Medical Laboratories Intertwining with Artificial Intelligence-A Narrative Review. Diagnostics (Basel) 2021;11:1399. [Crossref] [PubMed]
14. Herman DS, Rhoads DD, Schulz WL, et al. Artificial Intelligence and Mapping a New Direction in Laboratory Medicine: A Review. Clin Chem 2021;67:1466-82. [Crossref] [PubMed]
15. Zhou R, Liang YF, Cheng HL, et al. A highly accurate delta check method using deep learning for detection of sample mix-up in the clinical laboratory. Clin Chem Lab Med 2022;60:1984-92. [Crossref] [PubMed]
16. Farrell CJ. Identifying mislabelled samples: Machine learning models exceed human performance. Ann Clin Biochem 2021;58:650-2. [Crossref] [PubMed]
17. Cadamuro J, Cabitza F, Debeljak Z, et al. Potentials and pitfalls of ChatGPT and natural-language artificial intelligence models for the understanding of laboratory medicine test results. An assessment by the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) Working Group on Artificial Intelligence (WG-AI). Clin Chem Lab Med 2023;61:1158-66. [Crossref] [PubMed]
18. Singhal K, Tu T, Gottweis J, et al. Toward expert-level medical question answering with large language models. Nat Med 2025;31:943-50. [Crossref] [PubMed]
19. Acevedo A, Merino A, Alférez S, et al. A dataset of microscopic peripheral blood cell images for development of automatic recognition systems. Data Brief 2020;30:105474. [Crossref] [PubMed]
20. Alcazer V, Le Meur G, Roccon M, et al. Evaluation of a machine-learning model based on laboratory parameters for the prediction of acute leukaemia subtypes: a multicentre model development and validation study in France. Lancet Digit Health 2024;6:e323-33. [Crossref] [PubMed]
21. Gao F, Huang K, Xing Y. Artificial Intelligence in Omics. Genomics Proteomics Bioinformatics 2022;20:811-3. [Crossref] [PubMed]
22. Luo S, Shi Y, Chin LK, et al. Machine-Learning-Assisted Intelligent Imaging Flow Cytometry: A Review. Advanced Intelligent Systems 2021;3:2100073.
23. Zhao CL, Mou HZ, Pan JB, et al. AI-assisted mass spectrometry imaging with in situ image segmentation for subcellular metabolomics analysis. Chem Sci 2024;15:4547-55. [Crossref] [PubMed]
24. Mannino RG, Myers DR, Tyburski EA, et al. Smartphone app for non-invasive detection of anemia using only patient-sourced photos. Nat Commun 2018;9:4924. [Crossref] [PubMed]
25. Yasin OZ, Attia Z, Dillon JJ, et al. Noninvasive blood potassium measurement using signal-processed, single-lead ecg acquired from a handheld smartphone. J Electrocardiol 2017;50:620-5. [Crossref] [PubMed]
26. Corsi C, Cortesi M, Callisesi G, et al. Noninvasive quantification of blood potassium concentration from ECG in hemodialysis patients. Sci Rep 2017;7:42492. [Crossref] [PubMed]
27. Siam SI, Ahn H, Liu L, et al. Artificial Intelligence of Things: A Survey. ACM Trans Sen Netw 2025;21:1-75.
28. Ghosh I. AIoT: When Artificial Intelligence Meets the Internet of Things. 2020. Available online: https://www.visualcapitalist.com/aiot-when-ai-meets-iot-technology/. Accessed 13.12.2024.
29. Chen C, Ding S, Wang J. Digital health for aging populations. Nat Med 2023;29:1623-30. [Crossref] [PubMed]
30. Venkatesh DAN. Reimagining the Future of Healthcare Industry through Internet of Medical Things (IoMT), Artificial Intelligence (AI), Machine Learning (ML), Big Data, Mobile Apps and Advanced Sensors. International Journal of Engineering and Advanced Technology 2019;9:3014-9.
31. Bachtiger P, Petri CF, Scott FE, et al. Point-of-care screening for heart failure with reduced ejection fraction using artificial intelligence during ECG-enabled stethoscope examination in London, UK: a prospective, observational, multicentre study. Lancet Digit Health 2022;4:e117-25. [Crossref] [PubMed]
32. Yip M, Salcudean S, Goldberg K, et al. Artificial intelligence meets medical robotics. Science 2023;381:141-6. [Crossref] [PubMed]
33. Balloch J, Sridharan S, Oldham G, et al. Use of an ambient artificial intelligence tool to improve quality of clinical documentation. Future Healthc J 2024;11:100157. [Crossref] [PubMed]
34. Mantena S, Celi LA, Keshavjee S, et al. Improving community health-care screenings with smartphone-based AI technologies. Lancet Digit Health 2021;3:e280-2. [Crossref] [PubMed]
35. GitHub. What are AI agents? Available online: https://github.com/resources/articles/ai/what-are-ai-agents. Accessed 19.12.2024.
36. Ahmed S. AI Agents in 2025: A Comprehensive Review and Future Outlook. 2025. Available online: https://medium.com/@sahin.samia/current-trends-in-ai-agents-use-cases-and-the-future-ahead-1026c4d753fd. Accessed 10.12.2024.
37. Ghose S. The Next “Next Big Thing”: Agentic AI’s Opportunities and Risks. 2024. Available online: https://scet.berkeley.edu/the-next-next-big-thing-agentic-ais-opportunities-and-risks/. Accessed 18.12.2024.
38. Lippi G, Plebani M. Integrated diagnostics: the future of laboratory medicine? Biochem Med (Zagreb) 2020;30:010501. [Crossref] [PubMed]
39. Berikol GB, Yildiz O, Özcan IT. Diagnosis of Acute Coronary Syndrome with a Support Vector Machine. J Med Syst 2016;40:84. [Crossref] [PubMed]
40. Zhi M, Ding EL, Theisen-Toupal J, et al. The landscape of inappropriate laboratory testing: a 15-year meta-analysis. PLoS One 2013;8:e78962. [Crossref] [PubMed]
41. Mrazek C, Simundic AM, Salinas M, et al. Inappropriate use of laboratory tests: How availability triggers demand - Examples across Europe. Clin Chim Acta 2020;505:100-7. [Crossref] [PubMed]
42. Cadamuro J, Gaksch M, Wiedemann H, et al. Are laboratory tests always needed? Frequency and causes of laboratory overuse in a hospital setting. Clin Biochem 2018;54:85-91. [Crossref] [PubMed]
43. Cadamuro J, Simundic AM, von Meyer A, et al. Diagnostic Workup of Microcytic Anemia: An Evaluation of Underuse or Misuse of Laboratory Testing in a Hospital Setting Using the AlinIQ System. Arch Pathol Lab Med 2023;147:117-24. [Crossref] [PubMed]
44. Mrazek C, Haschke-Becher E, Felder TK, et al. Laboratory Demand Management Strategies-An Overview. Diagnostics (Basel) 2021;11:1141. [Crossref] [PubMed]
45. Hoffmann G, Aufenanger J, Födinger M, et al. Benefits and limitations of laboratory diagnostic pathways. Diagnosis (Berl) 2014;1:269-76. [Crossref] [PubMed]
46. Padoan A, Cadamuro J, Frans G, et al. Data flow in clinical laboratories: could metadata and peridata bridge the gap to new AI-based applications? Clin Chem Lab Med 2024;63:684-691. [Crossref] [PubMed]
47. Blatter TU, Witte H, Nakas CT, et al. Big Data in Laboratory Medicine-FAIR Quality for AI? Diagnostics (Basel) 2022;12:1923. [Crossref] [PubMed]
48. Agnello L, Vidali M, Padoan A, et al. Machine learning algorithms in sepsis. Clin Chim Acta 2024;553:117738. [Crossref] [PubMed]
49. Carobene A, Milella F, Famiglini L, et al. How is test laboratory data used and characterised by machine learning models? A systematic review of diagnostic and prognostic models developed for COVID-19 patients using only laboratory data. Clin Chem Lab Med 2022;60:1887-901. [Crossref] [PubMed]
50. Colacci M, Huang YQ, Postill G, et al. Sociodemographic bias in clinical machine learning models: a scoping review of algorithmic bias instances and mechanisms. J Clin Epidemiol 2025;178:111606. [Crossref] [PubMed]
51. Master SR, Badrick TC, Bietenbeck A, et al. Machine Learning in Laboratory Medicine: Recommendations of the IFCC Working Group. Clin Chem 2023;69:690-8. [Crossref] [PubMed]
52. Haymond S, Master SR. How Can We Ensure Reproducibility and Clinical Translation of Machine Learning Applications in Laboratory Medicine? Clin Chem 2022;68:392-5. [Crossref] [PubMed]
53. Miller HA, Valdes R. Rigorous validation of machine learning in laboratory medicine: guidance toward quality improvement. Crit Rev Clin Lab Sci 2025;62:327-46. [Crossref] [PubMed]
54. Lennerz JK, Salgado R, Kim GE, et al. Diagnostic quality model (DQM): an integrated framework for the assessment of diagnostic quality when using AI/ML. Clin Chem Lab Med 2023;61:544-57. [Crossref] [PubMed]
55. Carobene A, Cabitza F, Bernardini S, et al. Where is laboratory medicine headed in the next decade? Partnership model for efficient integration and adoption of artificial intelligence into medical laboratories. Clin Chem Lab Med 2023;61:1359. [Crossref] [PubMed]
56. Spies NC, Farnsworth CW, Wheeler S, et al. Validating, Implementing, and Monitoring Machine Learning Solutions in the Clinical Laboratory Safely and Effectively. Clin Chem 2024;70:1334-43. [Crossref] [PubMed]
57. Pennestrì F, Banfi G. Artificial intelligence in laboratory medicine: fundamental ethical issues and normative key-points. Clin Chem Lab Med 2022;60:1867-74. [Crossref] [PubMed]
58. OpenAI LP. Comment Regarding Request for Comments on Intellectual Property Protection for Artificial Intelligence Innovation. Docket No. PTO–C–2019–0038. United States Patent and Trademark Office, Washington, DC. 2019. Available online: https://www.uspto.gov/sites/default/files/documents/OpenAI_RFC-84-FR-58141.pdf. Accessed 11.12.2024.
59. Berger V. Ex OpenAI Researcher: How ChatGPT’s Training Violated Copyright Law. Forbes. 2024. Available online: https://www.forbes.com/sites/virginieberger/2024/10/29/ex-openai-researcher-how-chatgpts-training-violated-copyright-law/. Accessed 13.12.2024.
60. Negrini D, Pighi L, Tosi M, et al. Evaluating the accuracy of ChatGPT in classifying normal and abnormal blood cell morphology. Clin Chem Lab Med 2025;63:e143-5. [Crossref] [PubMed]
61. European Union. Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017 on medical devices, amending Directive 2001/83/EC, Regulation (EC) No 178/2002 and Regulation (EC) No 1223/2009 and repealing Council Directives 90/385/EEC and 93/42/EEC. Brussels: Official Journal of the European Union; 2017:1.
62. European Union. Regulation (EU) 2017/746 of the European Parliament and of the Council of 5 April 2017 on in vitro diagnostic medical devices and repealing Directive 98/79/EC and Commission Decision 2010/227/EU. Brussels: Official Journal of the European Union; 2017.
63. European Commission. CE marking – obtaining the certificate, EU requirements. In: Your Europe. 2025. Available online: https://europa.eu/youreurope/business/product-requirements/labels-markings/ce-marking/index_en.htm. Accessed 13.12.2024.
64. European Union. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending various regulations and directives (Artificial Intelligence Act). Brussels: Official Journal of the European Union; 2024.
65. Salloch S, Eriksen A. What Are Humans Doing in the Loop? Co-Reasoning and Practical Judgment When Using Machine Learning-Driven Decision Aids. Am J Bioeth 2024;24:67-78. [Crossref] [PubMed]
66. MacIntyre MR, Cockerill RG, Mirza OF, et al. Ethical considerations for the use of artificial intelligence in medical decision-making capacity assessments. Psychiatry Res 2023;328:115466. [Crossref] [PubMed]
67. World Health Organization. Ethics and governance of artificial intelligence for health: WHO guidance. Geneva: World Health Organization; 2021. Available online: https://iris.who.int/handle/10665/341996. Accessed 11.12.2024.
68. European Union. Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation). Brussels: Official Journal of the European Union; 2016.
69. Gungoren MS. Crossing the chasm: strategies for digital transformation in clinical laboratories. Clin Chem Lab Med 2023;61:570-5. [Crossref] [PubMed]
70. Jha N, Shankar PR, Al-Betar MA, et al. Undergraduate Medical Students' and Interns' Knowledge and Perception of Artificial Intelligence in Medicine. Adv Med Educ Pract 2022;13:927-37. [Crossref] [PubMed]
71. Alowais SA, Alghamdi SS, Alsuhebany N, et al. Revolutionizing healthcare: the role of artificial intelligence in clinical practice. BMC Med Educ 2023;23:689. [Crossref] [PubMed]
72. U.S. Department of Health and Human Services Food and Drug Administration. Guidance for Industry Part 11, Electronic Records; Electronic Signatures — Scope and Application. 2003. Available online: https://www.fda.gov/media/75414/download. Accessed 11.12.2024.
73. van der Velden BHM. Explainable AI: current status and future potential. Eur Radiol 2024;34:1187-9. [Crossref] [PubMed]
74. Athaluri SA, Manthena SV, Kesapragada VSRKM, et al. Exploring the Boundaries of Reality: Investigating the Phenomenon of Artificial Intelligence Hallucination in Scientific Writing Through ChatGPT References. Cureus 2023;15:e37432. [Crossref] [PubMed]
75. Banerjee M, Chiew D, Patel KT, et al. The impact of artificial intelligence on clinical education: perceptions of postgraduate trainee doctors in London (UK) and recommendations for trainers. BMC Med Educ 2021;21:429. [Crossref] [PubMed]
76. Cadamuro J, Carobene A, Cabitza F, et al. A comprehensive survey of artificial intelligence adoption in European laboratory medicine: current utilization and prospects. Clin Chem Lab Med 2024;63:692-703. [Crossref] [PubMed]
77. Bellini C, Padoan A, Carobene A, et al. A survey on Artificial Intelligence and Big Data utilisation in Italian clinical laboratories. Clin Chem Lab Med 2022;60:2017-26. [Crossref] [PubMed]
78. Adler J, Lenski M, Tolios A, et al. Digital competence in laboratory medicine. Journal of Laboratory Medicine 2023;47:143-8.
79. Paranjape K, Schinkel M, Hammer RD, et al. The Value of Artificial Intelligence in Laboratory Medicine. Am J Clin Pathol 2021;155:823-31. [Crossref] [PubMed]
80. U.S. Food and Drug Administration. Artificial Intelligence and Machine Learning (AI/ML)-Enabled Medical Devices. In: FDA. 2024. Available online: https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-aiml-enabled-medical-devices. Accessed 18.12.2024.
81. Bhargava A, López-Espina C, Schmalz L, et al. FDA-Authorized AI/ML Tool for Sepsis Prediction: Development and Validation. NEJM AI 2024. DOI: 10.1056/AIoa2400867.
82. Zhang L, Hu ZD. Clinical applications of machine learning in pre-analytical, analytical and post-analytical phases of laboratory medicine: a narrative review. AME Med J 2022;7:37.
83. Watson ID. Making test results more easily understood by patients. BMJ 2015;350:h1942. [Crossref] [PubMed]
84. Young MJ, Scheinberg E, Bursztajn H. Direct-to-patient laboratory test reporting: balancing access with effective clinical communication. JAMA 2014;312:127-8. [Crossref] [PubMed]
85. Poland DCW, Cobbaert CM. Blood self-sampling devices: innovation, interpretation and implementation in total lab automation. Clin Chem Lab Med 2025;63:3-13. [Crossref] [PubMed]
86. Pellegrini G, Lovati C. Stakeholders' engagement for improved health outcomes: a research brief to design a tool for better communication and participation. Front Public Health 2025;13:1536753. [Crossref] [PubMed]
87. Al Kuwaiti A, Nazer K, Al-Reedy A, et al. A Review of the Role of Artificial Intelligence in Healthcare. J Pers Med 2023;13:951. [Crossref] [PubMed]
88. Briganti G, Le Moine O. Artificial Intelligence in Medicine: Today and Tomorrow. Front Med (Lausanne) 2020;7:27. [Crossref] [PubMed]
89. Singhal K, Azizi S, Tu T, et al. Large language models encode clinical knowledge. Nature 2023;620:172-80. [Crossref] [PubMed]
90. Howell MD, Corrado GS, DeSalvo KB. Three Epochs of Artificial Intelligence in Health Care. JAMA 2024;331:242-4. [Crossref] [PubMed]
91. Aydin S, Karabacak M, Vlachos V, et al. Large language models in patient education: a scoping review of applications in medicine. Front Med (Lausanne) 2024;11:1477898. [Crossref] [PubMed]
92. Crossen SS, Romero CC, Lewis C, et al. Remote glucose monitoring is feasible for patients and providers using a commercially available population health platform. Front Endocrinol (Lausanne) 2023;14:1063290. [Crossref] [PubMed]
93. Teng R, Ding Y, See KC. Use of Robots in Critical Care: Systematic Review. J Med Internet Res 2022;24:e33380. [Crossref] [PubMed]
94. Hamet P, Tremblay J. Artificial intelligence in medicine. Metabolism 2017;69S:S36-40. [Crossref] [PubMed]
95. Ping An Good Doctor. Ping An Good Doctor, Shanghai. 2024. Available online: https://www.pagd.net/allPage/aboutUs/47?lang=EN_US. Accessed 13.12.2024.
96. Ping An Good Doctor. Ping An Good Doctor Launches Commercial Operation of One-minute Clinics and Signs Contracts For Nearly 1,000 Units, Serving More Than 3 Million Users. PR Newswire Asia. 2019. Available online: https://en.prnasia.com/releases/apac/ping-an-good-doctor-launches-commercial-operation-of-one-minute-clinics-and-signs-contracts-for-nearly-1-000-units-serving-more-than-3-million-users-234144.shtml. Accessed 09.12.2024.
97. Kiela D, Bartolo M, Nie Y, et al. Dynabench: Rethinking benchmarking in NLP. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Association for Computational Linguistics; 2021:4110-24. Available online: https://aclanthology.org/2021.naacl-main.324/. Accessed 11.12.4.
98. Fogel AL, Kvedar JC. Artificial intelligence powers digital medicine. NPJ Digit Med 2018;1:5. [Crossref] [PubMed]
99. Byju A. The 'Godfather of AI' Predicted I Wouldn't Have a Job. He Was Wrong. In: The New Republic. 2024. Available online: https://newrepublic.com/article/187203/ai-radiology-geoffrey-hinton-nobel-prediction. Accessed 04.02.2025.
100. Rawson JV, Smetherman D, Rubin E. Short-Term Strategies for Augmenting the National Radiologist Workforce. AJR Am J Roentgenol 2024;222:e2430920. [Crossref] [PubMed]
101. Topol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nat Med 2019;25:44-56. [Crossref] [PubMed]
102. Bohr A, Memarzadeh K. The rise of artificial intelligence in healthcare applications. Artificial Intelligence in Healthcare. London: Elsevier; 2020:25-60.
103. Carriero A, Groenhoff L, Vologina E, et al. Deep Learning in Breast Cancer Imaging: State of the Art and Recent Advancements in Early 2024. Diagnostics (Basel) 2024;14:848. [Crossref] [PubMed]
104. Plebani M, Laposata M, Lippi G. A manifesto for the future of laboratory medicine professionals. Clin Chim Acta 2019;489:49-52. [Crossref] [PubMed]
105. Kalidindi S, Gandhi S. Workforce Crisis in Radiology in the UK and the Strategies to Deal With It: Is Artificial Intelligence the Saviour? Cureus 2023;15:e43866. [Crossref] [PubMed]
106. Rock I, Palmer S. The legacy of Gestalt psychology. Sci Am 1990;263:84-90. [Crossref] [PubMed]
107. Xu W, Gao Z. Applying HCAI in Developing Effective Human-AI Teaming: A Perspective from Human-AI Joint Cognitive Systems. Interactions 2024;31:32-7.
108. Kotter JP. Leading Change. 1 ed. Munich: Franz Vahlen; 2011:20-2.
109. Babatunde AO, Shobanke HA, Akinade AA, et al. Enhancing preventive medicine over curative medicine: Role of telemedicine. Public Health Pract (Oxf) 2021;2:100130. [Crossref] [PubMed]