Artificial Intelligence in Medical Diagnostics: Applications and Legal Regulation

By Miglė Bernatonytė and supervised by Dr Neringa Gaubienė
Artificial intelligence (AI) is increasingly transforming medical diagnostics by helping healthcare professionals detect diseases more accurately and efficiently. AI systems are now being used across multiple medical fields such as radiology, pathology, cardiology and dermatology, primarily to support clinical decision-making.
At the same time, growing use of AI in healthcare raises important legal questions concerning patient safety and health care professionals’ responsibility and accountability. Under the EU’s Artificial Intelligence Act, most AI systems used for medical diagnostics are classified as high-risk and must comply with strict requirements regarding safety, accuracy, data quality, transparency and human oversight. Despite technological advancements, healthcare professionals remain responsible for critically evaluating AI-generated recommendations and ensuring that clinical decisions are made in the patient’s best interests.
This article also highlights Lithuania’s growing role in medical AI innovation, including development and deployment of AI systems such as ChestLink and Ligence Heart.
This article is written by Miglė Bernatonytė and supervised by Dr. Neringa Gaubienė. Miglė is a fourth-year Law student at Vilnius University with a particular interest in health law, pharmaceutical regulation, and the legal aspects of emerging technologies. Her research interests focus on the intersection of law, medicine, and innovation, particularly the regulation of artificial intelligence in healthcare and pharmaceutical development. Dr. Gaubienė is an Associate Professor at Vilnius University Faculty of Law with expertise in artificial intelligence regulation, digital technologies, and civil law. Her research focuses on the intersection of law, technology, and society, particularly AI governance, AI literacy, and the responsible use of artificial intelligence in justice and healthcare.
__________________________________________________________________________________
Artificial Intelligence in Medical Diagnostics: Applications and Legal Regulation
Imagine a radiologist who has to assess numerous chest X-rays during a single shift. Each image must be examined carefully, potential abnormalities evaluated, and a decision made that will guide the patient’s subsequent diagnostic assessment and treatment. Until recently, this was exclusively the physician’s responsibility. Today, radiologists at healthcare institutions such as Vilnius Šeškinė Polyclinic and Kaunas City Polyclinic, use artificial intelligence (AI) to assist them. Within seconds, AI algorithms analyse medical images, flag potential abnormalities and help radiologists make assessments that guide medical decisions.
In medical diagnosis—the process of identifying a disease, its stage and causes, test results and other clinical data—artificial intelligence achieves particularly strong results. For example, a systematic review of various medical AI applications found that deep learning models in radiology achieved an average sensitivity of 87% and specificity of 89% [1]. Another large-scale meta-analysis comparing radiologists’ performance with and without AI assistance found that AI-assisted specialists increased diagnostic sensitivity from 0.67 to 0.79 and specificity from 0.82 to 0.87 [2]. Artificial intelligence tools can also optimise clinical workflows, for example, by reducing the time required to interpret medical images by an average of 27.2% [3]. These benefits arise from AI’s ability to quickly analyse large volumes of standardised data—such as medical images, laboratory test results and electronic health records—and identify patterns that would be considerably more difficult for humans to recognise.
Nevertheless, using AI in medicine is not a recent phenomenon. Early AI expert systems, such as MYCIN, were applied in medicine in the second half of the twentieth century. Designed to diagnose infectious diseases, MYCIN operated using “if–then” rules and helped physicians select appropriate antibiotic treatments. Advances in machine and deep learning further extended the AI application use in clinical practice.
Legal clarity regarding what constitutes an AI system emerged when the Regulation (EU) 2024/1689 of the European Parliament and of the Council was adopted on 13 June 2024 (the AI Act). This AI Act harmonised the definition of an AI system at the European Union level for the first time. Under the Act, an AI system is:
“a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments” [4].
To ensure that the AI definition is applied consistently in practice, the European Commission published guidelines in 2025 explaining which systems qualify as AI systems and how they are different from conventional software.
Today, AI is prevalent in many major medical diagnostics areas. Its development is particularly rapid in medical image analysis, including radiology, pathology, dermatology, ophthalmology and endoscopy. Equally important areas include the analysis of electrocardiograms (ECGs), electroencephalograms (EEGs), echocardiography and other biomedical signals, the assessment of electronic health records, and laboratory diagnostics. Multimodal models capable of simultaneously analysing medical images, laboratory test results, clinical data and genetic information are also being developed, allowing physicians access to a more comprehensive picture of a patient’s health. However, these systems are not universal: each is designed to perform a specific diagnostic task and must meet strict safety, performance and quality requirements, particularly when making clinical decisions.
Despite their technological progress, AI systems in medicine are not intended to replace physicians. They generally function as clinical decision support systems (CDSS), analysing medical data, providing additional diagnostic or prognostic insights and supporting physician decision-making. Autonomous systems capable of independently performing certain diagnostic tasks remain relatively uncommon, with human oversight remaining an integral part of the clinical process.
Nonetheless, AI systems are moving toward making autonomous decisions in medical technology. For example, the LumineticsCore system, developed for ophthalmology, analyses retinal photographs and detects signs of diabetic retinopathy—one of the most common complications of diabetes and a leading cause of vision loss. LumineticsCore makes an independent diagnostic decision, meaning that a physician is no longer required to interpret the image at this stage. In 2018, it became the first autonomous AI diagnostic system authorised by the U.S. Food and Drug Administration (FDA). This development was significant from both a technological and a legal perspective, as it raised, for the first time, the question of how liability should be allocated when an initial diagnosis is provided by an algorithm rather than a human. In Europe, the UK-developed DERM system established an important regulatory precedent. In 2025, it became the first autonomous AI system authorised under the European Union Medical Devices Regulation (EU) 2017/745 (MDR) to make clinical decisions concerning skin cancer without physician review. The DERM system was assessed against the most stringent requirements applicable to Class III medical devices and received CE marking, meaning it was as good at skin cancer identification as a dermatologist.
ChestLink, one of the most advanced solutions in this field, was developed in Lithuania. Created by the Vilnius-based company Oxipit, which was acquired by the international medical technology company Sectra in 2026, ChestLink became the world’s first CE-certified Class IIb autonomous AI system for identifying normal chest X-rays. ChestLink assesses an X-ray within seconds and, if it detects no pathological abnormalities, produces a report that is ultimately confirmed by a physician. This model of human–AI collaboration leaves responsibility for the final decision with the physician, while AI provides an additional layer of diagnostic safety and operational efficiency. This model, used in Lithuanian healthcare institutions for more than four years, clearly illustrates the AI technology transformation: AI is becoming a reliable assistant in medical diagnostics and Lithuania is increasingly recognised as both a user and a developer of advanced AI technologies.
Clinical decision support systems are still more common in diagnostics than autonomous systems. For example, Paige Prostate Detect helps pathologists identify signs of prostate cancer more accurately in digital histopathology slides and reduces the likelihood of diagnostic errors. In Lithuania, an important milestone in pathology occurred in 2024 when the National Center of Pathology acceded to the European digital pathology network EMPAIA. In 2026, the National Center of Pathology at Vilnius University Hospital Santaros Klinikos began using the Ibex Breast HER2 AI system to assess the HER2 biomarker in breast cancer diagnostics. This is the first AI solution of its kind in the Baltic States to carry CE marking under the In Vitro Diagnostic Medical Devices Regulation (EU) 2017/746 (IVDR). In cardiology, the Lithuanian-developed Ligence Heart system automatically analyzes echocardiographic images, performs standardised cardiac measurements and helps prepare structured clinical reports.
Artificial intelligence is now used across a range of diagnostic fields, from neurology to endoscopy. For example, the auto SCORE EEG analysis system helps physicians identify abnormalities characteristic of epilepsy. The Olympus EVIS X1 endoscopy platform used at the Lithuanian University of Health Sciences Kaunas Hospital incorporates AI-based computer-aided detection and diagnosis (CAD) functions. It analyses colonoscopy images in real time and helps detect early mucosal abnormalities. In all these cases, AI serves as a clinical decision support tool designed to perform clearly defined diagnostic tasks. It helps assess examination results more accurately and consistently, while the final decision remains with the healthcare professional.
The EU AI Act adopts a risk-based regulatory approach. It divides AI systems into four risk categories: unacceptable risk, high risk, limited risk and minimal risk. The greater a system’s potential impact on a person’s health, safety or fundamental rights, the stricter the legal requirements governing its development, placement on the market and use.
AI systems used in medical diagnostics are generally classified as high-risk systems. AI systems most commonly acquire high-risk status because they are medical devices or in vitro diagnostic medical devices, or safety components of such devices, whose conformity assessment under the MDR or IVDR requires the involvement of a notified body. Certain systems listed in Annex III to the AI Act may also qualify as high-risk AI systems. In such cases, AI system providers must ensure not only a high level of diagnostic accuracy, but also demonstrate safety, clinical performance and the quality of the data used. They must implement risk management measures, ensure human oversight and continuously monitor the system throughout its life cycle.
The use of AI in medical diagnostics requires both highly accurate algorithms and a clear legal and organisational framework. Before deploying such a system, healthcare institutions must ensure appropriate human oversight, staff AI literacy, the quality of the data used, monitoring of the system’s operation, and use in accordance with the manufacturer’s intended purpose. The future of medical diagnostics will therefore depend on both technological progress and the ability to align that progress with legal regulation. Consequently, the deployment and use of AI systems cannot be based solely on technological capabilities. Compliance with EU legal requirements concerning safety, quality, transparency, human oversight and accountability must also be ensured.
Despite the rapid development of autonomous systems, ultimate responsibility for clinical decisions remains with healthcare professionals. Using AI systems does not alter physicians’ professional duty to exercise due care and comply with professional standards. Rather, AI becomes an additional decision-making tool whose recommendations must be critically assessed in light of the specific clinical circumstances. Thus, despite the AI system’s rapid development, human oversight is one of the most important legal safeguards in the deployment of AI systems in healthcare.
AI literacy among healthcare professionals is equally important. To ensure the lawful and safe use of high-risk AI systems, healthcare professionals must be able to operate such systems and understand how they work, the limits of their application, potential errors, the importance of data quality, the duty to exercise human oversight, and the legal and ethical implications of their use. These capabilities are necessary to fulfil the obligation under the EU AI Act to ensure a sufficient level of AI literacy (Article 4), properly implement human oversight requirements applicable to high-risk systems (Article 14), and ensure that systems are used in accordance with the provider’s instructions for use and intended purpose (Article 26). In practice, this means that physicians must be able to critically assess AI-generated recommendations, recognise situations in which they may be inaccurate or unreliable, and make independent clinical decisions grounded in professional competence and the patient’s interests.
References
1. AYEYEMI, B. M. et al. (2025). Applications of Artificial Intelligence in Medicine: A Comprehensive Systematic Review and Meta-Analysis. AROC in Pharmaceutical and Biotechnology, 5(3), pp. 1–12. Available at:https://doi.org/10.53858/arocpb05030112.
2. ZHAO, Dylan; PACKER, Thomas; JIE, Xiaobo; SHAHID, Muhammad; OKE, Jason; PLÜDDEMANN, Annette (2025). Diagnostic Accuracy of Artificial Intelligence-Assisted Radiology Assessment of Cancer: A Systematic Review. BJR|Artificial Intelligence, 2(1), ubaf016. Available at: https://doi.org/10.1093/bjrai/ubaf016.
3. CHEN, Mingyang; WANG, Yuting; WANG, Qiankun; SHI, Jingyi; WANG, Huike; YE, Zichen; XUE, Peng; QIAO, Youlin (2024). Impact of Human and Artificial Intelligence Collaboration on Workload Reduction in Medical Image Interpretation. npj Digital Medicine, 7, 349. Available at: https://doi.org/10.1038/s41746-024-01328-w.
4. 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 Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act). OJ L, 2024/1689, 2024.
_____________________________________________________________
Collaborate with us!
As always, we appreciate you taking the time to read our blog post.
If you have news relevant to our global WAI community or expertise in AI and law, we invite you to contribute to the WAI Legal Insights Blog in 2016! To explore this opportunity, please contact WAI editors Silvia A. Carretta - WAI Chief Legal Officer (via LinkedIn or silvia@womeninai.co) or Dina Blikshteyn (dina@womeninai.co).
Silvia A. Carretta and Dina Blikshteyn
- Editors



Comments