Co-governing lanthanide probes and AI-enabled analysis in biomedical imaging

Lanthanide probes and associated AI should be assessed together within a unified international framework, argue Profs Pingkun Yan and Ge Wang, involved in IEEE projects related to medical imaging, and Prof Mannudeep K. Kalra, radiologist.

DOI: https://doi.org/10.25453/plabs.33072212

Published on July 30th, 2026

Lanthanide-based carriers continue to revolutionize biomedical imaging, offering sharp emission lines, long luminescence lifetimes, new contrast agents, and strong paramagnetism across multiple imaging modalities including positron emission tomography (PET), single-photon emission computed tomography (SPECT), magnetic resonance imaging (MRI), computed tomography (CT), ultrasonography, and optical and near-infrared (NIR) imaging (1). Their applications extend beyond improved resolution and reduced radiation exposure to targeted diagnostics and theranostics. However, their quantitative outputs—including lifetime-based measurements, ratiometric temperature and force readouts, and relaxivity-based estimates of concentration—can challenge human interpretation. Such outputs often require computational support, including artificial intelligence (AI), to convert unfamiliar quantitative data into information that human users can interpret and relate to biological or clinical contexts. asynchronous introduction and implementation of the associated hardware, diagnostic and theranostic agents, and software have fragmented their governance. We argue that, when the resulting measurement depends materially on computational interpretation, next-generation lanthanide-based imaging agents and applications should be co-governed with the algorithms that interpret their outputs. We propose three internationally relevant actions: co-validate the ground truth, qualify the agent and its algorithm as one, and build human-relevant proving grounds that regulators worldwide can trust.

A computational signal and a cautionary precedent 

The Frontiers in Science lead article by Liu et al. shows how lanthanide carriers unify optical, X-ray, and magnetic functions within one materials family (1). Probes operating in the second near-infrared window (NIR-II) track molecules at depth (2), while engineered complexes and protein constructs increase magnetic resonance relaxivity and specificity (3). Increasingly, the diagnostic value lies not in a brighter picture but in a derived quantity, such as a lifetime-encoded temperature, force, or relaxivity-based concentration. Such quantities must be extracted computationally. Liu et al. anticipate this development, highlighting the emerging role of AI-enabled analysis in the use of these agents while noting that most advances remain at the proof-of-concept stage, with unresolved questions about long-term safety (1).

Clinical history shows why coordinated validation matters. Gadolinium, itself a lanthanide, is already used at scale in gadolinium-based contrast agents. After widespread adoption, they were linked to nephrogenic systemic fibrosis in patients with impaired kidney function (4), and later to tissue retention even in people with normal renal function (5), prompting regulatory restrictions roughly a decade later (6). The lesson is that safety must be validated before deployment, not discovered after the market has spoken. As AI has advanced over the past decade and become embedded across clinical imaging, validation of quantitative imaging systems should encompass not only the agent but also the algorithm used to interpret its signal.

Two products, one measurement, and a fractured landscape 

The current landscape splits an integrated measurement system across separate, slow regulatory tracks. A lanthanide agent may be evaluated as a drug or drug–device combination, subject to manufacturing, batch-reproducibility, and endotoxin-control requirements that were not written with nanoprobes in mind (1). Where an algorithm converts the agent’s signal into a clinical value, that algorithm is regulated separately as software. In 2025, the United States Food and Drug Administration (FDA) issued guidance on AI-enabled device software functions, including the use of predetermined change control plans to manage software modifications (7), while in the European Union, such software falls under both the AI Act and medical device legislation (8). The resulting quantity—the computed imaging biomarker—may travel a third route. Biomarker qualification is, by design, “interrelated but distinct” from device approval (9). Only a handful of imaging biomarkers have completed qualification, and the FDA’s Medical Device Development Tools program has qualified roughly 20 tools since 2017 (10). An agent whose readout is inseparable from its algorithm may therefore be difficult to evaluate efficiently through three uncoordinated pathways.

Some of the infrastructure needed to close this gap already exists, but it remains disconnected. The Radiological Society of North America (RSNA) Quantitative Imaging Biomarkers Alliance (QIBA) publishes profiles, phantoms, and conformance procedures that bind a biomarker to a stated level of precision (11). However, to our knowledge, none currently covers emerging lanthanide agents together with the AI used to interpret their signals. New approach methodologies (NAMs), such as organoids and organ-on-chip systems, are also advancing (12), but their regulatory qualification remains fragmented and case specific. A Viewpoint by Xu and Qu accompanying Liu et al. reinforces this broader direction, urging a move beyond brighter probes toward quantitatively reliable, biologically safe, and clinically translatable systems (13).

Recommended actions 

Addressing this fragmented landscape will require coordinated action across standardization, regulatory qualification, and preclinical validation. 

Co-validate the ground truth

Standardization must cover both halves of the integrated measurement system. We recommend extending QIBA-style profiles to next-generation optical and magnetic agents, pairing physical reference materials and phantoms with open reference datasets and reconstruction benchmarks so that the agent and the algorithm that interprets its signal are validated against the same traceable reference standards (11). Metrology institutes, standards bodies, and professional communities, including the Institute of Electrical and Electronics Engineers (IEEE) Engineering in Medicine and Biology Society, are well placed to lead. Without shared reference standards, neither measurements nor algorithms can be compared reliably across sites.

Qualify the pair, not the parts

Regulators should create a coordinated pathway that evaluates an agent together with its interpreting algorithm in a linked submission, drawing on the principles of device qualification, biomarker qualification, and predetermined change control (7,9,10). Long-term biodistribution, retention, and clearance studies should begin early in development and inform regulatory decision-making. Lifecycle surveillance should track not only the agent’s fate in the body but also algorithmic drift over time. A coordinated pathway would make safety-by-design enforceable across the lifecycle of a computational imaging system.

Build human-relevant proving grounds and make them global

Organoids and organ-on-chip systems can support co-validation of an agent and its readout before first-in-human studies by assessing imaging performance, biological responses, and reproducibility with fewer safety concerns than direct human use (12). However, these models should complement, rather than replace, the whole-body studies needed to assess biodistribution, clearance, and long-term retention. To support international acceptance, validation protocols should be advanced through relevant mechanisms of the Organisation for Economic Co-operation and Development (OECD) and the International Council for Harmonisation (ICH). They should align with the FDA’s 2025 roadmap to reduce animal testing (14).

Because lanthanides are geopolitically concentrated critical raw materials whose extraction can carry substantial environmental costs, responsible sourcing should form part of the same governance agenda (15). Harmonized standards and pooled proving grounds could reduce duplicated effort, cost, and fragmented validation, helping extend benefits beyond high-resource settings. Supplier scarcity will also require parallel action on responsible sourcing and supply-chain resilience.

Challenges 

Although harmonized governance of lanthanide applications is compelling, asynchronous development across highly specialized pharmaceutical, technical, and clinical domains—and across imaging modalities—makes a single co-governance structure difficult. Governance will also need to accommodate changing software versions, multiple algorithms for a single agent, and updates to acquisition hardware and protocols. Cross-domain representation across coordinated governance bodies could help ensure that resulting systems remain clinically relevant, usable, understandable, generalizable, and safe over the long term. Regulatory involvement is essential to ensure compliance and safety, but oversight should be proportionate and adaptive so that it does not unnecessarily slow scientific progress or delay benefits for patients.

Conclusion 

Algorithmic aids are increasingly indispensable in making information from next-generation imaging agents interpretable by humans. Governing the agent and its algorithm as an integrated system—through shared, traceable reference standards, a coordinated qualification pathway, and globally credible proving grounds—would let lanthanide imaging deliver quantitative, personalized diagnostics without repeating the gadolinium’s decade-late safety reckoning. The science is advancing rapidly. The task for policy is to govern the signal and its interpretation together.

Copyright statement 

Copyright: © 2026 [Yan, Wang, Kalra]. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in Frontiers Policy Labs is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.   

Generative AI statement
The authors declared that generative AI (Anthropic Claude and OpenAI ChatGPT) was used to assist in drafting and editing this manuscript. All scientific content, arguments, and references were reviewed, verified, and approved by the authors. 

References 

1. Liu Y, Chen Z, Sui M, Wang H, Bettinelli M, Carlos LD, et al. Lanthanide carriers: biomedical imaging redefined. Front Sci (2026) 4:1856377 doi: 10.3389/fsci.2026.1856377

2. Antaris AL, Chen H, Cheng K, Sun Y, Hong G, Qu C, et al. A small-molecule dye for NIR-II imaging. Nat Mater (2016) 15(2): 235–42. doi: 10.1038/nmat4476

3. Wahsner J, Gale EM, Rodríguez-Rodríguez A, Caravan P. Chemistry of MRI contrast agents: current challenges and new frontiers. Chem Rev (2019) 119(2):957–1057. doi: 10.1021/acs.chemrev.8b00363

4. Grobner T. Gadolinium – a specific trigger for the development of nephrogenic fibrosing dermopathy and nephrogenic systemic fibrosis? Nephrol Dial Transplant (2006) 21(4):1104–08. doi: 10.1093/ndt/gfk062

5. Kanda T, Ishii K, Kawaguchi H, Kitajima K, Takenaka D. High signal intensity in the dentate nucleus and globus pallidus on unenhanced T1-weighted MR images: relationship with increasing cumulative dose of a gadolinium-based contrast material. Radiology (2014) 270(3):834–41. doi: 10.1148/radiol.13131669

6. Sentinel. FDA drug safety communication: FDA warns that gadolinium-based contrast agents (GBCAs) are retained in the body; requires new class warnings [online] (2017). Available at: https://www.sentinelinitiative.org/news-events/fda-safety-communications-labeling-changes/fda-warns-gadolinium-based-contrast-agents

7. United States Food and Drug Administration. Artificial intelligence-enabled device software functions: lifecycle management and marketing submission recommendations [draft guidance]. Silver Spring, MD: FDA (2025). Available at: https://www.fda.gov/media/184856/download

8. European Parliament, Council of the 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 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) (Text with EEA relevance) [document 32024R1689]. European Union (2024). Available at: https://eur-lex.europa.eu/eli/reg/2024/1689/oj

9. Buckler AJ, Bresolin L, Dunnick NR, Sullivan DC. Quantitative imaging test approval and biomarker qualification: interrelated but distinct activities. Radiology (2011) 259(3):875–84. doi: 10.1148/radiol.10100800

10. United States Food and Drug Administration. Medical device development tools (MDDT) [online] (2026). Available at: https://www.fda.gov/medical-devices/medical-device-development-tools-mddt

11. Sullivan DC, Obuchowski NA, Kessler LG, Raunig DL, Gatsonis C, Huang EP, et al. Metrology standards for quantitative imaging biomarkers. Radiology (2015) 277(3):813–25. doi: 10.1148/radiol.2015142202

12. Ingber DE. Human organs-on-chips for disease modelling, drug development and personalized medicine. Nat Rev Genet (2022) 23(8):467–91. doi: 10.1038/s41576-022-00466-9

13. Xu J, Qu J. Lanthanide-based probes: beyond conventional imaging agents. Front Sci (2026) 4:1905638 doi: 10.3389/fsci.2026.1905638

14. United States Food and Drug Administration. Roadmap to reducing animal testing in preclinical safety studies. Silver Spring, MD: FDA (2025). Available at: https://www.fda.gov/files/newsroom/published/roadmap_to_reducing_animal_testing_in_preclinical_safety_studies.pdf

15. Balaram V. Rare earth elements: a review of applications, occurrence, exploration, analysis, recycling, and environmental impact. Geosci Front (2019) 10(4):1285–1303. doi: 10.1016/j.gsf.2018.12.005