As the buzz around artificial intelligence (AI) in imaging clinical trials begins to settle, the focus is shifting toward its practical deployment. Despite its promise, AI currently ranks as the third most impactful industry trend for 2025—behind anti-obesity medications and real-world evidence—according to respondents in GlobalData’s State of the Biopharmaceutical Industry 2024 Mid-Year Update. The reasons for this apparent disconnect stem from the industry’s historical caution in adopting new technologies and the skepticism of many clinicians. As highlighted in an NHS AI Lab report, this skepticism is partly fuelled by “uncertainty in the appropriate evidence required for AI technologies” which “can have a negative impact on the confidence of those commissioning and using AI technologies.” Robust, unbiased, and transparent validation of AI tools is critical to demonstrating both performance and safety, and ultimately boosting confidence among stakeholders.
Validation encompasses multiple levels, including internal, external, and local validation, prospective clinical studies and ongoing monitoring. Real-world evaluation is a crucial part of this process, yet it remains underexplored: of nearly 300 AI-enabled medical devices cleared by the US FDA, most have been validated retrospectively rather than through the more rigorous prospective randomized controlled trials (RCTs). While RCTs provide the gold standard for reliability, their high financial and time costs pose significant barriers. By investing more resources in real-world testing of AI tools in 2025, the industry can begin to bridge the current disconnect between their potential and practical implementation.
Radiopharmaceuticals may not be a new field - iodine-131 was first approved by the FDA way back in 1951 - but it has experienced significant advancements in recent years with applications in both diagnostic imaging and therapeutics. These compounds combine an imaging isotope, such as technetium-99m, with a targeting molecule, whether a drug or monoclonal antibody, to selectively locate a tissue of interest, such as cancer. This is followed by the use of a therapeutic isotope like iodine-131, which, once internalized, decays and releases energy that damages and destroys cancer cells. This mechanism of action provides a more targeted alternative to traditional radiation therapy and renders radiopharmaceuticals a powerful tool in precision medicine.
The field has seen a surge of activity in the past year around next-generation radiopharmaceuticals that deliver increasingly targeted treatments, particularly in oncology. This innovation has been matched by an influx of funding, with the market value projected to more than double to over $13 billion USD by 2033. The sector has also been marked by significant mergers and acquisitions, with companies like AstraZeneca, Novartis, and Eli Lilly all making strategic investments to expand their portfolios in this growing market. Will 2025 begin to see the fruit of these investments and more radiopharmaceutical drug candidates hitting the clinic?
While oncology is expected to remain the leading therapeutic area in 2025, with 946 clinical trials, the central nervous system (CNS) is running close behind with 686 trials. Both are fields that extensively leverage imaging techniques. In CNS research, imaging is likely to play an increasingly prominent role, not only in endpoints, but even more so during recruitment, stratification, and eligibility phases, as imaging biomarkers progressively make their way into more and more diagnostic criteria.
For example, the central vein sign and paramagnetic rim lesions - two highly specific MRI markers of Multiple Sclerosis (MS) - were recently proposed as additions to the revised 2024 McDonald Criteria. This represented the culmination of years of research into these two markers as well as accompanying advancements in imaging techniques. The expectation is that by improving the specificity of biomarkers, imaging can significantly enhance patient selection for trials, ensuring that treatments are tested in well-defined populations. This is especially important in a disease such as MS which presents with five overlapping phenotypes and a spectrum of clinical and radiographic features. In 2025, we hope to see these changes reflected in MS trial designs, with the diagnostic criteria for other diseases following suit, ultimately improving trial success rates.
Since it was first proposed for NASA’s Apollo program, the concept of the digital twin has evolved. In its simplest terms, a digital twin is a copy of a biological entity. It is a continuously evolving model that updates with real-world data from a person over time. In practical terms, this is conceived in projects such as the EU-funded The Virtual Brain for brain network simulations, or the Digital Twin Brain for intelligence research drawn from group-wise features, but few have yet come so close as to create replicas at the individual level.
Individualized digital twins could completely transform the clinical trial space. There’s the broader application of mimicking diseases to screen potential drugs and streamline the drug development process. But there’s also the possibility of combining multimodal imaging and multi-omics data to predict patient responsiveness prior to and during a trial. With a digital twin, a patient can belong to both treatment and placebo groups, so to speak. Such synthetic comparator arms could substantially cut the costs of patient recruitment and circumvent one key challenge in rare disease research - limited trial data. Several ongoing trials are aimed at the rigorous testing and validation of this evolving technology to inform the development of much-needed regulatory oversight. Will 2025 be the year that digital twins demonstrate their full potential?
Central reading of imaging data according to standardized guidelines is a key step in determining clinical trial subject eligibility and response assessment, and so read design is an important consideration for study coordinators. When deciding on the number and type of readers, coordinators are usually looking to minimize two things: speed and variability. 2023 saw several deep dives into how imaging trials can look to refine review practices and maximize the validity of result data. Iannessi and Beaumont found a high degree of variability in double reads at baseline in lung cancer trials with blinded independent central review. To decrease discrepancies, they suggest tracking each reader’s specific pattern of assessment and subsequent pairing optimization. Herzog et al argued that investigator-assessed review is not necessarily superior to blinded review; both have their respective advantages. The discordance rate between them can be improved from 39% to 12% by a simple training intervention and removal of reviewers with high reader variability. 2025 will see trial organizers further refine and put these recommendations into practice.
The 2024 US election and the plateauing investments since the pandemic have introduced a period of instability in the market, prompting sponsors to seek cost-effective solutions. In such a climate, there is growing interest in revisiting previously shelved drugs and data from past trials that seemingly led to dead ends. Retrospective data analyses, particularly using advanced analytics, are emerging as a powerful tool for re-evaluating clinical and imaging data from these earlier studies. Additionally, emerging technologies such as AI, which may not have been so advanced at the time of study completion, allow for new insights and potential re-interpretation of data that could lead to fresh opportunities in drug development. 2025 may see more sponsors seeking this approach in lieu of costly new trials in an effort to increase efficiency and reduce research expenses in an increasingly tight market.
As each of these trends increase data volume, impose stricter regulations and complicate trials analytics, 2025 and beyond will see an increasing reliance on end-to-end digital solution providers to tie it all together. Indeed, the growing demand for purpose-built electronic solutions is reflected in the rising use of digital tools by clinical trials sites from 57% in 2020 to 81% in 2022 and the growing number of platforms in the industry space. Sponsors are increasingly recognizing the benefits of Software as a Service (SaaS) solutions and all-in-one cloud-based platforms such as QMENTA's Imaging Hub to fulfill all of their imaging needs. These platforms facilitate the running of site-based, decentralized or hybrid trials by offering a centralized environment for site coordination, data harmonization, and secure, efficient image reading and analysis. They simplify the accessibility of metrics such as site data quality and protocol adherence in both previous and ongoing studies, enabling sponsors to collaborate with only the very best sites. 2025 will see platforms striving for innovation in an effort to stand out in an increasingly competitive market.
Find out more about what QMENTA’s Imaging Hub can do for you here.
Further reading: