Artificial Intelligence: Shaping the Future of Pediatric Cancer Imaging in Clinical Trials – Much Ado about Something

Lori
22.08.25 11:20 AM

Dr. Scott Bradfield is a clinical trial principal investigator (PI), practicing oncologist, and Medical Advisor to Yunu. His unique perspective offers valuable insight into the day-to-day realities of clinical research, patient care, and the evolving role of technology in advancing clinical trials. As both a physician and researcher, Dr. Bradfield brings firsthand knowledge of the challenges and opportunities shaping the future of clinical development.


In pediatric oncology with its limited patient numbers, every data point counts—and medical imaging is foundational to this data. From diagnosis through treatment response, imaging plays a crucial role in decision-making in clinical trials for solid tumors. And, clinical trials are the basis for giving children a better chance at survival. As AI accelerates its integration into clinical workflows, its impact on pediatric cancer imaging is poised to be transformative over the next 10 years. For cancer research professionals, CROs, and sponsors, understanding how AI will enhance imaging precision, streamline trials, and support better outcomes for children is imperative. The long-anticipated future is now, and AI incorporation will only continue to grow exponentially.

1. Enhancing Imaging Precision and Consistency


One of AI’s most anticipated future roles is in enhancing imaging precision and accuracy. Algorithms trained on large datasets promise to detect subtle patterns in MRI, CT, and PET scans—often with greater speed and sensitivity than the human eye can achieve. This aspect remains a future state, as validation and research are still being developed to reach a clinical use level. In pediatric oncology patients, where smaller anatomical structures and normal developmental changes can complicate interpretation, AI will offer a level of consistency that human radiologists may find difficult to maintain over time or across trial sites. Claims to eliminate radiologists’ involvement are exaggerated, but some added value is definitely possible now. For instance, AI-enabled segmentation tools can automatically delineate tumor boundaries with high reproducibility, reducing inter-reader variability. This is especially valuable in multicenter trials where standardizing imaging reads is crucial to maintaining data integrity. Radiologist oversight remains the standard, but enhanced readings are emerging and advancing rapidly. With better consistency, sponsors can achieve more reliable endpoints, shortening timelines to determine therapeutic efficacy.


2. Boosting Trial Efficiency

AI’s impact certainly goes beyond image analysis. It has the power to improve output across the imaging pipeline—from acquisition to interpretation to data entry. Automated quality control tools can flag suboptimal scans in real-time, reducing the need for costly reimaging. AI can also assist with eligibility screening by rapidly reviewing imaging data against protocol criteria, enabling faster patient enrollment and minimizing manual error. When incorporated into currently available high-value software platforms that already streamline radiology trial workflows, AI will dramatically improve timelines.

In pediatric trials—where enrollment is often slower and patients are rarer and more complex—these efficiencies can be the difference in achieving successful trial completion. Faster imaging review and fewer ineligible or protocol-inconsistent reports yield cleaner data, which in turn enables quicker interim analyses, more agile trial designs, and accelerated decision-making for go/no-go endpoints. Already, error reduction from trial imaging workflow platforms has been shown to reduce errors by 30-50%.1


Long Timelines and Protocol Consistency

Despite the optimism, integrating AI into pediatric oncology trials presents challenges. The majority of AI models are trained on adult data, limiting their generalizability to children. Pediatric-specific datasets are markedly smaller or nonexistent for some diagnoses, more heterogeneous, and subject to even stricter privacy protections, making algorithm development and validation more complex.

Ethically, transparency is paramount. Research sponsors must ensure that AI tools used in trials meet regulatory standards for bias, explainability, and performance. The stakes are high. The bar for acceptance of computer error is far lower than human error. For instance, a crash involving a self-driving car receives attention far out of proportion to the thousands that occur every day at the hands of humans.2  Even if the outcomes become statistically far superior, patients want to know who is responsible and won’t accept blame placed on an algorithm or computed hallucination.

Informed consent also takes on new dimensions in trials involving AI, especially when tools are updating with learning mid-study. Ensuring families understand how their child’s imaging data will be used—and possibly reused—is essential for ethical stewardship of sensitive information used to improve an evolving product or future products.

The Road Ahead

AI is not “taking over,” but it’s an increasingly powerful supplement in the pediatric oncology trial toolkit. Its ability to enhance imaging precision, streamline operations, and indirectly support better outcomes for children offers significant promise. For CROs and sponsors, investment in validated, ethically sound AI imaging technologies may soon become a competitive necessity rather than an experimental add-on.

As the field develops, collaboration between data scientists, pediatric oncologists, radiologists, and trial designers will be essential. The future of AI-enhanced pediatric cancer imaging isn’t just smarter-it’s responsive, evolving, yet misunderstood, and for now still needs human oversight. But the future is exciting… and far closer than you think.

References

1 Alsumidale M. Addressing the Accuracy Crisis in Clinical Trial Imaging. Clinical Trial Vanguard. Oct 9, 2024. Online.

2 Walsh B. The life-or-death case for self-driving cars. Vox. May 4, 2025. Online.

Lori