Introduction
Medical imaging plays a central role in childhood cancer care, from diagnosis and staging to treatment planning and long-term follow-up. However, pediatric imaging presents several unique challenges. Young patients can find it difficult to stay still, and even small amounts of movement can degrade image quality across imaging examinations. MRI is particularly susceptible to motion artifacts and often involves long acquisition times, so sedation or general anesthesia may be required. CT delivers ionizing radiation, to which children are more sensitive. For a child with cancer who may undergo repeated imaging examinations over the course of treatment, every unnecessary minute of scanning time, sedation, and radiation exposure matters. Access to specialized pediatric imaging expertise may also vary between institutions. These challenges have increased interest in approaches that can make pediatric imaging faster, safer, and more consistent. Artificial intelligence (AI) use is now being explored across the imaging pathway, including image acquisition, reconstruction, detection, segmentation, response assessment, reporting, and, increasingly, applications in the operating room.
Background / Clinical Context
Current Imaging Approaches
MRI is an important imaging modality in pediatric oncology because it provides excellent soft-tissue contrast without ionizing radiation. However, MRI examinations can be lengthy and are sensitive to patient motion, which can be particularly challenging in children. CT offers faster image acquisition and remains valuable for many clinical indications. However, its use involves ionizing radiation, which is an important consideration in pediatric patients who may undergo repeated examinations. Functional and molecular imaging can provide complementary information about tumor biology, although availability, cost, acquisition time, and, for some modalities, radiation exposure may limit their use.
AI is being investigated across pediatric oncologic imaging to address several of these challenges. Deep learning methods may improve image quality, reduce motion artifacts, and accelerate image acquisition. Additional applications include automated lesion detection, tumor segmentation, classification, and radiomics-based characterization. AI may also facilitate integration of imaging findings with clinical and molecular data. This integration may enable more individualized disease characterization, treatment response assessment, and clinical decision-making.
Current Challenges and Limitations
Pediatric cancer datasets are often relatively small and heterogeneous, which can make it difficult for AI models to perform consistently across different hospitals and age groups Furthermore, variation in acquisition protocols, patient size, tumor type, and disease presentation may further limit the generalizability of AI models. In neuro-oncology, treatment response assessment can also be complicated by differences in imaging protocols and variability in manual measurements. During surgery, fluorescence techniques such as indocyanine green (ICG) have limited tumor specificity in many non-hepatic malignancies.
Key Evidence / Recent Advances
AI Across Pediatric Oncology Imaging: De Donno et al.
De Donno G, de Vries ISA, Adriaansen LME, et al. Application of artificial intelligence in paediatric oncology imaging. Pediatr Radiol. Published online April 28, 2026. doi:10.1007/s00247-026-06606-1
De Donno et al. describe applications of AI across the pediatric oncology imaging pathway. These include improving image quality, reducing motion artifacts, harmonizing data across institutions, and accelerating image acquisition. In selected applications, these approaches may also enable reductions in radiation exposure. AI is also being studied for tumor segmentation, lesion detection, and classification. Natural language processing and large language models may further support reporting and the integration of imaging with clinical and molecular information.
Clinical relevance: Depending on the application, these approaches could shorten image acquisition times and improve image consistency while reducing some of the burden associated with repeated pediatric examinations, such as high radiation doses and the need for sedation.
AI for Response Assessment: Kann et al.
Kann BH, Vossough A, Brüningk SC, et al. Artificial Intelligence for Response Assessment in Pediatric Neuro-Oncology (AI-RAPNO), part 1: review of the current state of the art. Lancet Oncol. 2025;26(11):e597-e606. doi:10.1016/S1470-2045(25)00484-X
The AI-RAPNO Part 1 review focuses on the use of AI in pediatric neuro-oncology. Current research includes automated tumor segmentation, treatment response assessment, recurrence prediction, toxicity assessment, and multimodal analysis. These tools may improve the precision and reproducibility of imaging assessment and provide additional information to support treatment decisions. However, the authors highlight important barriers, including data heterogeneity and limited model generalizability, limited external validation, and difficulty integrating AI into routine clinical workflows.
Clinical relevance: AI may support more consistent and quantitative assessment of treatment response, but pediatric-specific validation is needed before these tools can be routinely relied upon.
Advanced Imaging in Pediatric Surgery: Fusco et al.
Fusco JC, Abdelhafeez AH, Krauel L, et al. Imaging adjuvants in pediatric surgical oncology. Pediatr Blood Cancer. 2025;72(suppl 2):e31241. doi:10.1002/pbc.31241
Fusco et al. review imaging technologies that can assist pediatric cancer surgery. These include intraoperative molecular imaging with ICG and newer tumor-targeted agents, radioguided surgery using tracers such as MIBG, dual-labeled agents, and 3D modeling with virtual or augmented reality. These technologies may help surgeons better localize tumors and facilitate more precise resections. Newer targeted agents, including pafolacianine and anti-GD2–IRDye800CW, are being investigated to overcome some of the limitations of ICG.
Clinical relevance: These approaches could extend the role of imaging from diagnosis and planning into real-time surgical guidance and potentially improve intraoperative tumor localization and margin assessment.
Clinical Implications
These technologies have several potential clinical applications. AI-assisted reconstruction, motion correction, and faster acquisition could reduce scan times and may help decrease the need for sedation in selected pediatric patients. AI-based segmentation and response assessment may also reduce measurement variability and provide more consistent follow-up, particularly in pediatric neuro-oncology.
Advanced molecular and radioguided imaging could provide surgeons with additional information during tumor resection. Three-dimensional models and virtual or augmented reality may further support surgical planning and localization.
However, these technologies should be viewed as tools that support rather than replace clinical expertise. Differences in imaging protocols, patient age, tumor type, and institutional practice can affect AI performance. Larger datasets, external validation, and integration into existing clinical workflows will be necessary before widespread adoption.
Key Takeaways
- AI is being studied throughout pediatric oncology imaging, from image acquisition and reconstruction to tumor detection, segmentation, reporting, and treatment response assessment.
- Pediatric-specific AI tools may improve the consistency of neuro-oncology response assessment, but limited datasets and differences between institutions remain important barriers.
- Molecular and radioguided imaging are expanding the role of imaging into the operating room.
- Targeted fluorophores and 3D or virtual/augmented reality technologies may improve tumor localization and surgical planning.
- Larger, more representative datasets, multicenter external validation, and prospective evaluation will be essential before these technologies can be routinely used in clinical practice.
Conclusion
Recent advances in AI and imaging technologies are expanding the role of imaging across pediatric oncology. Rather than being limited to detecting abnormalities, AI may help make examinations faster, improve image quality, automate measurements, and integrate imaging with other clinical information. At the same time, molecular and radioguided techniques are allowing imaging to play a more direct role during surgery.
However, strong technical performance does not necessarily translate into clinical adoption. Pediatric cancers are relatively uncommon, and differences in patient populations and imaging practices can make it difficult to generalize AI models. Future work will need to focus on larger and more representative datasets, standardized imaging,rigorous external validation, and prospective assessment of clinical utility. Ultimately, these technologies are most likely to be valuable when they complement the expertise of pediatric radiologists, oncologists, and surgeons and provide them with better information for clinical decision-making.
References
3 Main Papers discussed:
- De Donno G, de Vries ISA, Adriaansen LME, et al. Application of artificial intelligence in paediatric oncology imaging. Pediatr Radiol. Published online April 28, 2026. doi:10.1007/s00247-026-06606-1.
- Kann BH, Vossough A, Brüningk SC, et al. Artificial intelligence for response assessment in pediatric neuro-oncology (AI-RAPNO), part 1: review of the current state of the art. Lancet Oncol. 2025;26(11):e597-e606. doi:10.1016/S1470-2045(25)00484-X.
- Fusco JC, Abdelhafeez AH, Krauel L, et al. Imaging adjuvants in pediatric surgical oncology. Pediatr Blood Cancer. 2025;72(suppl 2):e31241. doi:10.1002/pbc.31241.
Supplemental Papers:
- Frush DP, Callahan MJ, Coley BD, Nadel HR, Paul Guillerman R. Comparison of the different imaging modalities used to image pediatric oncology patients: A COG diagnostic imaging committee/SPR oncology committee white paper. Pediatr Blood Cancer. 2023;70 Suppl 4(Suppl 4):e30298. doi:10.1002/pbc.30298
- Gallo-Bernal S, Bedoya MA, Gee MS, Jaimes C. Pediatric magnetic resonance imaging: faster is better. Pediatr Radiol. 2023;53(7):1270-1284. doi:10.1007/s00247-022-05529-x
- Singh SB, Sarrami AH, Gatidis S, Varniab ZS, Chaudhari A, Daldrup-Link HE. Applications of Artificial Intelligence for Pediatric Cancer Imaging. AJR Am J Roentgenol. 2024;223(2):e2431076. doi:10.2214/AJR.24.31076
