Application of artificial intelligence in paediatric oncology imaging.
Authors
Affiliations (6)
Affiliations (6)
- Division of Imaging and Oncology, University Medical Center Utrecht, Heidelberglaan 100, 3584CX, Utrecht, the Netherlands. [email protected].
- Princess Máxima Center for Pediatric Oncology, Utrecht, the Netherlands.
- Department of Radiology and Nuclear Medicine, Division of Imaging and Oncology, University Medical Center Utrecht, Utrecht, the Netherlands.
- Department of Radiation Oncology, Division of Imaging and Oncology, University Medical Center Utrecht, Utrecht, the Netherlands.
- Division of Imaging and Oncology, University Medical Center Utrecht, Heidelberglaan 100, 3584CX, Utrecht, the Netherlands.
- Department of Radiology and Nuclear Medicine, University of Amsterdam, Amsterdam UMC, Amsterdam, the Netherlands.
Abstract
Paediatric oncology relies heavily on medical imaging for diagnosis, treatment planning, and longitudinal disease monitoring. Yet the field faces unique challenges, including a limited number of patients, diverse anatomy, motion artefacts, and a global shortage of subspecialised radiologists. These constraints can compromise diagnostic accuracy, prolong workflows, and increase the risk of errors, highlighting a critical need for innovative solutions. Artificial intelligence (AI) has emerged as a transformative tool capable of enhancing the entire imaging pipeline. From acquisition to reporting, AI-driven methods show potential to improve image quality, correct motion artefacts, harmonise multicentre datasets, and accelerate scans while reducing radiation exposure. Deep learning models and radiomics have been shown capable of precise tumour segmentation, early lesion detection, and classification, while integration with clinical and molecular data supports individualised staging, prognosis, and therapeutic decision-making. Beyond analysis, natural language processing and large language models can streamline report generation and clinical documentation, potentially enabling more efficient communication and workflow optimisation. Despite these advances, paediatric applications remain constrained by small, heterogeneous datasets, limited paediatric-specific models, and challenges in generalisability, explainability, and regulatory approval. Strategies such as model generalisation across new datasets, the development of retrainable generic models, privacy-preserving training, and synthetic data generation can help overcome these barriers, thereby improving model robustness and promoting equity in care. By augmenting rather than replacing radiologists, AI holds the potential to transform paediatric oncology imaging, improving diagnostic precision, workflow efficiency, and enhancing access to high-quality care. Continued collaboration between clinicians, data scientists, and regulatory bodies will be essential to realise this promise safely and effectively.