Application of artificial intelligence in paediatric oncology imaging

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Paediatric Oncology

Background:

• Paediatric oncology imaging faces challenges like limited patient cohorts, diverse anatomy, motion artifacts, and radiologist shortages.
• These issues can reduce diagnostic accuracy, lengthen workflows, and increase error risks, necessitating innovative solutions.

Purpose of the Study:

• To explore the transformative potential of artificial intelligence (AI) in paediatric oncology medical imaging.
• To identify how AI can address current challenges and improve diagnostic and workflow efficiency.

Main Methods:

• Review of AI applications across the imaging pipeline, including acquisition, analysis, and reporting.
• Discussion of deep learning, radiomics, natural language processing, and large language models in paediatric imaging.
• Exploration of strategies to overcome limitations like data heterogeneity and regulatory hurdles.

Main Results:

• AI shows potential to improve image quality, correct motion artifacts, harmonize datasets, and accelerate scans.
• AI tools like deep learning and radiomics enable precise tumor segmentation, early detection, and classification.
• Natural language processing can streamline report generation and documentation.

Conclusions:

• AI can significantly enhance paediatric oncology imaging, improving diagnostic accuracy and workflow efficiency.
• Addressing challenges in data, model generalizability, and regulation is crucial for AI implementation.
• AI augmentation of radiologists promises to improve care quality and accessibility for paediatric cancer patients.