Abstract
The integration of artificial intelligence (AI) and real-world data (RWD) opens up a new paradigm for exploiting radiology data to develop advanced diagnostic and therapeutic support systems. This review explores the advantages and challenges of utilizing vast digital image datasets from routine clinical practice and computational AI capabilities to enhance cancer patient care. Particularly, the application of AI to radiology data has shown promise in developing tools that automate clinical processes, such as tumor detection, while also identifying novel biomarkers in cancer for potential treatment support. Deep learning models, crucial for this transformation, require substantial data, making RWD a valuable resource for accelerating assay development. RWD offer diverse, extensive data reflecting real-world clinical practices, complementing clinical trial data and providing a broader understanding of patient populations and treatment responses. However, challenges such as data access, variability in quality, and processing complexities must be addressed. Standardizing data processing protocols and feature extraction methods is essential to ensure reproducibility and clinical applicability. Moreover, building trust among clinicians, patients, and regulatory bodies is crucial for successful implementation. This review highlights the potential of AI to analyze RWD imaging data and radiology reports, extracting relevant information and enhancing biomarker discovery. To facilitate practical use, we offer tools to address the main challenges associated with utilizing real-world imaging data, such as key aspects of image access and data processing.