Harnessing data science to advance radiation oncology

Area of Science:

• Medical Physics and Informatics
• Oncology
• Data Science

Background:

• Radiation oncology is a technology-intensive field with potential for data science applications.
• Current data science applications in radiation oncology research are limited, with few successful patient care improvements.
• Data interoperability and semantic standardization are significant barriers to big data adoption.

Purpose of the Study:

• To discuss challenges hindering data science integration in radiation oncology.
• To identify methodological issues in data science and the use of health data registries.
• To evaluate the role of big data and machine learning against traditional research methods.

Main Methods:

• Review of data interoperability issues (semantic and organizational).
• Identification of methodological challenges in data science and population-based registries.
• Analysis of discrepancies between big data findings and randomized clinical trials.
• Discussion of machine learning, artificial intelligence, deep learning, and convolutional neural networks.

Main Results:

• Significant data interoperability and semantic issues impede big data and data science in radiation oncology.
• Big data analyses outcomes can conflict with randomized clinical trials, highlighting the latter's continued importance.
• Machine learning and AI show potential but require further development and validation.

Conclusions:

• Data science is a valuable complement to, not a replacement for, traditional research and randomized clinical trials in radiation oncology.
• Addressing data interoperability and standardization is crucial for future insights.
• Optimism exists for linking data sources to advance patient care through data science.