Machine Learning and Radiogenomics: Lessons Learned and Future Directions.

Area of Science:

• Precision Medicine
• Radiation Oncology
• Radiogenomics
• Machine Learning (ML)

Background:

• Increasing patient data fuels interest in precision medicine for personalized treatment plans.
• Radiation oncology generates vast diagnostic and therapeutic data, making it suitable for predictive ML models.
• Radiogenomics, studying genomic variations' impact on radiation sensitivity, is an emerging precision radiation oncology field.

Purpose of the Study:

• To provide an overview of machine learning (ML) applications in radiogenomics.
• To explore the relationship between radiogenomics, ML, and precision medicine.
• To discuss current ML approaches in genomics and their specific application to radiogenomics.

Main Methods:

• Review of ML applications in genomics, including genome-wide association studies.
• Examination of ML's current and potential roles within radiogenomics.
• Comparison of ML with statistical hypothesis testing for knowledge extraction in radiogenomics.

Main Results:

• Current uniform dose constraints in radiotherapy are suboptimal, leading to insufficient tumor control or normal tissue toxicity.
• Significant genetic contributions to radiation response are not fully understood.
• ML methods can extract complex knowledge from genomic data, offering potential for radiogenomics advancements.

Conclusions:

• ML has the potential to significantly advance radiogenomics and precision radiation oncology.
• Integrating ML into radiogenomics requires careful consideration of its methods and comparison to traditional statistical approaches.
• Further research is needed to fully leverage ML for personalized radiation therapy based on genetic profiles.