Radiomic Analysis: Study Design, Statistical Analysis, and Other Bias Mitigation Strategies

Area of Science:

• Medical Imaging Analysis
• Radiology
• Biostatistics

Background:

• Automated feature extraction from medical images has spurred growth in radiomic research.
• Translating radiomic findings into clinical practice is often impeded by study biases.

Purpose of the Study:

• To review common biases, sources of variability, and pitfalls in radiomic research.
• To emphasize considerations for study design and statistical analysis in radiomics.
• To describe approaches for avoiding these research pitfalls.

Main Methods:

• Literature review of statistical, radiologic, and machine learning research.
• Focus on biases in study design, analysis, and reporting.
• Identification of common pitfalls in radiomic studies.

Main Results:

• Radiomic studies are susceptible to various biases affecting translation to clinical practice.
• Study design and statistical analysis are critical areas prone to variability and pitfalls.
• Strategies exist to mitigate these issues and enhance research rigor.

Conclusions:

• Addressing biases and variability in radiomic research is crucial for clinical translation.
• Careful study design and robust statistical analysis are essential for reliable radiomic findings.
• Implementing described approaches can improve the quality and impact of radiomic studies.