Deep Learning for Synthetic Postcontrast T1-Weighted MRI: A Systematic Review With Targeted Meta-Analysis of Brain Tumor Studies

Key Points

• The aim is to identify barriers to clinical translation of deep learning techniques in postcontrast MRI synthesis for brain tumor studies.
• Conducted a systematic review and targeted meta-analysis of existing literature on synthetic postcontrast MRI.
• Evaluated the limitations in reader studies and external validation.
• Identified the necessity of standardized evaluation workflows.
• Limited reader studies hinder the clinical use of deep learning methodologies.
• Lack of external validation raises concerns regarding the reliability of synthesized images.
• Standardized evaluation workflows incorporating various metrics are crucial for successful clinical translation.

Abstract