Rectal MRI radiomics inter- and intra-reader reliability: should we worry about that?

Area of Science:

• Radiology
• Oncology
• Medical Imaging Analysis

Background:

• Radiomics is an emerging field in medical imaging.
• Radiomics aims to extract quantitative features from medical images for clinical use.
• Reliability of radiomics in rectal cancer diagnosis and treatment response prediction is crucial for clinical adoption.

Purpose of the Study:

• To review and summarize the existing literature on inter- and intra-reader reliability of radiomics applied to rectal MRI.
• To assess the consistency of radiomics feature extraction by different readers and within the same reader over time.

Main Methods:

• A systematic literature search was conducted on PubMed/Medline for studies published between January 2010 and December 2021.
• Included studies focused on rectal cancer patients receiving neoadjuvant therapy and utilizing rectal MRI-based radiomics.
• Studies reporting intra- and/or inter-reader reliability metrics were selected for this review.

Main Results:

• Thirteen studies were included, with an average of 145 patients per study.
• T2-weighted imaging (T2WI) and diffusion-weighted imaging (DWI) were commonly used sequences.
• Intraclass correlation coefficients (ICC) for intra-reader agreement ranged from 0.73 to 0.93, and for inter-reader agreement from 0.60 to 0.99.
• Higher agreement was observed for features from baseline rectal MRI, using 3D whole-tumor volume (VOI) segmentation and first-order features.

Conclusions:

• Current literature suggests good inter- and intra-reader reliability for radiomics on rectal MRI, particularly for low-order features from baseline scans.
• However, the evidence is based on a limited number of studies, some not specifically designed for reliability assessment.
• More targeted research is necessary to quantitatively validate radiomics reliability before widespread clinical implementation in rectal cancer management.