Measurement variability of radiologists when measuring brain tumors
Iris van der Loo1, Teresa M Tareco Bucho1, James A Hanley2
1Department of Radiology, The Netherlands Cancer Institute - Antoni van Leeuwenhoek Hospital, Amsterdam, the Netherlands; GROW - Research Institute for Oncology & Reproduction, Maastricht University, Maastricht, the Netherlands.
European Journal of Radiology
|December 10, 2024
Summary
Measurement variability in brain lesion imaging significantly impacts cancer treatment response classification. Eyeballing brain tumors leads to substantial misclassification of progression and response compared to diameter measurements.
Area of Science:
- Oncology
- Radiology
- Medical Imaging Analysis
Background:
- Response evaluation criteria are crucial in oncology clinical trials for assessing treatment efficacy.
- Established criteria for progression and response were defined prior to advanced medical imaging resolutions.
- This study investigates measurement variability in brain lesions and its effect on response classification.
Purpose of the Study:
- To replicate a previous study on measurement variability using contemporary radiological tools.
- To assess the impact of measurement error on the misclassification of progressive disease and treatment response.
- To quantify measurement error in brain lesion assessment using different methodologies.
Main Methods:
- Sixteen radiologists measured brain tumors using visual estimation (eyeballing), diameter measurements, and AI-assisted segmentation.
- Inter- and intraobserver variability analyses were replicated from a prior study.
- A computer simulation model evaluated the effect of measurement error on misclassifying progressive disease.
Main Results:
- Combined intra- and interobserver error ranged from 13.6-22.2% for eyeballing and 6.8-7.2% for diameter measurements.
- Eyeballing led to erroneous progression declaration in 25.5% of cases and response declaration in 12.3% of cases.
- Diameter measurements resulted in erroneous progression declaration in 1.1% and response declaration in 0% of cases.
- Measurement error disproportionately affected classification in cases with fewer lesions.
Conclusions:
- This study establishes a minimum expected measurement error for brain lesions based on real-world data.
- Measurement error significantly impacts response evaluation criteria misclassification, particularly with visual estimation (eyeballing).
- Future research should explore measurement error across diverse tumor types and its longitudinal impact on response classification.
