Related Experiment Video
Updated: Jul 15, 2026
10:26
A 3D Digital Model for the Diagnosis and Treatment of Pulmonary Nodules
Published on: May 19, 2023
Inter-observer variations of digital radiograph pulmonary nodule marking by using computer toolkit
Wei Song1, Ying Xu, Yong-Ming Xie
1Department of Radiology, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing 100730.
Summary
Expert radiologists showed significant variations in marking pulmonary nodules on digital radiographs, even when using computer tools. Nodule size agreement was high, but likelihood assessment varied, highlighting the need for standardized interpretation.
Area of Science:
- Radiology
- Medical Imaging
- Computer-Aided Diagnosis
Background:
- Pulmonary nodules are frequently detected on chest digital radiographs (DR).
- Accurate marking and characterization of these nodules are crucial for patient management.
- Inter-observer variability in nodule detection and assessment can impact diagnostic accuracy.
Purpose of the Study:
- To evaluate inter-observer variations in pulmonary nodule marking on chest DR softcopy reading.
- To assess the consistency of nodule likelihood determination using a computer toolkit.
- To compare independent readings with consensus readings among experienced radiologists.
Main Methods:
- Sixty-one chest posterior-anterior DR images were analyzed by two experienced radiologists independently using the IQQA-Chest toolkit.
- Radiologists marked nodules, indicated likelihood, and manually drew boundaries.
- A consensus reading was performed one week later without referring to previous results.
- Statistical analysis compared independent and consensus readings.
Main Results:
- Independent readings detected 409 (Reader 1) and 401 (Reader 2) nodules.
- Consensus reading identified 352 nodules.
- Only 14.4% of unique nodules were confirmed by both readers and consensus.
- Significant differences were found in nodule likelihood determination between readers and between consensus and one reader (P < 0.01).
- No significant difference in nodule size segmentation was observed.
Conclusions:
- Substantial variations exist in pulmonary nodule marking and likelihood assessment among expert radiologists.
- Nodule size measurements showed high consistency.
- Computer-assisted tools did not eliminate inter-observer variability in qualitative assessments.
- Standardization of interpretation protocols is warranted to improve diagnostic consistency.
