Related Experiment Video
Updated: Apr 27, 2026
10:32
Monitoring Dynamic Growth of Retinal Vessels in Oxygen-Induced Retinopathy Mouse Model
Published on: April 2, 2021
3.8K
OBSERVER AND FEATURE ANALYSIS ON DIAGNOSIS OF RETINOPATHY OF PREMATURITY
E Ataer-Cansizoglu1, S You1, J Kalpathy-Cramer2
1Cognitive Systems Laboratory, Northeastern University, Boston, MA.
IEEE International Workshop on Machine Learning for Signal Processing : [Proceedings]. IEEE International Workshop on Machine Learning for Signal Processing
|July 9, 2014
Summary
Retinopathy of prematurity diagnosis is subjective. This study introduces a computational method analyzing expert variability and image features, revealing consistent expert decisions and key diagnostic indicators for childhood blindness.
Area of Science:
- Ophthalmology
- Medical Imaging Analysis
- Computational Biology
Background:
- Retinopathy of prematurity (ROP) is a leading cause of childhood blindness in premature infants.
- Current diagnostic methods for ROP rely on subjective, qualitative assessments by human experts.
- This subjectivity can lead to variability in diagnoses and treatment decisions.
Purpose of the Study:
- To develop and evaluate a computational method for analyzing inter-expert variability in Retinopathy of Prematurity diagnosis.
- To investigate the relationship between specific retinal image features and expert diagnostic labels.
- To identify objective markers that correlate with expert consensus in ROP assessment.
Main Methods:
- Utilized Mutual Information and Kernel Density Estimation to analyze expert decisions on a dataset of 34 retinal images.
- Employed a dataset diagnosed by 22 independent experts to assess decision variability.
- Feature extraction and correlation analysis with diagnostic labels were performed.
Main Results:
- Demonstrated that a subset of experts exhibit consistent decision-making patterns.
- Identified specific image features that are highly correlated with expert diagnoses.
- Quantified the degree of agreement and disagreement among human observers.
Conclusions:
- The proposed computational method can objectively analyze diagnostic variability in Retinopathy of Prematurity.
- Consistent expert subgroups and key image features can be identified, potentially improving diagnostic accuracy.
- This approach may lead to more standardized and reliable ROP diagnosis, reducing childhood blindness.
