Skip to Main Content
An official website of the United States government

Variability of Observer Interpretations and Lung-RADS Assignment

Principal Investigator

Name
Devon Bernard

Degrees
B.S.

Institution
Hive Medical

Position Title
CTO

Email
devon@hivemd.ai

About this CDAS Project

Study
NLST (Learn more about this study)

Project ID
NLST-600

Initial CDAS Request Approval
Nov 13, 2019

Title
Variability of Observer Interpretations and Lung-RADS Assignment

Summary
There have been studies that report substantial variability in radiologists’ performance of screening for lung cancer and finding nodules in LDCTs. Our goal is to expand upon this research to provide more in-depth statistical literature and further develop a benchmark of interpretation variability and classification criteria performance. This study will involve labeling cases, a/b testing interpretations in isolated read and comparison read environments, and performing a multivariate statistical analysis of generated labels.

Aims

1. Measure inter-observer variability of radiologists detection of nodule presence, characteristics (size, calcification pattern, lobulation/spiculation), and Lung-RADS categorization.
2. Measure variability of interpretation between isolated reads (single time-series T0/T1/T2) and comparison reads (multi-series/longitudinal).
3. Measure correlations of nodule characteristics with confirmed diagnoses
4. Evaluate the performance of interpreters and classification criteria like Lung-RADS.

Collaborators

Hive Medical