Healthcare
Earth Observation
Industry 5.0
Diagnostics
Clinical Trials
Achieving 97% Precision in Diagnosing Lung Cancer at Every Stage
December 23, 2025
Median Technology: An overview
Median Technologies, a leading imaging CRO, wanted to expand its medical image reading capacity to develop and validate AI models for detecting pulmonary nodules in CT scans.
Lung cancer: The #1 cause of cancer-related deaths
This collaboration supported Median’s eyonis™ LCS with clinically precise data for lung cancer screening.
Project Goals
- Improve the accuracy of pulmonary nodule detection
- Scale lung CT reading and annotation capacity
- Deliver consistent, high-quality datasets for AI screening models
Introduction
Median Technologies, a leading imaging CRO, wanted to expand its medical image reading capacity to develop and validate AI models for detecting pulmonary nodules in CT scans.
Main Challenge
This collaboration supported Median’s eyonis™ LCS with clinically precise data for lung cancer screening.
Goals
- Improve the accuracy of pulmonary nodule detection
- Scale lung CT reading and annotation capacity
- Deliver consistent, high-quality datasets for AI screening models
Median achieved clinical-grade precision with eyonis™ LCS for lung cancer diagnosis
Built on IngeData’s workflow and expertise, the AI model delivered outstanding diagnostic precision across all stages of lung cancer.
Project Impact
97.7%
Cancer cases correctly detected and reported
96.8%
Stage I recall
(early-stage detection)
98.2%
Stage II–IV recall
(advanced-stage detection)
Standardising how lung nodules are defined, measured and segmented
Defining Nodules
Internally aligning on what is a pulmonary nodule: a focal, spherical opacity within the lung parenchyma.
0
1
Defining Nodules
Balancing Annotation Effort
Setting a minimum nodule size to balance the number of lesions detected with the time required per case.
0
2
Balancing Annotation Effort
Standdardising Measurement
The team standardised how the average diameter was measured across solid, part-solid and sub-solid nodules.
0
3
Standdardising Measurement
Assessing Density and Calcification
Comparing nodule density with the ribs and measuring calcification levels.
0
4
Assessing Density and Calcification
Consistent Segmentation
Define clear rules for part-solid and sub-solid nodules to ensure consistent labelling across all datasets.
0
5
Consistent Segmentation
We have successfully partnered with Ingedata on multiple projects covering hundreds of CT scans. By setting up a team of radiologists with project managers in the long run, we have achieved very high quality and swift execution of the projects. This setup builds trust and alleviates a lot of the burden during training and ramp up, with constant feedback helping to improve the annotation set up. We have obtained consistent and clean results both in complex protocols and precise segmentations.
Benjamin Renoust
Data and Knowledge Coordinator
Median Technologies
