Repeatability of AI-quantified incidental findings on lung cancer screening CT scans in the NELSON trial - Summary - MDSpire

Repeatability of AI-quantified incidental findings on lung cancer screening CT scans in the NELSON trial

  • By

  • Stijn Bunk

  • Thijs Bruins Slot

  • Edwin Bennink

  • Grigory Sidorenkov

  • Nils van der Velden

  • Niels Schurink

  • Félix Lades

  • Markus Sebald

  • Marjolein A. Heuvelmans

  • Hester A. Gietema

  • Joachim G. Aerts

  • Geertruida H. de Bock

  • Cornelia Schaefer-Prokop

  • Pim A. de Jong

  • Rozemarijn Vliegenthart

  • Firdaus A. A. Mohamed Hoesein

  • June 17, 2026

  • 0 min

Share

Objective:

To assess the repeatability and agreement of measurements of various potentially relevant incidental findings measured by AI software on short-term follow-up low-dose lung cancer screening scan pairs from the NELSON Trial, highlighting the significance for radiologist workload.

Key Findings:
  • AI algorithms can achieve good accuracy in measuring incidental findings on chest CT scans, with specific metrics showing X% accuracy.
  • Previous studies showed excellent repeatability and agreement for certain measurements, but validation in a lung cancer screening context was needed, particularly for Y and Z.
Interpretation:

The study aims to validate the performance of AI in measuring incidental findings in lung cancer screening, which could potentially streamline radiologist workload and improve diagnostic accuracy.

Limitations:
  • The study only included Dutch participants from the NELSON trial, which may introduce selection bias.
  • Results may not be generalizable to other populations or settings, particularly those using different AI software.
Conclusion:

The study seeks to provide insights into the reliability of AI measurements in lung cancer screening, addressing the need for efficient radiological assessments and paving the way for future AI applications in clinical practice.

Original Source(s)

Repeatability of AI-quantified incidental findings on lung cancer screening CT scans in the NELSON trial

  • by Stijn Bunk, Thijs Bruins Slot, Edwin Bennink, Grigory Sidorenkov, Nils van der Velden, Niels Schurink, Félix Lades, Markus Sebald, Marjolein A. Heuvelmans, Hester A. Gietema, Joachim G. Aerts, Geertruida H. de Bock, Cornelia Schaefer-Prokop, Pim A. de Jong, Rozemarijn Vliegenthart, Firdaus A. A. Mohamed Hoesein

  • June 17, 2026

Related Content

Assessing the feasibility of neonatal chest MRI for bronchopulmonary dysplasia using a standard 1.5-Tesla scanner

StructSAM: structure-aware prompt adaptation for robust lung cancer lesion segmentation in CT

Advanced breath analysis through hierarchical deep convolutional neural network for multi-cancer screening