Lung nodule detection and segmentation on chest CT
Detection and segmentation of pulmonary nodules on standard chest CT studies to power a CE-marked CAD module.
A radiology software vendor was extending its flagship platform with a lung nodule module and required a validation and training set of chest CT studies covering a wide range of acquisition protocols, kernels and clinical presentations.
Detect and segment pulmonary nodules within a defined size range on standardized chest CT acquisitions, with the level of consistency required for a CE-marked clinical decision support product.
Radiologists were briefed on the client's acquisition requirements (slice thickness, kernels, orientation, injection status) and worked case-by-case on the client's tool. Each annotator was dedicated to the project, ensuring stable inter-annotator agreement across manufacturers and vendors.
The client obtained a controlled, protocol-compliant dataset suitable for regulatory-grade evaluation, with clear traceability for each series. Weekly quality reviews kept sensitivity and false-positive rates aligned with clinical targets.