The Lung TIME: annotated lung nodule dataset and nodule detection framework

Abstract

ABSTRACT The Lung Test Images from Motol Environment (Lung TIME) is a new publicly available dataset of thoracic CTscans with manually annotated pulmonary nodules. It is largerthan other publicly available datasets. Pulmonarynodules are lesions in the lungs, which may indicate lung cancer. Their early detection signi“cantly improvessurvival rate of patients. Automatic nodule detecting systems using CT scans are being developed to reducephysicians load and to improve detection quality. Besides presenting our own nodule detection system, in thisarticle, we mainly address the problem of testing and comparison of automatic nodule detection methods. Ourpublicly available 157 CT scan dataset with 394 annotated nodules contains almost every nodule types (pleuraattached, vessel attached, solitary, regular, irregular) with 2-10mm in diameter, except ground glass opacities(GGO). Annotation was done consensually by two experienced radiologists. The data are in DICOM format,annotations are provided in XML format compatible with the Lung Imaging Database Consortium (LIDC). Ourcomputer aided diagnosis system (CAD) is based on mathematical morphology and “ltration with a subsequentclassi“cation step. We use Asymmetric AdaBoost cla ssi“er. The system was tested using TIME, LIDC andANODE09 databases. The performance was evaluated by cross-validation for Lung TIME and LIDC, and usingthe supplied evaluation procedure for ANODE09. The sensitivity at chosen working point was 94.27% with 7.57false positives/slice for TIME and LIDC datasets combined, 94.03% with 5.46 FPs/slice for the Lung TIME,89.62% sensitivity with 12.03 FPs/slice for LIDC, and 78.68% with 4,61 FPs/slice when applied on ANODE09.Keywords: Lungs, nodule detection, CAD develop ment, performance evaluation.