Computer-Aided Diagnosis Systems for Lung Cancer: Challenges and Methodologies

Table 5

Studies on volumetric nodule segmentation reported from 2011 to present. See Table 3 for description of captions.

Study	Purpose	Type	Method	Database	Validation and Performance
Farag et al. [182]	Juxtapleural	Deformable model, 3D	Variational LS segmentation with narrow band implementation	ELCAP database: 397 nodules of 50 patients, 115 juxtapleural cases	Success rate: 70% for juxtapleural cases
Zinoveva et al. [183]	General	Discriminative classification	Soft segmentation. CART decision-tree classifier trained with texture and intensity features. VI trimming after processing	LIDC2 dataset: 39 nodules with 3–30 mm in diameter; manual GTs by 4 radiologists	Median soft-overlap: 0.49 and 0.52 with VI trimming
Jirapatnakul et al. [184]	Juxtapleural	Surface analysis	Robust estimation of the pleural surface, surface removal by change detection over the estimated surface	150 solid juxtapleural nodules	Success rate: 98.0% (81.3% by [185])
Diciotti et al.; [186]	Juxtavascular	Shape analysis	Refine an initial rough segmentation based on a local shape analysis on; 3D geodesic distance map representations	ITALUNG dataset: 256 small nodules; LIDC12 datasets: 157 small nodules	Success rate: 84.8% (ITALUNG) and 88.5% (LIDC12) for automatic; 91.0% (ITALUNG) and 91.7% (LIDC12) for interactive mode