A fully automated and three-dimensional (3D) segmentationmethod for the identification of the pulmonaryparenchyma in thorax X-ray computed tomography (CT)datasets is proposed. It is meant to be used as preprocessingstep in the computer-assisted detection (CAD)system for malignant lung nodule detection that is beingdeveloped by the Medical Applications in a Grid InfrastructureConnection (MAGIC-5) Project. In this newapproach the segmentation of the external airways (tracheaand bronchi), is obtained by 3D region growing withwavefront simulation and suitable stop conditions, thusallowing an accurate handling of the hilar region, notoriouslydifficult to be segmented. Particular attention wasalso devoted to checking and solving the problem of theapparent ‘fusion’ between the lungs, caused by partialvolumeeffects, while 3D morphology operations ensurethe accurate inclusion of all the nodules (internal, pleural,and vascular) in the segmented volume. The newalgorithmwas initially developed and tested on a dataset of 130 CTscans fromthe Italung-CT trial, andwas then applied to theANODE09-competition images (55 scans) and to the LIDCdatabase (84 scans), giving very satisfactory results. Inparticular, the lung contour was adequately located in96% of the CT scans, with incorrect segmentation of theexternal airways in the remaining cases. Segmentationmetrics were calculated that quantitatively express theconsistency between automatic and manual segmentations:themean overlap degree of the segmentationmasksis 0.96±0.02, and the mean and the maximum distancebetween the mask borders (averaged on the whole dataset)are 0.74±0.05 and 4.5±1.5, respectively, whichconfirms that the automatic segmentations quite correctlyreproduce the borders traced by the radiologist. Moreover,no tissue containing internal and pleural nodules wasremoved in the segmentation process, so that this methodproved to be fit for the use in the framework of a CADsystem. Finally, in the comparison with a two-dimensionalsegmentation procedure, inter-slice smoothness was calculated,showing that the masks created by the 3Dalgorithm are significantly smoother than those calculatedby the 2D-only procedure.

Automatic Lung Segmentation in CT Images with Accurate Handling of the Hilar Region

I. De Mitri;
2011

Abstract

A fully automated and three-dimensional (3D) segmentationmethod for the identification of the pulmonaryparenchyma in thorax X-ray computed tomography (CT)datasets is proposed. It is meant to be used as preprocessingstep in the computer-assisted detection (CAD)system for malignant lung nodule detection that is beingdeveloped by the Medical Applications in a Grid InfrastructureConnection (MAGIC-5) Project. In this newapproach the segmentation of the external airways (tracheaand bronchi), is obtained by 3D region growing withwavefront simulation and suitable stop conditions, thusallowing an accurate handling of the hilar region, notoriouslydifficult to be segmented. Particular attention wasalso devoted to checking and solving the problem of theapparent ‘fusion’ between the lungs, caused by partialvolumeeffects, while 3D morphology operations ensurethe accurate inclusion of all the nodules (internal, pleural,and vascular) in the segmented volume. The newalgorithmwas initially developed and tested on a dataset of 130 CTscans fromthe Italung-CT trial, andwas then applied to theANODE09-competition images (55 scans) and to the LIDCdatabase (84 scans), giving very satisfactory results. Inparticular, the lung contour was adequately located in96% of the CT scans, with incorrect segmentation of theexternal airways in the remaining cases. Segmentationmetrics were calculated that quantitatively express theconsistency between automatic and manual segmentations:themean overlap degree of the segmentationmasksis 0.96±0.02, and the mean and the maximum distancebetween the mask borders (averaged on the whole dataset)are 0.74±0.05 and 4.5±1.5, respectively, whichconfirms that the automatic segmentations quite correctlyreproduce the borders traced by the radiologist. Moreover,no tissue containing internal and pleural nodules wasremoved in the segmentation process, so that this methodproved to be fit for the use in the framework of a CADsystem. Finally, in the comparison with a two-dimensionalsegmentation procedure, inter-slice smoothness was calculated,showing that the masks created by the 3Dalgorithm are significantly smoother than those calculatedby the 2D-only procedure.
CAD; image segmentation; lung nodules; region growing; grid; 3D imaging; biomedical image analysis
File in questo prodotto:
File Dimensione Formato  
2011_JDigitImaging_24_DeNunzio.pdf

non disponibili

Tipologia: Altro materiale allegato
Licenza: Non pubblico
Dimensione 797.61 kB
Formato Adobe PDF
797.61 kB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12571/2042
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 79
  • ???jsp.display-item.citation.isi??? 51
social impact