Automated volume analysis of head and neck lesions on CT scans using 3D level set segmentation

Ethan Street, Lubomir Hadjiiski, Berkman Sahiner, Sachin K Gujar, Mohannad Ibrahim, Suresh K. Mukherji, Heang Ping Chan

Research output: Contribution to journalArticle

Abstract

The authors have developed a semiautomatic system for segmentation of a diverse set of lesions in head and neck CT scans. The system takes as input an approximate bounding box, and uses a multistage level set to perform the final segmentation. A data set consisting of 69 lesions marked on 33 scans from 23 patients was used to evaluate the performance of the system. The contours from automatic segmentation were compared to both 2D and 3D gold standard contours manually drawn by three experienced radiologists. Three performance metric measures were used for the comparison. In addition, a radiologist provided quality ratings on a 1 to 10 scale for all of the automatic segmentations. For this pilot study, the authors observed that the differences between the automatic and gold standard contours were larger than the interobserver differences. However, the system performed comparably to the radiologists, achieving an average area intersection ratio of 85.4% compared to an average of 91.2% between two radiologists. The average absolute area error was 21.1% compared to 10.8%, and the average 2D distance was 1.38 mm compared to 0.84 mm between the radiologists. In addition, the quality rating data showed that, despite the very lax assumptions made on the lesion characteristics in designing the system, the automatic contours approximated many of the lesions very well.

Original languageEnglish (US)
Pages (from-to)4399-4408
Number of pages10
JournalMedical Physics
Volume34
Issue number11
DOIs
StatePublished - 2007
Externally publishedYes

Fingerprint

Neck
Head
Radiologists

Keywords

  • 3D segmentation
  • CT scans
  • Head and neck cancer
  • Level sets

ASJC Scopus subject areas

  • Biophysics

Cite this

Street, E., Hadjiiski, L., Sahiner, B., Gujar, S. K., Ibrahim, M., Mukherji, S. K., & Chan, H. P. (2007). Automated volume analysis of head and neck lesions on CT scans using 3D level set segmentation. Medical Physics, 34(11), 4399-4408. https://doi.org/10.1118/1.2794174

Automated volume analysis of head and neck lesions on CT scans using 3D level set segmentation. / Street, Ethan; Hadjiiski, Lubomir; Sahiner, Berkman; Gujar, Sachin K; Ibrahim, Mohannad; Mukherji, Suresh K.; Chan, Heang Ping.

In: Medical Physics, Vol. 34, No. 11, 2007, p. 4399-4408.

Research output: Contribution to journalArticle

Street, E, Hadjiiski, L, Sahiner, B, Gujar, SK, Ibrahim, M, Mukherji, SK & Chan, HP 2007, 'Automated volume analysis of head and neck lesions on CT scans using 3D level set segmentation', Medical Physics, vol. 34, no. 11, pp. 4399-4408. https://doi.org/10.1118/1.2794174
Street, Ethan ; Hadjiiski, Lubomir ; Sahiner, Berkman ; Gujar, Sachin K ; Ibrahim, Mohannad ; Mukherji, Suresh K. ; Chan, Heang Ping. / Automated volume analysis of head and neck lesions on CT scans using 3D level set segmentation. In: Medical Physics. 2007 ; Vol. 34, No. 11. pp. 4399-4408.
@article{cc3056293472455bba8566d800e72522,
title = "Automated volume analysis of head and neck lesions on CT scans using 3D level set segmentation",
abstract = "The authors have developed a semiautomatic system for segmentation of a diverse set of lesions in head and neck CT scans. The system takes as input an approximate bounding box, and uses a multistage level set to perform the final segmentation. A data set consisting of 69 lesions marked on 33 scans from 23 patients was used to evaluate the performance of the system. The contours from automatic segmentation were compared to both 2D and 3D gold standard contours manually drawn by three experienced radiologists. Three performance metric measures were used for the comparison. In addition, a radiologist provided quality ratings on a 1 to 10 scale for all of the automatic segmentations. For this pilot study, the authors observed that the differences between the automatic and gold standard contours were larger than the interobserver differences. However, the system performed comparably to the radiologists, achieving an average area intersection ratio of 85.4{\%} compared to an average of 91.2{\%} between two radiologists. The average absolute area error was 21.1{\%} compared to 10.8{\%}, and the average 2D distance was 1.38 mm compared to 0.84 mm between the radiologists. In addition, the quality rating data showed that, despite the very lax assumptions made on the lesion characteristics in designing the system, the automatic contours approximated many of the lesions very well.",
keywords = "3D segmentation, CT scans, Head and neck cancer, Level sets",
author = "Ethan Street and Lubomir Hadjiiski and Berkman Sahiner and Gujar, {Sachin K} and Mohannad Ibrahim and Mukherji, {Suresh K.} and Chan, {Heang Ping}",
year = "2007",
doi = "10.1118/1.2794174",
language = "English (US)",
volume = "34",
pages = "4399--4408",
journal = "Medical Physics",
issn = "0094-2405",
publisher = "AAPM - American Association of Physicists in Medicine",
number = "11",

}

TY - JOUR

T1 - Automated volume analysis of head and neck lesions on CT scans using 3D level set segmentation

AU - Street, Ethan

AU - Hadjiiski, Lubomir

AU - Sahiner, Berkman

AU - Gujar, Sachin K

AU - Ibrahim, Mohannad

AU - Mukherji, Suresh K.

AU - Chan, Heang Ping

PY - 2007

Y1 - 2007

N2 - The authors have developed a semiautomatic system for segmentation of a diverse set of lesions in head and neck CT scans. The system takes as input an approximate bounding box, and uses a multistage level set to perform the final segmentation. A data set consisting of 69 lesions marked on 33 scans from 23 patients was used to evaluate the performance of the system. The contours from automatic segmentation were compared to both 2D and 3D gold standard contours manually drawn by three experienced radiologists. Three performance metric measures were used for the comparison. In addition, a radiologist provided quality ratings on a 1 to 10 scale for all of the automatic segmentations. For this pilot study, the authors observed that the differences between the automatic and gold standard contours were larger than the interobserver differences. However, the system performed comparably to the radiologists, achieving an average area intersection ratio of 85.4% compared to an average of 91.2% between two radiologists. The average absolute area error was 21.1% compared to 10.8%, and the average 2D distance was 1.38 mm compared to 0.84 mm between the radiologists. In addition, the quality rating data showed that, despite the very lax assumptions made on the lesion characteristics in designing the system, the automatic contours approximated many of the lesions very well.

AB - The authors have developed a semiautomatic system for segmentation of a diverse set of lesions in head and neck CT scans. The system takes as input an approximate bounding box, and uses a multistage level set to perform the final segmentation. A data set consisting of 69 lesions marked on 33 scans from 23 patients was used to evaluate the performance of the system. The contours from automatic segmentation were compared to both 2D and 3D gold standard contours manually drawn by three experienced radiologists. Three performance metric measures were used for the comparison. In addition, a radiologist provided quality ratings on a 1 to 10 scale for all of the automatic segmentations. For this pilot study, the authors observed that the differences between the automatic and gold standard contours were larger than the interobserver differences. However, the system performed comparably to the radiologists, achieving an average area intersection ratio of 85.4% compared to an average of 91.2% between two radiologists. The average absolute area error was 21.1% compared to 10.8%, and the average 2D distance was 1.38 mm compared to 0.84 mm between the radiologists. In addition, the quality rating data showed that, despite the very lax assumptions made on the lesion characteristics in designing the system, the automatic contours approximated many of the lesions very well.

KW - 3D segmentation

KW - CT scans

KW - Head and neck cancer

KW - Level sets

UR - http://www.scopus.com/inward/record.url?scp=35648999057&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=35648999057&partnerID=8YFLogxK

U2 - 10.1118/1.2794174

DO - 10.1118/1.2794174

M3 - Article

C2 - 18072505

AN - SCOPUS:35648999057

VL - 34

SP - 4399

EP - 4408

JO - Medical Physics

JF - Medical Physics

SN - 0094-2405

IS - 11

ER -