Texture-based CAD improves diagnosis for low-dose CT colonography

Zhengrong Liang, Harris Cohen, Erica Posniak, Eddie Fiore, Zigang Wang, Bin Li, Joseph Andersen, Donald Harrington

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Computed tomography (CT)-based virtual colonoscopy or CT colonography (CTC) currently utilizes oral contrast solutions to tag the colonic fluid and possibly residual stool for differentiation from the colon wall and polyps. The enhanced image density of the tagged colonic materials causes a significant partial volume (PV) effect into the colon wall as well as the lumen space (filled with air or CO2). The PV effect on the colon wall can "bury" polyps of size as large as 5mm by increasing their image densities to a noticeable level, resulting in false negatives. It can also create false positives when PV effect goes into the lumen space. We have been modeling the PV effect for mixture-based image segmentation and developing text-based computer-aided detection of polyp (CADpolyp) by utilizing the PV mixture-based image segmentation. This work presents some preliminary results of developing and applying texture-based CADpolyp technique to low-dose CTC studies. A total of 114 studies of asymptomatic patients older than 50, who underwent CTC and then optical colonoscopy (OC) on the same day, were selected from a database, which was accumulated in the past decade and contains various bowel preparations and CT scanning protocols. The participating radiologists found ten polyps of greater than 5 mm from a total of 16 OC proved polyps, i.e., a detection sensitivity of 63%. They scored 23 false positives from the database, i.e., a 20% false positive rate. Approximately 70% of the datasets were marked as imperfect bowel cleansing and/or presence of image artifacts. The impact of imperfect bowel cleansing and image artifacts on VC performance is significant. The texture-based CADpolyp detected all the polyps with an average of 2.68 false positives per patient. This indicates that texture-based CADpolyp can improve the CTC performance in the cases of imperfect cleansed bowels and presence of image artifacts.

Original languageEnglish (US)
Title of host publicationProgress in Biomedical Optics and Imaging - Proceedings of SPIE
Volume6916
DOIs
StatePublished - 2008
Externally publishedYes
EventMedical Imaging 2008 - Physiology, Function, and Structure from Medical Images - San Diego, CA, United States
Duration: Feb 17 2008Feb 19 2008

Other

OtherMedical Imaging 2008 - Physiology, Function, and Structure from Medical Images
CountryUnited States
CitySan Diego, CA
Period2/17/082/19/08

Fingerprint

Tomography
Computer aided design
Textures
Image segmentation
Scanning
Fluids
Air

Keywords

  • Colonic material tagging
  • CTC
  • PV image segmentation
  • Texture based CADpolyp
  • Tissue mixture modeling

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Liang, Z., Cohen, H., Posniak, E., Fiore, E., Wang, Z., Li, B., ... Harrington, D. (2008). Texture-based CAD improves diagnosis for low-dose CT colonography. In Progress in Biomedical Optics and Imaging - Proceedings of SPIE (Vol. 6916). [69161N] https://doi.org/10.1117/12.768073

Texture-based CAD improves diagnosis for low-dose CT colonography. / Liang, Zhengrong; Cohen, Harris; Posniak, Erica; Fiore, Eddie; Wang, Zigang; Li, Bin; Andersen, Joseph; Harrington, Donald.

Progress in Biomedical Optics and Imaging - Proceedings of SPIE. Vol. 6916 2008. 69161N.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Liang, Z, Cohen, H, Posniak, E, Fiore, E, Wang, Z, Li, B, Andersen, J & Harrington, D 2008, Texture-based CAD improves diagnosis for low-dose CT colonography. in Progress in Biomedical Optics and Imaging - Proceedings of SPIE. vol. 6916, 69161N, Medical Imaging 2008 - Physiology, Function, and Structure from Medical Images, San Diego, CA, United States, 2/17/08. https://doi.org/10.1117/12.768073
Liang Z, Cohen H, Posniak E, Fiore E, Wang Z, Li B et al. Texture-based CAD improves diagnosis for low-dose CT colonography. In Progress in Biomedical Optics and Imaging - Proceedings of SPIE. Vol. 6916. 2008. 69161N https://doi.org/10.1117/12.768073
Liang, Zhengrong ; Cohen, Harris ; Posniak, Erica ; Fiore, Eddie ; Wang, Zigang ; Li, Bin ; Andersen, Joseph ; Harrington, Donald. / Texture-based CAD improves diagnosis for low-dose CT colonography. Progress in Biomedical Optics and Imaging - Proceedings of SPIE. Vol. 6916 2008.
@inproceedings{5d5f597101d9487998eb04508fedb606,
title = "Texture-based CAD improves diagnosis for low-dose CT colonography",
abstract = "Computed tomography (CT)-based virtual colonoscopy or CT colonography (CTC) currently utilizes oral contrast solutions to tag the colonic fluid and possibly residual stool for differentiation from the colon wall and polyps. The enhanced image density of the tagged colonic materials causes a significant partial volume (PV) effect into the colon wall as well as the lumen space (filled with air or CO2). The PV effect on the colon wall can {"}bury{"} polyps of size as large as 5mm by increasing their image densities to a noticeable level, resulting in false negatives. It can also create false positives when PV effect goes into the lumen space. We have been modeling the PV effect for mixture-based image segmentation and developing text-based computer-aided detection of polyp (CADpolyp) by utilizing the PV mixture-based image segmentation. This work presents some preliminary results of developing and applying texture-based CADpolyp technique to low-dose CTC studies. A total of 114 studies of asymptomatic patients older than 50, who underwent CTC and then optical colonoscopy (OC) on the same day, were selected from a database, which was accumulated in the past decade and contains various bowel preparations and CT scanning protocols. The participating radiologists found ten polyps of greater than 5 mm from a total of 16 OC proved polyps, i.e., a detection sensitivity of 63{\%}. They scored 23 false positives from the database, i.e., a 20{\%} false positive rate. Approximately 70{\%} of the datasets were marked as imperfect bowel cleansing and/or presence of image artifacts. The impact of imperfect bowel cleansing and image artifacts on VC performance is significant. The texture-based CADpolyp detected all the polyps with an average of 2.68 false positives per patient. This indicates that texture-based CADpolyp can improve the CTC performance in the cases of imperfect cleansed bowels and presence of image artifacts.",
keywords = "Colonic material tagging, CTC, PV image segmentation, Texture based CADpolyp, Tissue mixture modeling",
author = "Zhengrong Liang and Harris Cohen and Erica Posniak and Eddie Fiore and Zigang Wang and Bin Li and Joseph Andersen and Donald Harrington",
year = "2008",
doi = "10.1117/12.768073",
language = "English (US)",
isbn = "9780819471000",
volume = "6916",
booktitle = "Progress in Biomedical Optics and Imaging - Proceedings of SPIE",

}

TY - GEN

T1 - Texture-based CAD improves diagnosis for low-dose CT colonography

AU - Liang, Zhengrong

AU - Cohen, Harris

AU - Posniak, Erica

AU - Fiore, Eddie

AU - Wang, Zigang

AU - Li, Bin

AU - Andersen, Joseph

AU - Harrington, Donald

PY - 2008

Y1 - 2008

N2 - Computed tomography (CT)-based virtual colonoscopy or CT colonography (CTC) currently utilizes oral contrast solutions to tag the colonic fluid and possibly residual stool for differentiation from the colon wall and polyps. The enhanced image density of the tagged colonic materials causes a significant partial volume (PV) effect into the colon wall as well as the lumen space (filled with air or CO2). The PV effect on the colon wall can "bury" polyps of size as large as 5mm by increasing their image densities to a noticeable level, resulting in false negatives. It can also create false positives when PV effect goes into the lumen space. We have been modeling the PV effect for mixture-based image segmentation and developing text-based computer-aided detection of polyp (CADpolyp) by utilizing the PV mixture-based image segmentation. This work presents some preliminary results of developing and applying texture-based CADpolyp technique to low-dose CTC studies. A total of 114 studies of asymptomatic patients older than 50, who underwent CTC and then optical colonoscopy (OC) on the same day, were selected from a database, which was accumulated in the past decade and contains various bowel preparations and CT scanning protocols. The participating radiologists found ten polyps of greater than 5 mm from a total of 16 OC proved polyps, i.e., a detection sensitivity of 63%. They scored 23 false positives from the database, i.e., a 20% false positive rate. Approximately 70% of the datasets were marked as imperfect bowel cleansing and/or presence of image artifacts. The impact of imperfect bowel cleansing and image artifacts on VC performance is significant. The texture-based CADpolyp detected all the polyps with an average of 2.68 false positives per patient. This indicates that texture-based CADpolyp can improve the CTC performance in the cases of imperfect cleansed bowels and presence of image artifacts.

AB - Computed tomography (CT)-based virtual colonoscopy or CT colonography (CTC) currently utilizes oral contrast solutions to tag the colonic fluid and possibly residual stool for differentiation from the colon wall and polyps. The enhanced image density of the tagged colonic materials causes a significant partial volume (PV) effect into the colon wall as well as the lumen space (filled with air or CO2). The PV effect on the colon wall can "bury" polyps of size as large as 5mm by increasing their image densities to a noticeable level, resulting in false negatives. It can also create false positives when PV effect goes into the lumen space. We have been modeling the PV effect for mixture-based image segmentation and developing text-based computer-aided detection of polyp (CADpolyp) by utilizing the PV mixture-based image segmentation. This work presents some preliminary results of developing and applying texture-based CADpolyp technique to low-dose CTC studies. A total of 114 studies of asymptomatic patients older than 50, who underwent CTC and then optical colonoscopy (OC) on the same day, were selected from a database, which was accumulated in the past decade and contains various bowel preparations and CT scanning protocols. The participating radiologists found ten polyps of greater than 5 mm from a total of 16 OC proved polyps, i.e., a detection sensitivity of 63%. They scored 23 false positives from the database, i.e., a 20% false positive rate. Approximately 70% of the datasets were marked as imperfect bowel cleansing and/or presence of image artifacts. The impact of imperfect bowel cleansing and image artifacts on VC performance is significant. The texture-based CADpolyp detected all the polyps with an average of 2.68 false positives per patient. This indicates that texture-based CADpolyp can improve the CTC performance in the cases of imperfect cleansed bowels and presence of image artifacts.

KW - Colonic material tagging

KW - CTC

KW - PV image segmentation

KW - Texture based CADpolyp

KW - Tissue mixture modeling

UR - http://www.scopus.com/inward/record.url?scp=44349151684&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=44349151684&partnerID=8YFLogxK

U2 - 10.1117/12.768073

DO - 10.1117/12.768073

M3 - Conference contribution

AN - SCOPUS:44349151684

SN - 9780819471000

VL - 6916

BT - Progress in Biomedical Optics and Imaging - Proceedings of SPIE

ER -