Statistical shape and texture model of quadrature phase information for prostate segmentation

Abstract : Purpose: Prostate volume estimation from segmentation of transrectal ultrasound (TRUS) images aids in diagnosis and treatment of prostate hypertro- phy and cancer. Computer-aided accurate and compu- tationally efficient prostate segmentation in TRUS im- ages is a challenging task, owing to low signal-to-noise ratio, speckle noise, calcifications and heterogeneous in- tensity distribution in the prostate region. Method: A multi-resolution framework using texture features in a parametric deformable statistical model of shape and appearance was developed to segment the prostate. Local phase information of log-Gabor quadra- ture filter extracted texture of the prostate region in TRUS images. Large bandwidth of log-Gabor filter en- sures easy estimation of local orientations and zero re- sponse for a constant signal provides invariance to gray level shift. This aids in enhanced representation of the underlying texture information of the prostate unaf- fected by speckle noise and imaging artifacts. The para- metric model of the propagating contour is derived from principal component analysis of prior shape and texture information of the prostate from the training data. The Soumya Ghose*, Jhimli Mitra*, Arnau Oliver, Robert Mart'ı, Xavier Llad'o and Jordi Freixenet Computer Vision and Robotics Group, University of Girona Campus Montilivi, Edifici P-IV,17071 Girona, Spain. E-mail: soumyaghose@gmail.com, jhimlimitra@yahoo.com, {aoliver, marly, llado, and jordif}@eia.udg.edu Joan C.Vilanova Clinica Girona, Calle Joan Maragall 26, 17002 Girona, Spain. Josep Comet University Hospital Dr. Josep Trueta, Av. Frana, 17007 Girona, Spain. Fabrice Meriaudeau *Laboratoire Le2I - UMR CNRS 5158, Universit'e de Bour- gogne,12 Rue de la Fonderie, 71200 Le Creusot, Bourgogne, France. E-mail: fabrice.meriaudeau@u-bourgogne.fr. parameters were modified using prior knowledge of the optimization space to achieve segmentation. Results: The proposed method achieves a mean Dice similarity coefficient value of 0.95±0.02, and mean ab- solute distance of 1.26±0.51 millimeter when validated with 24 TRUS images of 6 datasets in a leave-one- patient-out validation framework. Conclusions: The proposed method for prostate TRUS image segmentation is computationally efficient and pro- vides accurate prostate segmentations in presence of in- tensity heterogeneities and imaging artifacts.
Document type :
Journal articles
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-00612739
Contributor : Fabrice Meriaudeau <>
Submitted on : Saturday, July 30, 2011 - 3:42:14 PM
Last modification on : Wednesday, September 12, 2018 - 1:27:10 AM

Links full text

Identifiers

Collections

Citation

Soumya Ghose, Arnau Oliver, Robert Marti, Xavier Llado, Jhimli Mitra, et al.. Statistical shape and texture model of quadrature phase information for prostate segmentation. International Journal of Computer Assisted Radiology and Surgery, Springer Verlag, 2011, http://www.springerlink.com/content/c841254357h734r1/. ⟨10.1007/s11548-011-0616-y⟩. ⟨hal-00612739⟩

Share

Metrics

Record views

232