Bidirectional segmentation of prostate capsule from ultrasound volumes: an improved strategy

Liyang Wei; Ramkrishnan Narayanan; Dinesh Kumar; Aaron Fenster; Albaha Barqawi; Priya Werahera; E. David Crawford; Jasjit S. Suri

doi:10.1117/12.770569

Translator Disclaimer

19 March 2008 Bidirectional segmentation of prostate capsule from ultrasound volumes: an improved strategy

Liyang Wei, Ramkrishnan Narayanan, Dinesh Kumar, Aaron Fenster, Albaha Barqawi, Priya Werahera, E. David Crawford, Jasjit S. Suri

Author Affiliations +

Proceedings Volume 6914, Medical Imaging 2008: Image Processing; 69143W (2008) https://doi.org/10.1117/12.770569
Event: Medical Imaging, 2008, San Diego, California, United States

Abstract

Prostate volume is an indirect indicator for several prostate diseases. Volume estimation is a desired requirement during prostate biopsy, therapy and clinical follow up. Image segmentation is thus necessary. Previously, discrete dynamic contour (DDC) was implemented in orthogonal unidirectional on the slice-by-slice basis for prostate boundary estimation. This suffered from the glitch that it needed stopping criteria during the propagation of segmentation procedure from slice-to-slice. To overcome this glitch, axial DDC was implemented and this suffered from the fact that central axis never remains fixed and wobbles during propagation of segmentation from slice-to-slice. The effect of this was a multi-fold reconstructed surface. This paper presents a bidirectional DDC approach, thereby removing the two glitches. Our bidirectional DDC protocol was tested on a clinical dataset on 28 3-D ultrasound image volumes acquired using side fire Philips transrectal ultrasound. We demonstrate the orthogonal bidirectional DDC strategy achieved the most accurate volume estimation compared with previously published orthogonal unidirectional DDC and axial DDC methods. Compared to the ground truth, we show that the mean volume estimation errors were: 18.48%, 9.21% and 7.82% for unidirectional, axial and bidirectional DDC methods, respectively. The segmentation architecture is implemented in Visual C++ in Windows environment.

Citation Download Citation

Liyang Wei, Ramkrishnan Narayanan, Dinesh Kumar, Aaron Fenster, Albaha Barqawi, Priya Werahera, E. David Crawford, and Jasjit S. Suri "Bidirectional segmentation of prostate capsule from ultrasound volumes: an improved strategy", Proc. SPIE 6914, Medical Imaging 2008: Image Processing, 69143W (19 March 2008); https://doi.org/10.1117/12.770569

ACCESS THE FULL ARTICLE