Skip to Main content Skip to Navigation

3D segmentation and registration for minimal invasive prostate cancer therapy

Abstract : The work of this Thesis is focused on image guided focal therapy of prostate cancer by High Intensity Focused Ultrasound (HIFU). Currently MRI is the only imaging technique that can locate the tumor in prostate. In contrast, the tumor is not visible in the ultrasound image which is used to guide the HIFU planning and therapy. The aim of the Thesis is to provide registration techniques of T2 MRI to ultrasound. Two approaches were explored: 1) Region-based registration. More particularly, we studied an ultrasound texture descriptors based on moments invariant to rotation and scaling. These descriptors are sensitive to speckle distribution regardless of the scale or the orientation. As we expected, some of these descriptors can be used to characterize regions sharing a similar speckle spatial distribution. But, we also found that some other descriptors were sensitive to the contours of these regions. This property seems very useful to adapt the classical boundary-based or mixed region/boundary-based segmentation methods (active contours, graph cut, etc.) to process US images. 2) Surface-based registration approach.. We adapted the Optimal Definition Surface (OSD) method to the segmentation of the prostate in T2 MRI, Furthermore, we proposed the multiple-objects OSD which is a concurrent segmentation of the prostate, bladder and rectum. Finally we used the prostate surface extracted from the ultrasound volume and from T2 MRI in a surface-to-surface elastic registration scheme. This registration allowed us to merge the preoperative MR information in the peroperative US volume.
Complete list of metadata

Cited literature [75 references]  Display  Hide  Download
Contributor : ABES STAR :  Contact
Submitted on : Friday, August 29, 2014 - 10:18:08 AM
Last modification on : Wednesday, September 14, 2022 - 10:20:04 AM


Version validated by the jury (STAR)


  • HAL Id : tel-00962028, version 2


Ke Wu. 3D segmentation and registration for minimal invasive prostate cancer therapy. Signal and Image processing. Université Rennes 1, 2014. English. ⟨NNT : 2014REN1S015⟩. ⟨tel-00962028v2⟩



Record views


Files downloads