Semi-automatic segmentation of 3D ultrasound images of the prostate
This project aims at facilitating the segmentation of the prostate in ultrasound 3D images using shape priors. An average prostate coming from a statistical atlas is positioned in the ultrasound volume and deformed to fit to user-defined points of the prostate contours. Generally, thirty to forty points are sufficient to guarantee a precise result. Where 20 to 30 minutes were necessary in a completely manual segmentation, a similarly good segmentation can be obtained in five minutes with this method.
Reference: K. Dowlut, M. Baumann, S. Martin, J. Troccaz - Semi-automatic segmentation of prostate ultrasound images using shape priors - Poster presented to: Engineering and Urology Society 24^th Annual Meeting, Chicago, 25 Avril 2009 – on-line proceedings freely available
Collaborations: Clinical partners and KOELIS company.
Status: FINISHED This project was developed by Kevin Dowlut during a training session of the « Magistère informatique” de l’UJF (Grenoble University). The developed tools have been integrated in the Urostation platform of KOELIS.
Contact : jocelyne.troccaz@imag.fr