Skip to Main content Skip to Navigation
Theses

Méthodes et outils pour le détourage des malformations artério-veineuses cérébrales dans un contexte multimodalité

Abstract : Radiosurgery of brain arteriovenous malformations (AVM) requires a precise localization of the volume of the AVM. This volume is at present reconstructed by cone intersection from the manual delineations performed by a medical expert in 2D digital subtracted angiographic (2DSA) views. Regarding the often complex shape of the AVM and the few number of views the resulting volume is far from the actual shape. Moreover the difficulties in the delineation led to an important inter and intra-observer variability. The aim of this study was to develop a new method using 3D imaging and improving the accuracy of this volume determination. We first focused on the validation of 3D X-ray angiography (3DXA) as a 3D imaging modality for delineation of the AVM. This validation presented with difficult problems of methodology. We compared the delineations performed in 3DXA with those performed in 2DSA taken as the reference. In practice we constructed statistical models derived from a set of 3 experts delineations on 2DSA in a clinical database of AVMs treated by radiosurgery. Then, using this shape database as a reference, we evaluate 3DXA delineations performed by the same experts. For that purpose we designed a statistical test for acceptation/rejection of these 3DXA delinations. The results proved that 3DXA could be used for AVMs delineation. We then developed a new method using deformable surfaces applied to 3DXA volume in order to refine the initial estimate of the volume provided by the delineations in 2DSA views. The resulting volumes well matched the reference shape database and demonstrated an important reduction of the inter-observer variability. We now plan to test this method on a large range of MAVs. Moreover we have to keep on the beginning works on the use of 3D MRI.
Document type :
Theses
File URL :
http://docnum.univ-lorraine.fr/prive/SCD_T_2003_0154_ANXIONNAT.pdf
Complete list of metadata

https://hal.univ-lorraine.fr/tel-01746830
Contributor : Thèses Ul <>
Submitted on : Thursday, March 29, 2018 - 10:45:40 AM
Last modification on : Thursday, September 24, 2020 - 2:08:16 PM

Identifiers

  • HAL Id : tel-01746830, version 1

Collections

Citation

René Anxionnat. Méthodes et outils pour le détourage des malformations artério-veineuses cérébrales dans un contexte multimodalité. Autre [cs.OH]. Université Henri Poincaré - Nancy 1, 2003. Français. ⟨NNT : 2003NAN10154⟩. ⟨tel-01746830⟩

Share

Metrics

Record views

20