Résumé
To achieve needle steering and guide a flexible needle,it is necessary to localize it first. In the case of 3Dultrasound (US) in B-mode, the poor imaging quality andartefacts make it difficult to determine the needle pose.Needle localization issues specific to needle steering arereviewed in [1]. In [2], we proposed an observer of theneedle tip pose. In [3], we detailed a machine learningapproach for needle localization in 3D US volumes. Inthis paper, we describe the interconnection of bothmethods for precise curved needle localization in thecontext of needle steering.