X-rays remain the preferred imaging modality for orthopedic surgery for surgical planning, intra-operative control or patient follow-up. Nevertheless, it does not allow anatomical bone structures referencing. It is then impossible to control geometrical modifications of bone structures during the surgical process. However, surgical tools are referenced in the operating-room space and allow the surgeon to define anatomical structures geometrically by defining landmark positions. This process is only allowed during surgical procedures because it requires to do cuts on the patient. In this work, we propose a new approach using an ultrasound probe that is referenced in the operating-room space. We present an image processing algorithm to extract anatomical landmark position in a surgical context. It is a crucial improvement because it allows a complete patient follow-up from pre-operative planning to post-operative consults. To determine anatomical landmark positions on ultrasound images we added an intermediate step to extract the bone/soft tissues interface via several segmentation methods as active contours. Due to the low quality of ultrasound images we decided to design a innovative cooperative process. As the surgeon positions the ultrasound probe on the patient, the bone interface appears on the system screen in real time. Then the clinician can help the segmentation result to converge to the final solution by a soft movement of the probe. The validation of our work was performed on a database of 651 ultrasound images, in a non-cooperative way. The best algorithm that extracts the bone interface by defining the optimal path in a graph of potential candidates was validated with a cooperative protocol on a prototype called PhysioPilot, in order to perform physiological measurements in a non-surgical context
