Navigation for percutaneous surgical interventions: ultrasound data processing, feature extraction and 3D organ reconstruction

Surgical practice is constantly replacing traditional invasive approaches with minimally invasive techniques, which provide many benefits for the patient such as reduced postoperative complications, faster recovery and shorten hospitalization. Percutaneous procedures are probably the most widely used minimally invasive technique, this approach is used both for diagnosis (i.e. biopsy) and for the treatment of localized pathological areas in many regions of the human body: head and neck, limbs, lungs, hearth, abdominal and pelvic structures. In percutaneous ablation a surgical tool that resembles a thick needle is accurately inserted through the skin of the patient in correspondence of a pathological area to destroy diseased tissue thanks to high energy or low temperature. The latter approach is called percutaneous cryoablation and provides very encouraging mid-term outcome compared with other minimally invasive techniques. Percutaneous ablation requires accurate positioning of the tools to guarantee the complete treatment of diseased area, thus reducing the possible re-insurgence of same pathological condition or other intra or post operative complications. Although these procedures could be performed blindly the introduction of image guidance could improve the outcome of the procedure even in the case of inexpert users or critical cases. Therefore percutaneous ablation could greatly benefit from the introduction of a computer navigation system, able to provide integrated guidance with imaging sensor and tools localization. Required characteristics are the ability of performing automatic registration with other intra-operative and pre-operative dataset, planning needle trajectory and guidance during the insertion with effective and user-friendly interface. Furthermore the accurate monitoring of the needle position and ablation area is fundamental to obtain a full treatment of the pathological area. Ultrasound guided systems provide the needed characteristics in a compact and coste ffective device. Advance processing techniques are necessary to overcome some limitations of ultrasound based system, mainly connected to limited quality and resolution of these images. The introduction of a tracking system able to measure the position and the orientation of ultrasound probe and tools is a key component of a navigation system, and could enable the automatic 3D reconstruction of patient anatomy based on the extraction of feature points from 2D images. To enable the correct 3D reconstruction a calibration procedure is necessary to estimate an unknown rigid transformation between tracking and image reference systems. In percutaneous ablation procedures, monitoring the ablation area is fundamental to guarantee to complete suppression of the diseased tissues and avoid short term recurrence. The specific case of cryoablation is very challenging since the frozen zone produces strong re ections and shadows that make it unfeasible to guarantee the proper monitoring of the ablated area. In this thesis we describe different methods able to improve the current ultrasound based navigation systems. The first contribution of this thesis is the introduction of an ultrasound calibration procedure based on data acquired with different acquisition depth settings for the computation of a single rigid transformation. The procedure is based on the automatic estimation of the beam width to improve the integration of feature extracted from different depth settings. The proposed method enables the calibration of ultrasound probe with less strict control of the acquisition parameters and without significant effect on the calibration accuracy and precision. This would ease the adoption of tracked ultrasound system in the real clinical condition, where acquisition parameters are changed very often to guarantee the correct visualization of patient anatomy. The second contribution of this thesis is the development of a feature detector and descriptor able to localize and match salient points from ultrasound image. These methods have been designed and tested with specific attention to ultrasound images, where the intensity information are not stable. The detector is based a local energy model in place of the widely adopted gradient methods, where feature points are localized based on the phase congruency of Fourier components. The detector is based on Local Binary Pattern operator computed over a local angle and direction of the phase congruency. These choices enable the robust localization of feature point in ultrasound image in presence of intensity and geometrical transformation. These feature points could be used for the image based localization of tools or pathological areas, or to enable the registration with other imaging dataset acquired before, during or after the surgical procedure. The third contribution is the introduction of a compact navigation device that integrates a small display directly into the ablation tool to provide indication on how to move the needle in the proper way and to guarantee the correct insertion along the planned trajectory. Providing indications to the user in a confortable and effective manner is fundamental to guarantee the correct execution of the planned trajectory, consequently improving the final result of the procedure with reduced risks and reduced procedure duration. The fourth contribution overcomes the unfeasible monitoring of the ablation area with ultrasound image during cryoablation procedures, and it is based on ultrasound elastography. Ultrasound elastography is used for the measurement of tissue elasticity; since the freezing produces cellular structural damage we measure the tissue elastic properties before and after the complete thawing of the tissue. This method would enable the monitoring of cryo ablated area immediately after the intervention, thus enabling the effective adoption of corrective actions in case of not complete treatment or complications. We believe that the contributions described in this thesis, if integrated in an ultrasound guided navigation system, will improve the characteristics of these advanced systems with specific attention to real clinical requirements.