Section: New Results
Image Guided Intervention
Automatic geometrical and statistical detection of acoustic shadows
Participants : Pierre Hellier, Xavier Morandi.
In ultrasound images, acoustic shadows appear as regions of low signal intensity linked to boundaries with very high acoustic impedance differences. Acoustic shadows can be viewed either as informative features to detect lesions or calcifications, or as damageable artifacts for image processing tasks such as segmentation, registration or 3D reconstruction. In both cases, the detection of these acoustic shadows is useful. We have designed a new method [20] to detect these shadows that combines a geometrical approach to estimate the B-scans shape, followed by a statistical test based on a dedicated modeling of ultrasound image statistics. Results demonstrate that this detection improves the reconstruction and registration of tracked intraoperative brain ultrasound images.
Automatic steps recognition in neurosurgical procedures by microscope video analysis
Participants : Florent Lalys, Xavier Morandi, Laurent Riffaud, Pierre Jannin.
Analysis of surgical procedures is a recent field allowing the creation of complex patient-specific and generic surgical process models. With the increased number of technological tools incorporate in the Operating Room (OR), the need for new computer-assisted system has emerged. By extracting different signals (at different levels of granularity) from the OR, it's possible to extract helpful information such as the surgical workflow followed by the surgeons. Our project, in collaboration with Carl Zeiss Medical Systems (Oberkochen, Germany) is based on the extraction of information from digital microscope videos. We decided in a first step to use static information, without taking into account the motion, by extracting image features from videos and by training models with machine learning techniques. Image data-bases have been constructed for different types of neurosurgical procedures. The problem is thus reduced to an image classification problem which allow us to automatically detect the different steps of a procedure but also to recognize other helpful information. We evaluated the approach with a database of 18 videos of transsphenoidal pituitary surgeries. 400 images were extracted and we obtained for the step segmentation a correct classification rate of 82 percent.
Comparison of classification methods for modeling neurosurgical process
Participants : Brivael Trelhu, Laurent Riffaud, Xavier Morandi, Pierre Jannin, Florent Lalys.
We performed a study for the analysis of neurosurgical procedures with the aim to improve their comprehension and optimize the surgery. We first used a XML Database included 157 tumors surgeries descriptions. 85 variables were identified and classified into predictive variables (i.e., known before operation) and into variables to be predicted (i.e., pertaining for the surgical gestures). We investigated a six classify methods comparison to determine the mostly adapted method for classifying and predicting neurosurgical procedures steps. Five criteria (correct rate (CR), sensitivity (SEN), specificity (SPC), positive predictive value (PPV) and negative predictive value (NPV)) were used to compare the methods against. Results were studied and interpreted by an expert neurosurgeon to estimate and to validate medical applications.
Cognitive analysis of surgical planning and information requirements in image guided neurosurgery
Participants : Pierre Jannin, Xavier Morandi.
With the GRESICO (Groupe de REcherche en Sciences de l'Information et de la COgnition) laboratory (Thierry Morineau, Nadege Le Moellic) from the Université de Bretagne Sud in Vannes (France), we have defined a methodology for identifying differences in cognitive behaviour between neurosurgeons with different expertise levels. 9 neurosurgeons were interviewed. First results indicate a clear distinction between surgeons and provide a basis for further analysis. We also developed a method for analysis the surgical work domain and used it for assessing information provided by image guided surgery systems for surgical planning. We are currently applying the results in a study comparing different display modes (2-D vs. 3-D display) with different users (engineers, medical students, and surgeons).
Post operative assessment of Deep Brain Stimulation (DBS) based on multimodal images
Participants : Pierre Jannin, Florent Lalys, Claire Haegelen, Jean-Christophe Ferré, Xavier Morandi, Jean-Yves Gauvrit.
Deep Brain Stimulation (DBS) is a surgical procedure used from about 20 years mainly for functional neurosurgery of Parkinson disease. It consists of inserting and stimulating an electrode within deep brain structures such as the sub thalamic nucleus (STN). For patients suffering of movement disorders, medical therapy could be not effective. In that case high frequency electrical stimulation via the electrode will considerably reduce the functional pathology. The targeting of the STN is based on anatomic, imaging and stastistical data obtained on anatomic and clinical studies. By using pre and post operative multimodal images and clinical scores, we developed an approach for post operative assessment of DBS. It includes the automatic segmentation of the electrode from post operative CT, and the development of a registration workflow in order to express the coordinates of the electrodes and the stimulated electrode contacts in a common coordinates system for different patients. We built a 3T MR mono subject template in order to make easier comparison between different subjects which serves as the common coordinate system. This template allow visualisation of spatially complex structures as well as increased contrast. We also showed that it greatly improved the accuracy of template based registration. The registration workflow between pre and post operative images and the anatomical MR template was validated on clinical data. Finally, we developed a methodology for building an anatomical and clinical digital atlas gathering information about stimulated electrode contacts and related pre and post operative clinical scores. Motor scores as well as neuro-psychological tests were included in the study. We performed statistical analysis of relationships between clusters of 3D points and improvements of clinical scores. This methodology was tested for a population of 9 parkinsonian patients implanted in the STN. It aims at highlighting anatomical areas with better clinical results and lower clinical side effects in order to find the optimal site for STN DBS.
Comparison of Piece-Wise Linear, Linear and Nonlinear Atlas-to-Patient Warping Techniques: Analysis of the labeling of subcortical nuclei for functional neurosurgical applications
Participant : Pierre Hellier.
Digital atlases are commonly used in pre-operative planning in functional neurosurgical procedures performed to minimize the symptoms of Parkinson's disease. These atlases can be customized to fit an individual patient's anatomy through atlas-to-patient warping procedures. We have participated in an evaluation study [13] of eight different registration methods for atlas-to- patient customization of a new digital atlas of the basal ganglia and thalamus to demonstrate the value of non-linear registration for automated atlas-based subcortical target identification in functional neurosurgery. Since a gold standard of the subcortical anatomy is not available, manual segmentations of the striatum, globus pallidus, and thalamus are used to derive a silver standard for evaluation. The results show that nonlinear techniques perform statistically better than linear and piece-wise linear techniques.
Automated Surgical Planning
Participants : Caroline Essert-Villard, Omar El Ganaoui, Xavier Morandi, Claire Haegelen, Pierre Jannin.
Surgical Planning consists in identifying optimal access to the target based on anatomical references and constrained by healthy functional areas. For helping this process, we aim at automatically computing possible surgical approaches, respecting patient specific constraints expressed from preoperative images (MR and CT) and generic constraints expressed from patient-adapted atlases. The first application, with the participation of Dr. C. Haegelen from the neurosurgical department of the university hospital, focuses on the automatic planning of the implant of deep brain stimulation electrodes (DBS) for the treatment of Parkinson's disease. The purpose is to find an optimal trajectory for a cylindrical electrode to a target located in deep structures of the brain (e.g. sub thalamus nucleus). The method we are developing is using a formalization of the expertise of the surgeon as well as preoperative images (MR and CT), sent to a geometrical constraint solver to produce a space of possible solutions weighted with a quantification of their quality. Our latest results allow us to define in a few milliseconds the areas of possible insertion points of DBS electrodes, according to the brain anatomy extracted from the pre-operative images.
