Claire Chalopin

Active contours driven by Cuckoo Search strategy for brain tumour images segmentation

An alternative Active Contour Model solution for medical images is introduced.A multi-population Cuckoo Search Strategy (MCSS) is implemented to boost ACM.Proposed method was applied on Magnetic Resonance Imaging (MRI) data.MCSS outperforms traditional ACM and ACM driven by multi-population PSO. In this paper, an alternative Active Contour Model (ACM) driven by Multi-population Cuckoo Search (CS) algorithm is introduced. This strategy assists the converging of control points towards the global minimum of the energy function, unlike the traditional ACM version which is often trapped in a local minimum. In the proposed methodology, each control point is constrained in a local search window, and its energy minimisation is performed through a Cuckoo Search via Levy flights paradigm. With respect to local search window, two shape approaches have been considered: rectangular shape and polar coordinates. Results showed that the CS method using polar coordinates is generally preferable to CS performed in rectangular shapes. Real medical and synthetic images were used to validate the proposed strategy, through three performance metrics as the Jaccard index, the Dice index and the Hausdorff distance. Applied specifically to Magnetic Resonance Imaging (MRI) images, the proposed method enables to reach better accuracy performance than the traditional ACM formulation, also known as Snakes and the use of Multi-population Particle Swarm Optimisation (PSO) algorithm.

Integration of a 3D ultrasound probe into neuronavigation.

BackgroundIntraoperative ultrasound (iUS) allows the generation of real-time data sets during surgical interventions. The recent innovation of 3D ultrasound probes permits the acquisition of 3D data sets without the need to reconstruct the volume by 2D slices. This article describes the integration of a tracked 3D ultrasound probe into a neuronavigation.MethodsAn ultrasound device, provided with both a 2D sector probe and a 3D endocavity transducer, was integrated in a navigation system with an optical tracking device. Navigation was performed by fusion of preoperatively acquired MRI data and intraoperatively acquired ultrasound data throughout an open biopsy. Data sets with both probes were acquired transdurally and compared.ResultsThe acquisition with the 3D probe, processing and visualization of the volume only took about 2 min in total. The volume data set acquired by the 3D probe appears more homogeneous and offers better image quality in comparison with the image data acquired by the 2D probe.ConclusionsThe integration of a 3D probe into neuronavigation is possible and has certain advantages compared with a 2D probe. The risk of injury can be reduced, and the application can be recommended for certain cases, particularly for small craniotomies.

Localized active contour model with background intensity compensation applied on automatic MR brain tumor segmentation

Abstract This paper presents a Localized Active Contour Model (LACM) integrating an additional step of background intensity compensation. The region-based active contour models that use statistical intensity information are more sensitive to the high mean intensity distance between consecutive regions. In Magnetic Resonance Imaging (MRI) this distance is great between the foreground and the background, hence it leads to an incorrect delineation of the target. In order to resolve this problem, an automatic process is introduced in our model for balancing the mean intensity distance between an image foreground and its background. The aim is to minimize the attraction effect of the active contour model to the undesired borderlines defined by these two mentioned image regions. By using this approach not only the obtained accuracy outperforms the traditional localized mean separation active contour model, but also it reduces the computation time of the segmentation task. In addition, this method was efficiently applied on automatic brain tumor segmentation in multimodal MRI data. The Hierarchical Centroid Shape Descriptor (HCSD) was used for detecting the region of interest i.e. abnormal tissue so as to automatically initialize the active contour. The validation of experiments was carried out on synthetic images and the quantitative evaluation was performed on the BRATS2012 database. Finally, the accuracy achieved by the proposed method was compared to the localized mean separation intensity, the localized Chan-Vese, the local Gaussian distribution fitting and the local binary fitting models by using the Dice coefficient, Sensitivity, Specificity and the Hausdorff distance. The computation time of the methods was also measured for comparison purposes. The obtained results show that the proposed model outperforms the accuracy of the selected state of the art methods. Moreover, it is also faster than the comparative methods in the medical image segmentation task.

Localization and tracking of aortic valve prosthesis in 2D fluoroscopic image sequences

This paper presents a new method for localization and tracking of the aortic valve prosthesis (AVP) in 2D fluoroscopic image sequences to assist the surgeon to reach the safe zone of implantation during transapical aortic valve implantation. The proposed method includes four main steps: First, the fluoroscopic images are preprocessed using a morphological reconstruction and an adaptive Wiener filter to enhance the AVP edges. Second, a target window, defined by a user on the first image of the sequences which includes the AVP, is tracked in all images using a template matching algorithm. In a third step the corners of the AVP are extracted based on the AVP dimensions and orientation in the target window. Finally, the AVP model is generated in the fluoroscopic image sequences. Although the proposed method is not yet validated intraoperatively, it has been applied to different fluoroscopic image sequences with promising results.

A neurosurgical phantom-based training system with ultrasound simulation

BackgroundBrain tumor surgeries are associated with a high technical and personal effort. The required interactions between the surgeon and the technical components, such as neuronavigation, surgical instruments and intraoperative imaging, are complex and demand innovative training solutions and standardized evaluation methods. Phantom-based training systems could be useful in complementing the existing surgical education and training.MethodsA prototype of a phantom-based training system was developed, intended for standardized training of important aspects of brain tumor surgery based on real patient data. The head phantom consists of a three-part construction that includes a reusable base and adapter, as well as a changeable module for single use. Training covers surgical planning of the optimal access path, the setup of the navigation system including the registration of the head phantom, as well as the navigated craniotomy with real instruments. Tracked instruments during the simulation and predefined access paths constitute the basis for the essential objective training feedback.ResultsThe prototype was evaluated in a pilot study by assistant physicians at different education levels. They performed a complete simulation and a final assessment using an evaluation questionnaire. The analysis of the questionnaire showed the evaluation result as “good” for the phantom construction and the used materials. The learning effect concerning the navigated planning was evaluated as “very good”, as well as having the effect of increasing safety for the surgeon before planning and conducting craniotomies independently on patients.ConclusionsThe training system represents a promising approach for the future training of neurosurgeons. It aims to improve surgical skill training by creating a more realistic simulation in a non-risk environment. Hence, it could help to bridge the gap between theoretical and practical training with the potential to benefit both physicians and patients.

Evaluation of a semi-automatic segmentation algorithm in 3D intraoperative ultrasound brain angiography.

Abstract In this work, we adapted a semi-automatic segmentation algorithm for vascular structures to extract cerebral blood vessels in the 3D intraoperative contrast-enhanced ultrasound angiographic (3D-iUSA) data of the brain. We quantitatively evaluated the segmentation method with a physical vascular phantom. The geometrical features of the segmentation model generated by the algorithm were compared with the theoretical tube values and manual delineations provided by observers. For a silicon tube with a radius of 2 mm, the results showed that the algorithm overestimated the lumen radii values by about 1 mm, representing one voxel in the 3D-iUSA data. However, the observers were more hindered by noise and artifacts in the data, resulting in a larger overestimation of the tube lumen (twice the reference size). The first results on 3D-iUSA patient data showed that the algorithm could correctly restitute the main vascular segments with realistic geometrical features data, despite noise, artifacts and unclear blood vessel borders. A future aim of this work is to provide neurosurgeons with a visualization tool to navigate through the brain during aneurysm clipping operations.

Vascular Structure Identification in Intraoperative 3D Contrast-Enhanced Ultrasound Data

In this paper, a method of vascular structure identification in intraoperative 3D Contrast-Enhanced Ultrasound (CEUS) data is presented. Ultrasound imaging is commonly used in brain tumor surgery to investigate in real time the current status of cerebral structures. The use of an ultrasound contrast agent enables to highlight tumor tissue, but also surrounding blood vessels. However, these structures can be used as landmarks to estimate and correct the brain shift. This work proposes an alternative method for extracting small vascular segments close to the tumor as landmark. The patient image dataset involved in brain tumor operations includes preoperative contrast T1MR (cT1MR) data and 3D intraoperative contrast enhanced ultrasound data acquired before (3D-iCEUSstart) and after (3D-iCEUSend) tumor resection. Based on rigid registration techniques, a preselected vascular segment in cT1MR is searched in 3D-iCEUSstart and 3D-iCEUSend data. The method was validated by using three similarity measures (Normalized Gradient Field, Normalized Mutual Information and Normalized Cross Correlation). Tests were performed on data obtained from ten patients overcoming a brain tumor operation and it succeeded in nine cases. Despite the small size of the vascular structures, the artifacts in the ultrasound images and the brain tissue deformations, blood vessels were successfully identified.

Brain tumor enhancement revealed by 3D intraoperative ultrasound imaging in a navigation system

Introduction. Brain tumour resections are commonly guided using a navigation system based on preoperative MR data. In order to update the brain status which deforms during the surgery, 3D intraoperative ultrasound (3D-iUS) data are acquired. Tumour interpretation in the 3D-iUS data is however complex because of the image noise, artefacts and missing tumour parts. We propose therefore to enhance the visualization of the brain tumour in the 3D-iUS data using the navigation system. Methods. Our method is based on a tumour model generated by semi-automatic segmentation of the tumour in the preoperative MR data. The tumour model is pre-registered into the patient coordinate system using the transform matrix computed by the navigation system. The brain shift is then estimated using a rigid block matching algorithm between the pre-registered tumour model and the 3D-iUS data. The obtained fine-registered tumour model surface is finally described as a mesh and visualized in the navigation system overlapped on the 3D-iUS data. Results. This method was off-line tested on five patient data including 3D-iUS data and enhanced 3D-iUS data acquired with a contrast agent. In order to evaluate our method the Dice Similarity Index (DSI) was respectively computed between manual delineations of the tumours in the 3D-iUS data considered as references and the pre-registered and fineregistered tumour model. Results demonstrated that the DSI values increased after fine registration and in the enhanced 3D-iUS data. Conclusion. The advantage of using a tumour model combined to a rigid registration technique is to be robust to image artefacts and to be able to reconstitute the tumour information when it is missing in the 3D-iUS data. In order to better match the tumour model with the 3D-iUS data, the fine registered model will be used as initialization to a deformable segmentation method.

Automatic selection of localized region-based active contour models using image content analysis applied to brain tumor segmentation.

Brain tumor segmentation is a routine process in a clinical setting and provides useful information for diagnosis and treatment planning. Manual segmentation, performed by physicians or radiologists, is a time-consuming task due to the large quantity of medical data generated presently. Hence, automatic segmentation methods are needed, and several approaches have been introduced in recent years including the Localized Region-based Active Contour Model (LRACM). There are many popular LRACM, but each of them presents strong and weak points. In this paper, the automatic selection of LRACM based on image content and its application on brain tumor segmentation is presented. Thereby, a framework to select one of three LRACM, i.e., Local Gaussian Distribution Fitting (LGDF), localized Chan-Vese (C-V) and Localized Active Contour Model with Background Intensity Compensation (LACM-BIC), is proposed. Twelve visual features are extracted to properly select the method that may process a given input image. The system is based on a supervised approach. Applied specifically to Magnetic Resonance Imaging (MRI) images, the experiments showed that the proposed system is able to correctly select the suitable LRACM to handle a specific image. Consequently, the selection framework achieves better accuracy performance than the three LRACM separately.

Real Time Issues for usage of Vision and Image Data in the Future Operating Room

The operating room (OR) is more and more equipped with surgical assist systems which make use of modern image acquisition and processing technologies. Within this scenario, the surgeon has to deal with a large amount of information which must be available at the right time and in the right place. This results in various real-time problems which must be addressed while building surgical assist systems. The components of such systems and their real time requirements are presented in this paper. Several concrete systems are presented and their real time properties discussed