Comparison of manual and semi-automatic registration in augmented reality image-guided liver surgery: a clinical feasibility study.

BACKGROUND: The laparoscopic approach to liver resection may reduce morbidity and hospital stay. However, uptake has been slow due to concerns about patient safety and oncological radicality. Image guidance systems may improve patient safety by enabling 3D visualisation of critical intra- and extrahepatic structures. Current systems suffer from non-intuitive visualisation and a complicated setup process. A novel image guidance system (SmartLiver), offering augmented reality visualisation and semi-automatic registration has been developed to address these issues. A clinical feasibility study evaluated the performance and usability of SmartLiver with either manual or semi-automatic registration. METHODS: Intraoperative image guidance data were recorded and analysed in patients undergoing laparoscopic liver resection or cancer staging. Stereoscopic surface reconstruction and iterative closest point matching facilitated semi-automatic registration. The primary endpoint was defined as successful registration as determined by the operating surgeon. Secondary endpoints were system usability as assessed by a surgeon questionnaire and comparison of manual vs. semi-automatic registration accuracy. Since SmartLiver is still in development no attempt was made to evaluate its impact on perioperative outcomes. RESULTS: The primary endpoint was achieved in 16 out of 18 patients. Initially semi-automatic registration failed because the IGS could not distinguish the liver surface from surrounding structures. Implementation of a deep learning algorithm enabled the IGS to overcome this issue and facilitate semi-automatic registration. Mean registration accuracy was 10.9 ± 4.2 mm (manual) vs. 13.9 ± 4.4 mm (semi-automatic) (Mean difference - 3 mm; p = 0.158). Surgeon feedback was positive about IGS handling and improved intraoperative orientation but also highlighted the need for a simpler setup process and better integration with laparoscopic ultrasound. CONCLUSION: The technical feasibility of using SmartLiver intraoperatively has been demonstrated. With further improvements semi-automatic registration may enhance user friendliness and workflow of SmartLiver. Manual and semi-automatic registration accuracy were comparable but evaluation on a larger patient cohort is required to confirm these findings.

Performance of image guided navigation in laparoscopic liver surgery - A systematic review.

BACKGROUND: Compared to open surgery, minimally invasive liver resection has improved short term outcomes. It is however technically more challenging. Navigated image guidance systems (IGS) are being developed to overcome these challenges. The aim of this systematic review is to provide an overview of their current capabilities and limitations. METHODS: Medline, Embase and Cochrane databases were searched using free text terms and corresponding controlled vocabulary. Titles and abstracts of retrieved articles were screened for inclusion criteria. Due to the heterogeneity of the retrieved data it was not possible to conduct a meta-analysis. Therefore results are presented in tabulated and narrative format. RESULTS: Out of 2015 articles, 17 pre-clinical and 33 clinical papers met inclusion criteria. Data from 24 articles that reported on accuracy indicates that in recent years navigation accuracy has been in the range of 8-15 mm. Due to discrepancies in evaluation methods it is difficult to compare accuracy metrics between different systems. Surgeon feedback suggests that current state of the art IGS may be useful as a supplementary navigation tool, especially in small liver lesions that are difficult to locate. They are however not able to reliably localise all relevant anatomical structures. Only one article investigated IGS impact on clinical outcomes. CONCLUSIONS: Further improvements in navigation accuracy are needed to enable reliable visualisation of tumour margins with the precision required for oncological resections. To enhance comparability between different IGS it is crucial to find a consensus on the assessment of navigation accuracy as a minimum reporting standard.

Motion artifact correction in free-breathing abdominal MRI using overlapping partial samples to recover image deformations.

This article presents a method to reconstruct liver MRI data acquired continuously during free breathing, without any external sensor or navigator measurements. When the deformations associated with k-space data are known, generalized matrix inversion reconstruction has been shown to be effective in reducing the ghosting and blurring artifacts of motion. This article describes a novel method to obtain these nonrigid deformations. A breathing model is built from a fast dynamic series: low spatial resolution images are registered and their deformations parameterized by overall superior-inferior displacement. The correct deformation for each subset of the subsequent imaging data is then found by comparing a few lines of k-space with the equivalent lines from a deformed reference image while varying the deformation over the model parameter. This procedure is known as image deformation recovery using overlapping partial samples (iDROPS). Simulations using 10 rapid dynamic studies from volunteers showed the average error in iDROPS-derived deformations within the liver to be 1.43 mm. A further four volunteers were imaged at higher spatial resolution. The complete reconstruction process using data from throughout several breathing cycles was shown to reduce blurring and ghosting in the liver. Retrospective respiratory gating was also demonstrated using the iDROPS parameterization.

3-D Pose Estimation of Articulated Instruments in Robotic Minimally Invasive Surgery.

Estimating the 3-D pose of instruments is an important part of robotic minimally invasive surgery for automation of basic procedures as well as providing safety features, such as virtual fixtures. Image-based methods of 3-D pose estimation provide a non-invasive low cost solution compared with methods that incorporate external tracking systems. In this paper, we extend our recent work in estimating rigid 3-D pose with silhouette and optical flow-based features to incorporate the articulated degrees-of-freedom (DOFs) of robotic instruments within a gradient-based optimization framework. Validation of the technique is provided with a calibrated ex-vivo study from the da Vinci Research Kit (DVRK) robotic system, where we perform quantitative analysis on the errors each DOF of our tracker. Additionally, we perform several detailed comparisons with recently published techniques that combine visual methods with kinematic data acquired from the joint encoders. Our experiments demonstrate that our method is competitively accurate while relying solely on image data.

Tracking 'differential organ motion' with a 'breathing' multileaf collimator: magnitude of problem assessed using 4D CT data and a motion-compensation strategy.

Intrafraction tumour (e.g. lung) motion due to breathing can, in principle, be compensated for by applying identical breathing motions to the leaves of a multileaf collimator (MLC) as intensity-modulated radiation therapy is delivered by the dynamic MLC (DMLC) technique. A difficulty arising, however, is that irradiated voxels, which are in line with a bixel at one breathing phase (at which the treatment plan has been made), may move such that they cease to be in line with that breathing bixel at another phase. This is the phenomenon of differential voxel motion and existing tracking solutions have ignored this very real problem. There is absolutely no tracking solution to the problem of compensating for differential voxel motion. However, there is a strategy that can be applied in which the leaf breathing is determined to minimize the geometrical mismatch in a least-squares sense in irradiating differentially-moving voxels. A 1D formulation in very restricted circumstances is already in the literature and has been applied to some model breathing situations which can be studied analytically. These are, however, highly artificial. This paper presents the general 2D formulation of the problem including allowing different importance factors to be applied to planning target volume and organ at risk (or most generally) each voxel. The strategy also extends the literature strategy to the situation where the number of voxels connecting to a bixel is a variable. Additionally the phenomenon of 'cross-leaf-track/channel' voxel motion is formally addressed. The general equations are presented and analytic results are given for some 1D, artificially contrived, motions based on the Lujan equations of breathing motion. Further to this, 3D clinical voxel motion data have been extracted from 4D CT measurements to both assess the magnitude of the problem of 2D motion perpendicular to the beam-delivery axis in clinical practice and also to find the 2D optimum breathing-leaf strategy. Issues relating to the practical calculation of the strategy, including effects on leaf velocity and effects of different spatial-sampling frequencies, have been investigated, and unattenuated-fluence maps have been produced showing the effects of the differential motion and tracking. It was discovered that large distances between adjacent leaf-ends could cause the tracking to fail when there was tissue motion across the leaf channels. To overcome this problem the use of 'synchronized' leaf trajectories, which ensure that adjacent leaf-ends are always close enough to each other to facilitate tracking, has also been investigated.

MRI-based measurements of respiratory motion variability and assessment of imaging strategies for radiotherapy planning.

Respiratory organ motion has a significant impact on the planning and delivery of radiotherapy (RT) treatment for lung cancer. Currently widespread techniques, such as 4D-computed tomography (4DCT), cannot be used to measure variability of this motion from one cycle to the next. In this paper, we describe the use of fast magnetic resonance imaging (MRI) techniques to investigate the intra- and inter-cycle reproducibility of respiratory motion and also to estimate the level of errors that may be introduced into treatment delivery by using various breath-hold imaging strategies during lung RT planning. A reference model of respiratory motion is formed to enable comparison of different breathing cycles at any arbitrary position in the respiratory cycle. This is constructed by using free-breathing images from the inhale phase of a single breathing cycle, then co-registering the images, and thereby tracking landmarks. This reference model is then compared to alternative models constructed from images acquired during the exhale phase of the same cycle and the inhale phase of a subsequent cycle, to assess intra- and inter-cycle variability ('hysteresis' and 'reproducibility') of organ motion. The reference model is also compared to a series of models formed from breath-hold data at exhale and inhale. Evaluation of these models is carried out on data from ten healthy volunteers and five lung cancer patients. Free-breathing models show good levels of intra- and inter-cycle reproducibility across the tidal breathing range. Mean intra-cycle errors in the position of organ surface landmarks of 1.5(1.4)-3.5(3.3) mm for volunteers and 2.8(1.8)-5.2(5.2) mm for patients. Equivalent measures of inter-cycle variability across this range are 1.7(1.0)-3.9(3.3) mm for volunteers and 2.8(1.8)-3.3(2.2) mm for patients. As expected, models based on breath-hold sequences do not represent normal tidal motion as well as those based on free-breathing data, with mean errors of 4.4(2.2)-7.7(3.9) mm for volunteers and 10.1(6.1)-12.5(6.3) mm for patients. Errors are generally larger still when using a single breath-hold image at either exhale or inhale to represent the lung. This indicates that account should be taken of intra- and inter-cycle respiratory motion variability and that breath-hold-based methods of obtaining data for RT planning may potentially introduce large errors. This approach to analysis of motion and variability has potential to inform decisions about treatment margins and optimize RT planning.

Cadaver validation of intensity-based ultrasound to CT registration.

A method is presented for the rigid registration of tracked B-mode ultrasound images to a CT volume of a femur and pelvis. This registration can allow tracked surgical instruments to be aligned with the CT image or an associated preoperative plan. Our method is fully automatic and requires no manual segmentation of either the ultrasound images or the CT volume. The parameter which is directly related to the speed of sound through tissue has also been included in the registration optimisation process. Experiments have been carried out on six cadaveric femurs and three cadaveric pelves. Registration results were compared with a "gold standard" registration acquired using bone implanted fiducial markers. Results show the registration method to be accurate, on average, to 1.6 mm root-mean-square target registration error.

Anisotropic multi-scale fluid registration: evaluation in magnetic resonance breast imaging.

Registration using models of compressible viscous fluids has not found the general application of some other techniques (e.g., free-form-deformation (FFD)) despite its ability to model large diffeomorphic deformations. We report on a multi-resolution fluid registration algorithm which improves on previous work by (a) directly solving the Navier-Stokes equation at the resolution of the images, (b) accommodating image sampling anisotropy using semi-coarsening and implicit smoothing in a full multi-grid (FMG) solver and (c) exploiting the inherent multi-resolution nature of FMG to implement a multi-scale approach. Evaluation is on five magnetic resonance (MR) breast images subject to six biomechanical deformation fields over 11 multi-resolution schemes. Quantitative assessment is by tissue overlaps and target registration errors and by registering using the known correspondences rather than image features to validate the fluid model. Context is given by comparison with a validated FFD algorithm and by application to images of volunteers subjected to large applied deformation. The results show that fluid registration of 3D breast MR images to sub-voxel accuracy is possible in minutes on a 1.6 GHz Linux-based Athlon processor with coarse solutions obtainable in a few tens of seconds. Accuracy and computation time are comparable to FFD techniques validated for this application.

Tissue deformation and shape models in image-guided interventions: a discussion paper.

This paper promotes the concept of active models in image-guided interventions. We outline the limitations of the rigid body assumption in image-guided interventions and describe how intraoperative imaging provides a rich source of information on spatial location of anatomical structures and therapy devices, allowing a preoperative plan to be updated during an intervention. Soft tissue deformation and variation from an atlas to a particular individual can both be determined using non-rigid registration. Established methods using free-form deformations have a very large number of degrees of freedom. Three examples of deformable models--motion models, biomechanical models and statistical shape models--are used to illustrate how prior information can be used to restrict the number of degrees of freedom of the registration algorithm and thus provide active models for image-guided interventions. We provide preliminary results from applications for each type of model.

Detection and modelling of contacts in explicit finite-element simulation of soft tissue biomechanics.

PURPOSE: Realistic modelling of soft tissue biomechanics and mechanical interactions between tissues is an important part of biomechanically-informed surgical image-guidance and surgical simulation. This submission details a contact-modelling pipeline suitable for implementation in explicit matrix-free FEM solvers. While these FEM algorithms have been shown to be very suitable for simulation of soft tissue biomechanics and successfully used in a number of image-guidance systems, contact modelling specifically for these solvers is rarely addressed, partly because the typically large number of time steps required with this class of FEM solvers has led to a perception of them being a poor choice for simulations requiring complex contact modelling. METHODS: The presented algorithm is capable of handling most scenarios typically encountered in image-guidance. The contact forces are computed with an evolution of the Lagrange-multiplier method first used by Taylor and Flanagan in PRONTO 3D extended with spatio-temporal smoothing heuristics for improved stability and edge-edge collision handling, and a new friction model. For contact search, a bounding-volume hierarchy (BVH) is employed, which is capable of identifying self-collisions by means of the surface-normal bounding cone of Volino and Magnenat-Thalmann, in turn computed with a novel formula. The BVH is further optimised for the small time steps by reducing the number of bounding-volume refittings between iterations through identification of regions with mostly rigid motion and negligible deformation. Further optimisation is achieved by integrating the self-collision criterion in the BVH creation and updating algorithms. RESULTS: The effectiveness of the algorithm is demonstrated on a number of artificial test cases and meshes derived from medical image data. It is shown that the proposed algorithm reduces the cost of BVH refitting to the point where it becomes a negligible part of the overall computation time of the simulation. It is also shown that the proposed surface-normal cone computation formula leads to about 40 % fewer BVH subtrees that must be checked for self-collisions compared with the widely used method of Provot. The proposed contact-force formulation and friction model are evaluated on artificial test cases that allow for a comparison with a ground truth. The quality of the proposed contact forces is assessed in terms of trajectories and energy conservation; a [Formula: see text]0.4 % drop off in total energy and highly plausible trajectories are found in the experiments. The friction model is evaluated through a benchmark problem with an analytical solution and a maximum displacement error of 8.2 %, and excellent agreement in terms of the stick/slip boundary is found. Finally, we show with realistic image-guidance examples that the entire contact-modelling pipeline can be executed within a timeframe that is of the same order of magnitude as that required for standard FEM computations.

Validation of volume blood flow measurements using three-dimensional distance-concentration functions derived from digital x-ray angiograms.

RATIONALE AND OBJECTIVES: The authors present phantom validation of a method for computing pulsatile flow waveforms in arterial vessels from high-frame-rate biplane x-ray angiograms. METHODS: The three-dimensional course of a blood vessel is constructed from biplane digital x-ray angiograms. A parametric image of contrast mass versus time and true three-dimensional path length is generated. Adjacent contrast mass-distance profiles are matched to compute instantaneous velocity, which is multiplied by cross-sectional area to yield volume flow. An electromagnetic flowmeter was used to validate flow estimates in a phantom consisting of 150-mm tubes 3, 4, and 6 mm in diameter, orientated 15 degrees, 30 degrees, and 35 degrees to the imaging plane, with flow rates and waveforms expected in vivo. RESULTS: Mean and peak flows were accurate to within 9% and 10%, respectively, for velocities of less than 1 meter/second at a frame rate of 25 frames per second. CONCLUSIONS: A practical method for computing highly pulsatile flow waveforms in vivo in tortuous vessels is presented.

Automated three-dimensional registration of magnetic resonance and positron emission tomography brain images by multiresolution optimization of voxel similarity measures.

Approaches using measures of voxel intensity similarity are showing promise in fully automating magnetic resonance (MR) and positron emission tomography (PET) image registration in the head, without requiring extraction and identification of corresponding structures. In this paper a method of multiresolution optimization of these measures is described and five alternative measures are compared: cross correlation, minimization of corresponding PET intensity variation, moments of the distribution of values in the intensity feature space, entropy of the intensity feature space and mutual information. Their ability to recover registration is examined for ten clinically acquired image pairs with respect to the size of initial misregistration, the precision of the final result, and the accuracy assessed by visual inspection. The mutual information measure proved the most robust to initial starting estimate, successfully registering 98.8% of 900 trial misregistrations. Success is defined as providing a visually acceptable solution to a trained observer. A high resolution search (1/16 mm step size) of 30 trial misregistrations showed that optimization using the mutual information measure provided solutions with 0.13 mm, 0.11 mm and 0.17 mm standard deviations in the three Cartesian axes of the translation vector and 0.2 degree, 0.3 degree and 0.2 degree standard deviations for rotations about the three axes. The algorithm takes between 4 and 8 minutes to run on a typical workstation, including visual inspection of the result.

Validation of a two- to three-dimensional registration algorithm for aligning preoperative CT images and intraoperative fluoroscopy images.

We present a validation of an intensity based two- to three-dimensional image registration algorithm. The algorithm can register a CT volume to a single-plane fluoroscopy image. Four routinely acquired clinical data sets from patients who underwent endovascular treatment for an abdominal aortic aneurysm were used. Each data set was comprised of two intraoperative fluoroscopy images and a preoperative CT image. Regions of interest (ROI) were drawn around each vertebra in the CT and fluoroscopy images. Each CT image ROI was individually registered to the corresponding ROI in the fluoroscopy images. A cross validation approach was used to obtain a measure of registration consistency. Spinal movement between the preoperative and intraoperative scene was accounted for by using two fluoroscopy images. The consistency and robustness of the algorithm when using two similarity measures, pattern intensity and gradient difference, was investigated. Both similarity measures produced similar results. The consistency values were rotational errors below 0.74 degree and in-plane translational errors below 0.90 mm. These errors approximately relate to a two-dimensional projection error of 1.3 mm. The failure rate was less than 8.3% for three of the four data sets. However, for one of the data sets a much larger failure rate (28.5%) occurred.

Combined and three-dimensional rendered multimodal data for planning cranial base surgery: a prospective evaluation.

Magnetic resonance (MR), X-ray computed tomography (CT), and angiographic images best depict soft tissue, bone, and blood vessels respectively. No one on its own is sufficient in the preoperative assessment of cranial base lesions. We have developed and evaluated a computational technique for the three-dimensional (3D) combination and display of multimodality images for planning cranial base surgery. This evaluation was prospective and performed in such a way that the results could be quantified. Eight patients (three acoustic neuromas, four subfrontal and suprasellar meningiomas, and one petrous apex meningioma) underwent MR, CT, and MR angiographic investigations. These images were registered with anatomical landmarks rather than an external frame. Two techniques were used to display the resulting combined images: multiple slices in which bone from CT was overlaid on soft tissue from registered MR and pseudo-3D-rendered movie sequences showing bone from CT, lesions and optic nerves from MR, and blood vessels from MR angiography. The advantages of the combined displays compared with those of conventional methods of viewing were assessed prospectively by the operating surgeon and by an independent surgeon, and the results were compared with operative findings. The preoperative assessment showed a significant improvement (P < 0.05, sign test) in the depiction of both individual structures (lesion and bone from overlaid slices and lesion and vasculature from 3D-rendered displays) and structural relationships (tumor-bone relationships from overlaid slices and of tumor-vasculature relationships from 3D-rendered displays). The operative findings indicated that a more accurate interpretation of this information was possible from the combined images.

Sources of error in comparing functional magnetic resonance imaging and invasive electrophysiological recordings.

OBJECT: Several authors have recently reported studies in which they aim to validate functional magnetic resonance (fMR) imaging against the accepted gold standard of invasive electrophysiological monitoring. The authors have conducted a similar study, and in this paper they identify and quantify two characteristics of these data that can make such a comparison problematic. METHODS: Eight patients in whom surgery for epilepsy was performed and five healthy volunteers underwent fMR imaging to localize the part of the sensorimotor cortex responsible for hand movement. In the patient group subdural electrode mats were subsequently implanted to identify eloquent regions of the brain and the epileptogenic zone. The fMR imaging data were processed to correct for motion during the study and then registered with a postimplantation computerized tomography (CT) scan on which the electrodes were visible. The motion during imaging in the two groups studied, and the deformation of the brain between the preoperative images and postoperative scans were measured. The patients who underwent epilepsy surgery moved significantly more during fMR imaging experiments than healthy volunteers performing the same motor task. This motion had a particularly increased out-of-plane component and was significantly more correlated with the stimulus than in the volunteers. This motion was especially increased when the patients were performing a task on the side affected by the lesion. The additional motion is hard to correct and substantially degrades the quality of the resulting fMR images, making it a much less reliable technique for use in these patients than in others. Also, the authors found that after electrode implantation, the brain surface can shift more than 10 mm relative to the skull compared with its preoperative location, substantially degrading the accuracy of the comparison of electrophysiological measurements made in the deformed brain and fMR studies obtained preoperatively. CONCLUSIONS: These two findings indicate that studies of this sort are currently of limited use for validating fMR imaging and should be interpreted with care. Additional image analysis research is necessary to solve the problems caused by patients' motion and brain deformation.

A comparison of similarity measures for use in 2-D-3-D medical image registration.

A comparison of six similarity measures for use in intensity-based two-dimensional-three-dimensional (2-D-3-D) image registration is presented. The accuracy of the similarity measures are compared to a "gold-standard" registration which has been accurately calculated using fiducial markers. The similarity measures are used to register a computed tomography (CT) scan of a spine phantom to a fluoroscopy image of the phantom. The registration is carried out within a region-of-interest in the fluoroscopy image which is user defined to contain a single vertebra. Many of the problems involved in this type of registration are caused by features which were not modeled by a phantom image alone. More realistic "gold-standard" data sets were simulated using the phantom image with clinical image features overlaid. Results show that the introduction of soft-tissue structures and interventional instruments into the phantom image can have a large effect on the performance of some similarity measures previously applied to 2-D-3-D image registration. Two measures were able to register accurately and robustly even when soft-tissue structures and interventional instruments were present as differences between the images. These measures were pattern intensity and gradient difference. Their registration accuracy, for all the rigid-body parameters except for the source to film translation, was within a root-mean-square (rms) error of 0.54 mm or degrees to the "gold-standard" values. No failures occurred while registering using these measures.

Nonrigid registration using free-form deformations: application to breast MR images.

In this paper we present a new approach for the nonrigid registration of contrast-enhanced breast MRI. A hierarchical transformation model of the motion of the breast has been developed. The global motion of the breast is modeled by an affine transformation while the local breast motion is described by a free-form deformation (FFD) based on B-splines. Normalized mutual information is used as a voxel-based similarity measure which is insensitive to intensity changes as a result of the contrast enhancement. Registration is achieved by minimizing a cost function, which represents a combination of the cost associated with the smoothness of the transformation and the cost associated with the image similarity. The algorithm has been applied to the fully automated registration of three-dimensional (3-D) breast MRI in volunteers and patients. In particular, we have compared the results of the proposed nonrigid registration algorithm to those obtained using rigid and affine registration techniques. The results clearly indicate that the nonrigid registration algorithm is much better able to recover the motion and deformation of the breast than rigid or affine registration algorithms.

Measures of folding applied to the development of the human fetal brain.

Previous work has suggested the existence of differences between the cerebral cortex of normal individuals, and those of patients with diseases such as epilepsy and schizophrenia. These shape abnormalities may be of developmental origin. Improved shape measures could provide useful tools for neuroscience research and patient diagnosis. We consider the theoretically desirable properties of measures of brain shape. We have implemented seven measures, three from the neuroscience literature, and four new to this field. Three of the measures are zero-order and four are second-order with respect to the surface. We validate the measures using simple geometrical shapes, and a collection of magnetic resonance scans of ten histologically normal ex vivo fetal brains with gestational ages from 19-42 weeks. We then apply the measures to MR scans from two histologically abnormal ex vivo brains. We demonstrate that our implementation of the measures is sensitive to anatomical variability rather than to the discreteness of the image data. All the measures were sensitive to changes in shape during fetal development. Several of the measures could distinguish between the normal and abnormal fetal brains. We propose a multivariate approach to studying the shape of the cerebral cortex, in which both zero-order and second-order measures are used to quantify folding.

Voxel similarity measures for 3-D serial MR brain image registration.

We have evaluated eight different similarity measures used for rigid body registration of serial magnetic resonance (MR) brain scans. To assess their accuracy we used 33 clinical three-dimensional (3-D) serial MR images, with deformable extradural tissue excluded by manual segmentation and simulated 3-D MR images with added intensity distortion. For each measure we determined the consistency of registration transformations for both sets of segmented and unsegmented data. We have shown that of the eight measures tested, the ones based on joint entropy produced the best consistency. In particular, these measures seemed to be least sensitive to the presence of extradural tissue. For these data the difference in accuracy of these joint entropy measures, with or without brain segmentation, was within the threshold of visually detectable change in the difference images.

Measurement and analysis of brain deformation during neurosurgery.

Recent studies have shown that the surface of the brain is deformed by up to 20 mm after the skull is opened during neurosurgery, which could lead to substantial error in commercial image-guided surgery systems. We quantitatively analyze the intraoperative brain deformation of 24 subjects to investigate whether simple rules can describe or predict the deformation. Interventional magnetic resonance images acquired at the start and end of the procedure are registered nonrigidly to obtain deformation values throughout the brain. Deformation patterns are investigated quantitatively with respect to the location and magnitude of deformation, and to the distribution and principal direction of the displacements. We also measure the volume change of the lateral ventricles by manual segmentation. Our study indicates that brain shift occurs predominantly in the hemisphere ipsi-lateral to the craniotomy, and that there is more brain deformation during resection procedures than during biopsy or functional procedures. However, the brain deformation patterns are extremely complex in this group of subjects. This paper quantitatively demonstrates that brain deformation occurs not only at the surface, but also in deeper brain structure, and that the principal direction of displacement does not always correspond with the direction of gravity. Therefore, simple computational algorithms that utilize limited intraoperative information (e.g., brain surface shift) will not always accurately predict brain deformation at the lesion.