Image registration of pre-procedural MRI and intra-procedural CT images to aid CT-guided percutaneous cryoablation of renal tumors

SlicerCBM: automatic framework for biomechanical analysis of the brain

PURPOSE

Brain shift that occurs during neurosurgery disturbs the brain's anatomy. Prediction of the brain shift is essential for accurate localisation of the surgical target. Biomechanical models have been envisaged as a possible tool for such predictions. In this study, we created a framework to automate the workflow for predicting intra-operative brain deformations.

METHODS

We created our framework by uniquely combining our meshless total Lagrangian explicit dynamics (MTLED) algorithm for computing soft tissue deformations, open-source software libraries and built-in functions within 3D Slicer, an open-source software package widely used for medical research. Our framework generates the biomechanical brain model from the pre-operative MRI, computes brain deformation using MTLED and outputs results in the form of predicted warped intra-operative MRI.

RESULTS

Our framework is used to solve three different neurosurgical brain shift scenarios: craniotomy, tumour resection and electrode placement. We evaluated our framework using nine patients. The average time to construct a patient-specific brain biomechanical model was 3 min, and that to compute deformations ranged from 13 to 23 min. We performed a qualitative evaluation by comparing our predicted intra-operative MRI with the actual intra-operative MRI. For quantitative evaluation, we computed Hausdorff distances between predicted and actual intra-operative ventricle surfaces. For patients with craniotomy and tumour resection, approximately 95% of the nodes on the ventricle surfaces are within two times the original in-plane resolution of the actual surface determined from the intra-operative MRI.

CONCLUSION

Our framework provides a broader application of existing solution methods not only in research but also in clinics. We successfully demonstrated the application of our framework by predicting intra-operative deformations in nine patients undergoing neurosurgical procedures.

Developing and testing a robotic MRI/CT fusion biopsy technique using a purpose-built interventional phantom

BACKGROUND

Magnetic resonance imaging (MRI) can be used to target tumour components in biopsy procedures, while the ability to precisely correlate histology and MRI signal is crucial for imaging biomarker validation. Robotic MRI/computed tomography (CT) fusion biopsy offers the potential for this without in-gantry biopsy, although requires development.

METHODS

Test-retest T1 and T2 relaxation times, attenuation (Hounsfield units, HU), and biopsy core quality were prospectively assessed (January-December 2021) in a range of gelatin, agar, and mixed gelatin/agar solutions of differing concentrations on days 1 and 8 after manufacture. Suitable materials were chosen, and four biopsy phantoms were constructed with twelve spherical 1-3-cm diameter targets visible on MRI, but not on CT. A technical pipeline was developed, and intraoperator and interoperator reliability was tested in four operators performing a total of 96 biopsies. Statistical analysis included T1, T2, and HU repeatability using Bland-Altman analysis, Dice similarity coefficient (DSC), and intraoperator and interoperator reliability.

RESULTS

T1, T2, and HU repeatability had 95% limits-of-agreement of 8.3%, 3.4%, and 17.9%, respectively. The phantom was highly reproducible, with DSC of 0.93 versus 0.92 for scanning the same or two different phantoms, respectively. Hit rate was 100% (96/96 targets), and all operators performed robotic biopsies using a single volumetric acquisition. The fastest procedure time was 32 min for all 12 targets.

CONCLUSIONS

A reproducible biopsy phantom was developed, validated, and used to test robotic MRI/CT-fusion biopsy. The technique was highly accurate, reliable, and achievable in clinically acceptable timescales meaning it is suitable for clinical application.

Synthesis of magnetic resonance images from computed tomography data using convolutional neural network with contextual loss function

BACKGROUND

Magnetic resonance imaging (MRI) images synthesized from computed tomography (CT) data can provide more detailed information on pathological structures than that of CT data alone; thus, the synthesis of MRI has received increased attention especially in medical scenarios where only CT images are available. A novel convolutional neural network (CNN) combined with a contextual loss function was proposed for synthesis of T1- and T2-weighted images (T1WI and T2WI) from CT data.

METHODS

A total of 5,053 and 5,081 slices of T1WI and T2WI, respectively were selected for the dataset of CT and MRI image pairs. Affine registration, image denoising, and contrast enhancement were done on the aforementioned multi-modality medical image dataset comprising T1WI, T2WI, and CT images of the brain. A deep CNN was then proposed by modifying the ResNet structure to constitute the encoder and decoder of U-Net, called double ResNet-U-Net (DRUNet). Three different loss functions were utilized to optimize the parameters of the proposed models: mean squared error (MSE) loss, binary crossentropy (BCE) loss, and contextual loss. Statistical analysis of the independent-sample t-test was conducted by comparing DRUNets with different loss functions and different network layers.

RESULTS

DRUNet-101 with contextual loss yielded higher values of peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), and Tenengrad function (i.e., 34.25±2.06, 0.97±0.03, and 17.03±2.75 for T1WI and 33.50±1.08, 0.98±0.05, and 19.76±3.54 for T2WI respectively). The results were statistically significant at P<0.001 with a narrow confidence interval of difference, indicating the superiority of DRUNet-101 with contextual loss. In addition, both image zooming and difference maps presented for the final synthetic MR images visually reflected the robustness of DRUNet-101 with contextual loss. The visualization of convolution filters and feature maps showed that the proposed model can generate synthetic MR images with high-frequency information.

CONCLUSIONS

The results demonstrated that DRUNet-101 with contextual loss function provided better high-frequency information in synthetic MR images compared with the other two functions. The proposed DRUNet model has a distinct advantage over previous models in terms of PSNR, SSIM, and Tenengrad score. Overall, DRUNet-101 with contextual loss is recommended for synthesizing MR images from CT scans.

Computer simulation of tumour resection-induced brain deformation by a meshless approach

Tumour resection requires precise planning and navigation to maximise tumour removal while simultaneously protecting nearby healthy tissues. Neurosurgeons need to know the location of the remaining tumour after partial tumour removal before continuing with the resection. Our approach to the problem uses biomechanical modelling and computer simulation to compute the brain deformations after the tumour is resected. In this study, we use meshless Total Lagrangian explicit dynamics as the solver. The problem geometry is extracted from the patient-specific magnetic resonance imaging (MRI) data and includes the parenchyma, tumour, cerebrospinal fluid and skull. The appropriate non-linear material formulation is used. Loading is performed by imposing intra-operative conditions of gravity and reaction forces between the tumour and surrounding healthy parenchyma tissues. A finite frictionless sliding contact is enforced between the skull (rigid) and parenchyma. The meshless simulation results are compared to intra-operative MRI sections. We also calculate Hausdorff distances between the computed deformed surfaces (ventricles and tumour cavities) and surfaces observed intra-operatively. Over 80% of points on the ventricle surface and 95% of points on the tumour cavity surface were successfully registered (results within the limits of two times the original in-plane resolution of the intra-operative image). Computed results demonstrate the potential for our method in estimating the tissue deformation and tumour boundary during the resection.

Radiation dose reduction in CT-guided cryoablation of renal tumors

PURPOSE

We aimed to evaluate the effect on the radiation dose to the patient by reducing the tube current during the placement of the ablation needles (reduced dose group) compared with the patient doses delivered when scanning at the standard fully diagnostic level (full dose group) in computed tomography (CT)-guided percutaneous cryoablation.

METHODS

We conducted a retrospective study of 103 patients undergoing cryoablation in a tertiary cancer center. Overall, 62 patients were scanned with standard exposure parameters (full dose group) set on a 64-slice multidetector CT scanner, while 41 patients were scanned on a reduced dose protocol. Dose levels were retrieved from the hospital picture and archiving communication system including the volumetric CT dose index (CTDIvol), total dose length product (DLP), length of cryoablation procedure, number of cryoablation needles and patient size. Wilcoxon Mann-Whitney (rank-sum) tests were used to compare the median DLP, CTDIvol and skin dose between the two groups.

RESULTS

Median total DLP for the full dose group was 6025 mGy•cm (1909-13353 mGy•cm) compared with 3391 mGy•cm (1683-6820 mGy•cm) for the reduced dose group. The reduced dose group had a 44% reduction in total DLP and 42% reduction in total CTDIvol (p < 0.001). The estimated skin doses were 384 mGy for the full dose group and 224 mGy for the reduced dose group (42% reduction) (p < 0.001). At 12-month follow-up, the technical success for the full dose (n=62) was 97% with 2 patients requiring a further cryoablation treatment for residual tumor. The technical success for the reduced dose group (n=41) was 100%.

CONCLUSION

CT dose reduction technique during image-guided cryoablation treatment of renal tumors can achieve significant radiation dose reduction whilst maintaining sufficient image quality.

Three-dimensional quantitative assessment of ablation margins based on registration of pre- and post-procedural MRI and distance map

PURPOSE

Contrast-enhanced MR images are widely used to confirm the adequacy of ablation margin after liver ablation for early prediction of local recurrence. However, quantitative assessment of the ablation margin by comparing pre- and post-procedural images remains challenging. We developed and tested a novel method for three-dimensional quantitative assessment of ablation margin based on non-rigid image registration and 3D distance map.

METHODS

Our method was tested with pre- and post-procedural MR images acquired in 21 patients who underwent image-guided percutaneous liver ablation. The two images were co-registered using non-rigid intensity-based registration. After the tumor and ablation volumes were segmented, target volume coverage, percent of tumor coverage, and Dice similarity coefficient were calculated as metrics representing overall adequacy of ablation. In addition, 3D distance map around the tumor was computed and superimposed on the ablation volume to identify the area with insufficient margins. For patients with local recurrences, the follow-up images were registered to the post-procedural image. Three-dimensional minimum distance between the recurrence and the areas with insufficient margins was quantified.

RESULTS

The percent tumor coverage for all nonrecurrent cases was 100 %. Five cases had tumor recurrences, and the 3D distance map revealed insufficient tumor coverage or a 0-mm margin. It also showed that two recurrences were remote to the insufficient margin.

CONCLUSIONS

Non-rigid registration and 3D distance map allow us to quantitatively evaluate the adequacy of the ablation margin after percutaneous liver ablation. The method may be useful to predict local recurrences immediately following ablation procedure.

Patient-specific biomechanical model as whole-body CT image registration tool

Whole-body computed tomography (CT) image registration is important for cancer diagnosis, therapy planning and treatment. Such registration requires accounting for large differences between source and target images caused by deformations of soft organs/tissues and articulated motion of skeletal structures. The registration algorithms relying solely on image processing methods exhibit deficiencies in accounting for such deformations and motion. We propose to predict the deformations and movements of body organs/tissues and skeletal structures for whole-body CT image registration using patient-specific non-linear biomechanical modelling. Unlike the conventional biomechanical modelling, our approach for building the biomechanical models does not require time-consuming segmentation of CT scans to divide the whole body into non-overlapping constituents with different material properties. Instead, a Fuzzy C-Means (FCM) algorithm is used for tissue classification to assign the constitutive properties automatically at integration points of the computation grid. We use only very simple segmentation of the spine when determining vertebrae displacements to define loading for biomechanical models. We demonstrate the feasibility and accuracy of our approach on CT images of seven patients suffering from cancer and aortic disease. The results confirm that accurate whole-body CT image registration can be achieved using a patient-specific non-linear biomechanical model constructed without time-consuming segmentation of the whole-body images.

Development of a software for quantitative evaluation radiotherapy target and organ-at-risk segmentation comparison

Modern radiotherapy requires accurate region of interest (ROI) inputs for plan optimization and delivery. Target delineation, however, remains operator-dependent and potentially serves as a major source of treatment delivery error. In order to optimize this critical, yet observer-driven process, a flexible web-based platform for individual and cooperative target delineation analysis and instruction was developed in order to meet the following unmet needs: (1) an open-source/open-access platform for automated/semiautomated quantitative interobserver and intraobserver ROI analysis and comparison, (2) a real-time interface for radiation oncology trainee online self-education in ROI definition, and (3) a source for pilot data to develop and validate quality metrics for institutional and cooperative group quality assurance efforts. The resultant software, Target Contour Testing/Instructional Computer Software (TaCTICS), developed using Ruby on Rails, has since been implemented and proven flexible, feasible, and useful in several distinct analytical and research applications.

Optimisation of the coagulation zone for thermal ablation procedures: a theoretical approach with considerations for practical use

PURPOSE

This paper outlines a theoretical approach for optimisation of the coagulation zone for thermal ablation procedures and considerations for its practical application.

METHODS

The theoretical approach is outlined in the Cartesian coordinate system. Considerations for practical application are implemented. The optimised coagulation zone is defined as the bare coverage of tumour mass plus a safety margin. The eccentricity of coagulation centre (ECC) is defined as the distance between the coagulation centre and the tumour centre. The direction of the applicator shaft is determined based on the x-axis direction. The tumour centre and coagulation centre are defined within the x/y-plane. The distance between coagulation margin (applicator tip) and tumour margin is called parallel offset (PAO).

RESULTS

For spherical coagulation shapes, a linear relationship exists between optimised coagulation diameter and ECC. An exponential relationship exists between optimised coagulation volume and ECC. A complex relationship was found between PAO and determinants of ECC, which are ex and ey. PAO is an extremely important parameter, which allows for determination of the optimal applicator tip position in relation to the tumour margin. It can be calculated in such a manner that the optimised coagulation zone is minimised by neutralising dislocation of the coagulation centre in applicator shaft direction. The latter can be realised by withdrawing or further inserting the applicator shaft.

CONCLUSIONS

The presented concept can be used to optimise the extent of the coagulation zone for thermal ablation procedures after positioning of the applicator. Its inherent advantage is the simple adjustment of the applicator shaft, which obviates the need for a repuncture.

Beyond finite elements: a comprehensive, patient-specific neurosurgical simulation utilizing a meshless method

To be useful in clinical (surgical) simulations, a method must use fully nonlinear (both geometric and material) formulations to deal with large (finite) deformations of tissues. The method must produce meaningful results in a short time on consumer hardware and not require significant manual work while discretizing the problem domain. In this paper, we showcase the Meshless Total Lagrangian Explicit Dynamics Method (MTLED) which meets these requirements, and use it for computing brain deformations during surgery. The problem geometry is based on patient-specific MRI data and includes the parenchyma, tumor, ventricles and skull. Nodes are distributed automatically through the domain rendering the normally difficult problem of creating a patient-specific computational grid a trivial exercise. Integration is performed over a simple, regular background grid which does not need to conform to the geometry boundaries. Appropriate nonlinear material formulation is used. Loading is performed by displacing the parenchyma surface nodes near the craniotomy and a finite frictionless sliding contact is enforced between the skull (rigid) and parenchyma. The meshless simulation results are compared to both intraoperative MRIs and Finite Element Analysis results for multiple 2D sections. We also calculate Hausdorff distances between the computed deformed surfaces of the ventricles and those observed intraoperatively. The difference between previously validated Finite Element results and the meshless results presented here is less than 0.2mm. The results are within the limits of neurosurgical and imaging equipment accuracy (~1 mm) and demonstrate the method's ability to fulfill all of the important requirements for surgical simulation.

Patient-specific non-linear finite element modelling for predicting soft organ deformation in real-time: application to non-rigid neuroimage registration

Long computation times of non-linear (i.e. accounting for geometric and material non-linearity) biomechanical models have been regarded as one of the key factors preventing application of such models in predicting organ deformation for image-guided surgery. This contribution presents real-time patient-specific computation of the deformation field within the brain for six cases of brain shift induced by craniotomy (i.e. surgical opening of the skull) using specialised non-linear finite element procedures implemented on a graphics processing unit (GPU). In contrast to commercial finite element codes that rely on an updated Lagrangian formulation and implicit integration in time domain for steady state solutions, our procedures utilise the total Lagrangian formulation with explicit time stepping and dynamic relaxation. We used patient-specific finite element meshes consisting of hexahedral and non-locking tetrahedral elements, together with realistic material properties for the brain tissue and appropriate contact conditions at the boundaries. The loading was defined by prescribing deformations on the brain surface under the craniotomy. Application of the computed deformation fields to register (i.e. align) the preoperative and intraoperative images indicated that the models very accurately predict the intraoperative deformations within the brain. For each case, computing the brain deformation field took less than 4 s using an NVIDIA Tesla C870 GPU, which is two orders of magnitude reduction in computation time in comparison to our previous study in which the brain deformation was predicted using a commercial finite element solver executed on a personal computer.