Detection Of Soft-Tissue Deformation Based On Multi-modality Imaging

Traditional image-guided navigation systems are mainly based on theestablishment of a3D model of patient’s anatomical structure that using3Dreconstruction and visualization of pre-operative images. For soft tissuenavigation, due to organ deformation caused by respiration, body motion andintervention of operation equipment, etc, the model established pre-operativelycould not reflect the change of underlying anatomical structure during theoperation and result in inaccurate navigation, even failure of navigation.Therefore, the accurate tracking of soft tissue deformation has significantresearch and clinical value for accurate navigation of chest/abdominal surgeriesand interventional therapies.In recent years, intraoperative ultrasound has been widely used inchest/abdominal surgeries and interventional therapies. However, ultrasoundimages may not reflect the surrounding tissue comprehensively due to its2Dand structural selectivity properties. If preoperative high-resolution volume images and intraoperative2D ultrasound images could be integrated effectivelyfor soft tissue deformation tracking, the navigation accuracy would be greatlyimproved. In this study, with the integration of high resolution and high tissuecontrast preoperative MRI images, real-time intraoperative ultrasound imaging,electromagnetic tracking system and improved HAMMER algorithm, thescheme for the detection and correction of soft tissue deformation based on MRIand ultrasound imaging is proposed and evaluated.Due to the complexity of human structure and tissue deformation, it isdifficult to perform a quantitative evaluation of the detection and correctionsystem without ground truth of the deformation. Considering the difficulty onusing physical phantoms to reflect deformation characteristics of soft tissue, anUltrasound Computer Simulation Module was first designed for quantitativeevaluation of the proposed detection algorithm. In this module,2D ultrasoundimages could be simulated based on pre-operation3D MRI images, providingground truth for quantitative evaluation of deformation detection and correctionin the following experiments.In the simulation experiments, the performance of the proposed algorithmwas evaluated with simulated global and local deformation of abdominalsoft-tissue. In the global deformation experiment, three rigid and non-rigidtransformations were applied on a MRI image and then deformed ultrasoundimages were simulated from deformed MRI images. To detect and registerdeformed ultrasound image with original MRI image, an improved HAMMERalgorithm was applied, and the average displacement error and normalizedmutual information were calculated to evaluate the performance of deformationdetection and correction. The results indicate that the average displacement errorof all corrected images is less than4pixels and their normalized mutual information is more than0.6. For comparison, in the local deformationexperiment, only the deformation of the region of interest (ROI, i.e., liver) wasperformed on the MRI image using rigid and non-rigid transformation, andsimulated ultrasound images were generated from deformed MRI ROI. Then theimproved HAMMER algorithm was applied to detect and register deformedultrasound with original ROI image. The experiment results indicate that theaverage displacement error of corrected images is less than2pixels, and theirnormalized mutual information is more than0.8. Finally, a volunteer experimentwas performed using the proposed detection scheme and correction method.These preliminary experimental results indicate the feasibility of the proposedsystem for the detection and correction of tissue deformation.