C-arm pose estimation using intensity-based registration of imaging modalities

#This invention relates to methods and systems for improving accuracy of clinical procedures that rely on C-arm fluoroscopy images. In embodiments of the methods and systems described herein the relative pose of C-arm images may be estimated by registering image data from imaging modalities using markers or a fiducial. The embodiments described herein are particularly useful in brachytherapy procedures.

#Although C-arm fluoroscopy is widely used for visual assessment during clinical procedures, a limitation is that it cannot show soft tissues or organs of interest and other relevant structures. On the other hand, use of other imaging modalities such as ultrasound during clinical procedures provides real-time visualization of the tissue/organ of interest but does not provide radiographic information. Fusion of these complementary imaging modalities would provide benefits during many procedures.

#For example, interventional tool registration with soft tissue would benefit from being able to accurately view the tool and the soft tissue of interest simultaneously during a medical procedure. The procedure could be performed with greater precision through improved tool visualization and registration with the soft tissues undergoing the procedure. Further, the ability to visualize the soft tissue using another imaging modality such as ultrasound provides real time information to the clinician performing the procedure.

#For a better understanding of the invention, and to show more clearly how it may be carried into effect, embodiments will now be described, by way of example, with reference to the accompanying drawings, wherein:

#FIG. 1A shows a cylindrical sheath design, and FIG. 1B shows a wire model, of a specific instance of a fluoroscopy tracking (FTRAC) fiducial. FIG. 1C shows a radiographic fiducial of size 3×5 cm attached to the needle insertion template during intervention.

#FIG. 2 is a flowchart showing a C-arm pose recovery method according to one embodiment.

#In the embodiments described herein, the relative pose of a C-arm fluoroscopy image is estimated by registration of the 2D fluoroscopy image to a 3D image volume. The 3D image volume may be generated by ultrasound or other suitable imaging technology. The pose of the C-arm images is estimated using a coordinate system fixed to the tissue/organ of interest. This allows the C-arm pose to be estimated with an accuracy that is sufficient for subsequent reconstruction of the implanted seeds and thus enable 2D/3D intensity based registration. In one embodiment the coordinate system may be established through use of a fiducial. In another embodiment the coordinate system may be established by one or more markers implanted in or otherwise attached or fixed to the tissue/organ of interest. For example, one or more of numerous imaging techniques may be used to accurately determine the placement of seeds during a brachytherapy procedure. These seeds represent the equivalent of a fiducial to subsequent modifications of C-arm pose. In this manner, dynamic pose estimation during a procedure may be carried out by iteratively updating the locations of seeds and using those seed locations to compute pose.

#Algorithms for 2D/3D registration of 2D radiographic projections to 3D CT images can be broadly categorized as feature-based methods, methods that use intensity-derived features, intensity-based methods, and hybrid methods. Feature based methods use distance between corresponding point pairs or surfaces as a measure to be optimized. Establishing point correspondences and minimizing the distance between them is alternated and repeated iteratively until convergence. Consequently, prior segmentation of the image data is required. Intensity-based methods compare the 2D image with a digitally reconstructed radiograph (DRR) created from the 3D volume. Key to intensity-based registration is the choice of an appropriate similarity measure to quantify the spatial alignment between the images. An understanding of the image formation process and the underlying relationship between intensity values in the modalities being matched is often useful in choosing the similarity measure. The hybrid registrations usually exploit both image and geometric information and are based on two alternated registrations.

#As described herein, the relative pose of a C-arm fluoroscopy image is estimated by registering the fluoroscopy image to a fiducial or marker image without segmentation. In particular, intensity-based registration techniques as described herein do not require prior segmentation, a substantial improvement over segmentation-based registration that usually requires some level of manual intervention and is hampered by segmentation errors. Intensity-based registration may use the entire image volumes and thus both false positive and false negative appearances may average out.