US-12620169-B2 - Dense non-rigid volumetric mapping of image coordinates using sparse surface-based correspondences

Abstract

Examples of the presently disclosed technology provide systems and methods for improved image registration for image-guided brain interventions. The disclosed systems and methods use sparse surface-based correspondences between vertices of patient-specific 3D mesh representations to fit a non-rigid transformation function for estimating a deformation field that maps one cranial image (e.g., a source image used for surgical trajectory planning) to another cranial image (e.g., a reference image obtained during an image-guided brain intervention).

Inventors

• Lyubomir ZAGORCHEV

Assignees

• CLEARPOINT NEURO, INC.

Dates

Publication Date

20260505

Application Date

20240111

Claims (20)

1. 1 . A method, comprising: adapting a shape-constrained deformable cranial region model to a first cranial image to generate a first patient-specific 3D mesh representation; adapting the shape-constrained deformable cranial region model to a second cranial image to generate a second patient-specific 3D mesh representation; using correspondences between the first and second patient-specific 3D mesh representations to fit a non-rigid transformation function; estimating a deformation field using the non-rigid transformation function; detecting non-rigid brain shift in a patient responsive to determining at least one of the following: length of at least one displacement vector of the estimated deformation field exceeds a threshold length value, or over a threshold number of displacement vectors of the estimated deformation field have lengths exceeding the threshold length value; mapping coordinates of planned target points and entry points in the first cranial image to corresponding locations in the second cranial image using the estimated deformation field; and displaying, on a graphical user interface (GUI), a visual representation of the second cranial image overlaid with: visual highlights for the mapped planned target points and entry points at their corresponding locations in the second cranial image, and a visual warning indicating the detected non-rigid brain shift.
2. 2 . The method of claim 1 , wherein the correspondences between the first and second patient-specific 3D mesh representations comprise surface-based correspondences between mesh vertices of the first and second patient-specific 3D mesh representations.
3. 3 . The method of claim 1 , wherein the first cranial image is a pre-operative image of a patient's cranial region used for surgical trajectory planning and the second cranial image is an intra-operative image of the patient's cranial region obtained during a surgical intervention into the patient's cranial region.
4. 4 . The method of claim 1 , wherein the first cranial image is a brain atlas used for surgical trajectory planning and the second cranial image is an intra-operative image of a patient's cranial region obtained during a surgical intervention into the patient's cranial region.
5. 5 . The method of claim 1 , wherein the shape-constrained deformable cranial region model comprises a computerized 3D mesh representation of a non-patient-specific human cranial region that preserves point-based correspondences during mesh adaption to patient images.
6. 6 . The method of claim 1 , wherein the first and second cranial images comprise at least one of the following types of images: a magnetic resonance (MR) image; a computerized tomography (CT) image; and a positron emission tomography (PET) image.
7. 7 . The method of claim 6 , wherein the first and second cranial images comprise different types of images.
8. 8 . The method of claim 1 , wherein the estimated deformation field is a dense deformation field comprising an individual displacement vector associated with each voxel of the first cranial image, the displacement vectors mapping voxels of the first cranial image to corresponding voxels of the second cranial image.
9. 9 . Non-transitory computer-readable storage medium including instructions that, when executed by one or more processors of a computing system, cause the computing system to: adapt a shape-constrained deformable cranial region model to a first cranial image to generate a first patient-specific 3D mesh representation; adapt the shape-constrained deformable cranial region model to a second cranial image to generate a second patient-specific 3D mesh representation; use surface-based correspondences between mesh vertices of the first and second patient-specific 3D mesh representations to fit a non-rigid transformation function; estimate a deformation field using the non-rigid transformation function; detect non-rigid brain shift in a patient responsive to determining at least one of the following: length of at least one displacement vector of the estimated deformation field exceeds a threshold length value, or over a threshold number of displacement vectors of the estimated deformation field have lengths exceeding the threshold length value; map planned surgical target point and entry point coordinates from the first cranial image to the second cranial image using the estimated deformation field; and display, on a graphical user interface (GUI), a visual representation of the second cranial image overlaid with: visual highlights for the mapped planned target points and entry points at their corresponding locations in the second cranial image, and a visual warning indicating the detected non-rigid brain shift.
10. 10 . The non-transitory computer-readable medium of claim 9 , wherein the first cranial image is a pre-operative image of a patient's cranial region used for surgical trajectory planning and the second cranial image is an intra-operative image of the patient's cranial region obtained during a surgical intervention into the patient's cranial region.
11. 11 . The non-transitory computer-readable medium of claim 9 , further comprising instructions that, when executed by the one or more processors, cause the computing system to map the first cranial image to the second cranial image.
12. 12 . The non-transitory computer-readable medium of claim 11 , wherein the estimated deformation field is a dense deformation field comprising an individual displacement vector associated with each voxel of the first cranial image including voxels associated with the planned surgical target point and entry point coordinates, the displacement vectors mapping voxels of the first cranial image to corresponding voxels of the second cranial image.
13. 13 . The non-transitory computer-readable medium of claim 1 , wherein the shape-constrained deformable cranial region model comprises a computerized 3D mesh representation of a non-patient-specific human cranial region that preserves point-based correspondences during mesh adaption to patient images.
14. 14 . The non-transitory computer-readable medium of claim 9 , wherein the first and second cranial images comprise at least one of the following types of images: a MR image; a CT image; and a PET image.
15. 15 . The non-transitory computer-readable medium of claim 13 , wherein the first and second cranial images comprise different types of images.
16. 16 . A system comprising: one or more processing resources; and non-transitory computer-readable medium, coupled to the one or more processing resources, having stored therein instructions that when executed by the one or more processing resources cause the system to: adapt a shape-constrained deformable cranial region model to a first cranial image to generate a first patient-specific 3D mesh representation; adapt the shape-constrained deformable cranial region model to a second cranial image to generate a second patient-specific 3D mesh representation; use correspondences between mesh vertices of the first and second patient-specific 3D mesh representations to fit a non-rigid transformation function; estimate deformation field using the non-rigid transformation function; detect non-rigid brain shift in a patient responsive to determining at least one of the following: length of at least one displacement vector of the estimated deformation field exceeds a threshold length value, or over a threshold number of displacement vectors of the estimated deformation field have lengths exceeding the threshold length value; and display, on a graphic user interface (GUI), a visual warning indicating the detected non-rigid brain shift.
17. 17 . The system of claim 16 , wherein: the estimated deformation field is a dense deformation field comprising an individual displacement vector associated with each voxel of the first cranial image, the displacement vectors mapping voxels of the first cranial image to corresponding voxels of the second cranial image; and estimating non-rigid brain shift in the patient comprises determining magnitudes of displacement vectors of the estimated deformation field.
18. 18 . The system of claim 16 , wherein the shape-constrained deformable cranial region model comprises a computerized 3D mesh representation of a non-patient-specific human cranial region that preserves point-based correspondences during mesh adaption to patient images.
19. 19 . The system of claim 16 , wherein the correspondences between the first and second patient-specific 3D mesh representations comprise surface-based correspondences between mesh vertices of the first and second patient-specific 3D mesh representations.
20. 20 . The system of claim 16 , wherein the visual warning indicating the detected non-rigid brain shift overlays the second cranial image of the second patient-specific 3D mesh representation.

Description

REFERENCE TO RELATED APPLICATION The present application claims priority to U.S. Provisional Patent Application No. 63/479,733, filed Jan. 12, 2023, and titled “DENSE NON-RIGID VOLUMETRIC MAPPING OF IMAGE COORDINATES USING SPARSE SURFACE-BASED CORRESPONDENCE,” which is incorporated herein by reference in its entirety. TECHNICAL FIELD The present disclosure relates generally to medical technologies, and more particularly, some examples relate to non-rigid volumetric mapping for cranial images. BACKGROUND Image registration may refer to a computational method for determining a spatial transformation that maps one image (sometimes referred to as a “source or moving image”) to another image (sometimes referred to as a “reference or fixed image”) point-by-point (or more specifically, voxel-by-voxel for 3D images). Image registration can be either rigid or non-rigid. Rigid registration preserves distance between image points and involves only rotation and translation. By contrast, non-rigid registration is spatially unconstrained and in addition to rotation and translation may involve at least some “stretching” of images (here distances between image points of a stretched image may not be preserved). Image registration (rigid or non-rigid) typically utilizes a deformation field (i.e., a geometric transformation) to map features of a source/moving image to features in a reference/fixed image. In dense image registrations, the deformation field assigns a displacement vector to each image point of a source image. The displacement vector for a given point of a source image maps the given point to its corresponding spatial location in the reference image. BRIEF DESCRIPTION OF THE DRAWINGS The present disclosure, in accordance with one or more various examples, is described in detail with reference to the following figures. The figures are provided for purposes of illustration only and merely depict examples. FIG. 1 depicts cross sections of an overlay of two example cranial images without image registration, in accordance with various examples of the presently disclosed technology. FIG. 2 depicts cross sections of an overlay of two example MR cranial region images after rigid registration, in accordance with various examples of the presently disclosed technology. FIG. 3 depicts an example shape-constrained deformable cranial region model, in accordance with various examples of the presently disclosed technology. FIG. 4 depicts adaption of a shape-constrained deformable cranial region model to two example cranial images, in accordance with examples of the presently disclosed technology. FIG. 5 depicts a first example checkerboard overlay of two cranial images after rigid registration, and a second example checkerboard overlay of the two cranial images after non-rigid registration performed in accordance with the presently disclosed technology. FIG. 6 depicts an example flow diagram that may be used to non-rigidly transform a first cranial image to a second cranial image, in accordance with various examples of the presently disclosed technology. FIG. 7 depicts an example flow diagram that may be used to detect/estimate non-rigid brain shift during a brain intervention using an estimated deformation field, in accordance with various examples of the presently disclosed technology. FIG. 8 is an example computing component that may be used to implement various features of examples described in the present disclosure. The figures are not exhaustive and do not limit the present disclosure to the precise form disclosed. DETAILED DESCRIPTION In image-guided brain interventions, image registration can be used to align (in the same image space) cranial images acquired at different times (e.g., pre-operative vs. intra-operative) or using different modalities (e.g., a brain atlas, MR, computerized tomography (CT), positron emission tomography (PET), etc.). In conventional image-guided brain interventions, rigid registration is used to reduce differences in position and orientation across cranial images. An example overlay of two cranial images before image registration is shown in the third row of FIG. 1. An example overlay of the two cranial images after rigid registration is illustrated in the third row of FIG. 2. As depicted, rigid registration can bring the two cranial images into approximate alignment, but it can be less effective in aligning images of soft (brain) tissues within the skull. This misalignment of soft tissue images can result from “non-rigid brain shift” during a surgical intervention. Non-rigid brain shift may refer to a deformation/shift of soft tissues of the brain that is more complex than mere rotation or translation. During a surgical intervention into the brain, non-rigid brain shift may result from various causes including cerebrospinal fluid (CSF) leakage, a change in intercranial pressure, etc. As alluded to above (and as depicted in FIG. 2), rigid registrations have trouble accounting for non-rigid br