EP-4490696-B1 - OPTIMAL PATH FINDING BASED SPINAL CENTER LINE EXTRACTION

EP4490696B1EP 4490696 B1EP4490696 B1EP 4490696B1EP-4490696-B1

Inventors

CHEN, HONGXIN
BUERGER, CHRISTIAN
KLINDER, TOBIAS
LORENZ, CRISTIAN

Dates

Publication Date: 20260506
Application Date: 20230307

Claims (15)

A method of generating an optimal spinal canal path, the method comprising: receiving (302) scan data as a three-dimensional array of a scanned spine image; and executing (304) a three-dimensional voxel wise spinal canal segmentation from the received scan data to generate a binary three-dimensional array corresponding to the scanned spine image; characterized in that the method further comprises: generating (306) an inverse density three-dimensional array corresponding to the generated binary three-dimensional array having a size the same as the generated binary three-dimensional array; down sampling (308) the inverse density three-dimensional array to a predetermined size; performing programming based optimal spinal canal path finding (310) to generate an optimal spinal canal path; sampling (312) the inverse three-dimensional array around the generated optimal spinal canal path with higher resolution and limited size; performing programming based optimal spinal canal path finding (314) to generate a revised optimal spinal canal path; repeating the sampling and the programming based optimal spinal canal path finding by incrementally increasing the resolution and decreasing the size; determining (316) whether an accuracy of the optimal spinal canal path satisfies a predetermined threshold; and generating (318) a final optimal spinal canal path as a spinal canal center line, wherein a reformatted image is generated according to the final optimal spinal canal path as a spinal canal center line.
The method of generating an optimal spinal canal path according to claim 1, wherein spinal canal segmentation is performed using machine or deep learning techniques, wherein a value of 1 is assigned to each spinal canal voxel and a value of 0 is assigned to background in the generated binary three-dimensional array.
The method of generating an optimal spinal canal path according to claim 2, wherein generating the inverse density three-dimensional array includes assigning a float value to each voxel where the higher a density value of 1 voxels are in the binary three-dimensional array, the lower the float value assigned to a corresponding voxel in the inverse density three-dimensional array.
The method of generating an optimal spinal canal path according to claim 1, wherein the programming based optimal spinal canal path finding comprises: creating a cost-pointer array that is a three-dimensional array corresponding to the inverse density three-dimensional array, wherein each element in the cost-pointer array comprises a cost component and a pointer component, wherein each cost component in the cost-pointer array corresponds to a voxel in a same position in the inverse density three-dimensional array, and wherein the pointer component is a three-dimensional vector which points to a physical coordinate in the binary three-dimensional array.
The method of generating an optimal spinal canal path according to claim 1, wherein the programming based optimal spinal canal path finding is dynamic programming based optimal spinal canal path finding.
The method of generating an optimal spinal canal path according to claim 1, wherein a density estimation is applied to generate the inverse density three-dimensional array.
The method of generating an optimal spinal canal path according to claim 6, wherein the density estimation is performed via kernel density estimation techniques, and wherein the kernel is at least one of gaussian, linear, cosine, uniform, epanechnikov, exponential, and tophat.
An optimal spinal canal path generating apparatus, comprising: a memory configured to store computer executable instructions; and at least one processor configured to execute the computer executable instructions to cause the optimal path finding apparatus to: receive (302) scan data as a three-dimensional array of a scanned spine image; and execute (304) a three-dimensional voxel wise spinal canal segmentation from the received scan data to generate a binary three-dimensional array corresponding to the scanned spine image; characterized in that the computer executable instructions further cause the optimal path finding apparatus to: generate (306) an inverse density three-dimensional array corresponding to the generated binary three-dimensional array having a size the same as the generated binary three-dimensional array; down sample (308) the inverse density three-dimensional array to a predetermined size; perform programming based optimal spinal canal path finding (310) to generate an optimal spinal canal path; sample (312) the inverse three-dimensional array around the generated optimal spinal canal path with higher resolution and limited size; perform (314) programming based optimal spinal canal path finding (314) to generate a revised optimal spinal canal path; repeat the sampling and the programming based optimal spinal canal path finding by incrementally increasing the resolution and decreasing the size; determine (316) whether an accuracy of the optimal spinal canal path satisfies a predetermined threshold; and generate (318) a final optimal spinal canal path as a spinal canal center line, wherein a reformatted image is generated according to the final optimal spinal canal path as a spinal canal center line.
The optimal spinal canal path generating apparatus according to claim 8, wherein the spinal canal segmentation is performed using machine or deep learning techniques, wherein a value of 1 is assigned to each spinal canal voxel and a value of 0 is assigned to background in the generated binary three-dimensional array.
The optimal spinal canal path generating apparatus according to claim 9, wherein the processor is configured to assign a float value to each voxel in the generated inverse density three-dimensional array, where the higher a density value of 1 voxels are in the binary three-dimensional array, the lower the float value assigned to a corresponding voxel in the inverse density three-dimensional array
The optimal spinal canal path generating apparatus according to claim 8, wherein the processor is further configured to create a cost-pointer array that is a three-dimensional array corresponding to the inverse density three-dimensional array, wherein each element in the cost-pointer array comprises a cost component and a pointer component, wherein each cost component in the cost-pointer array corresponds to a voxel in a same position in the inverse density three-dimensional array, and wherein the pointer component is a three-dimensional vector which points to a physical coordinate in the binary three-dimensional array.
The optimal spinal canal path generating apparatus according to claim 8, wherein the processor is configured to perform the programming based optimal spinal canal path finding via dynamic programming techniques.
The optimal spinal canal path generating apparatus according to claim 8, wherein the processor is configured to apply a density estimation to generate the inverse density three-dimensional array.
The optimal spinal canal path generating apparatus according to claim 13, wherein the density estimation is performed via kernel density estimation techniques, and wherein the kernel is at least one of gaussian, linear, cosine, uniform, epanechnikov, exponential, and tophat.
A non-transitory computer-readable medium having stored thereon instructions for causing processing circuitry to execute a process, the process comprising: receiving (302) scan data as a three-dimensional array of a scanned spine image; and executing (304 a three-dimensional voxel wise spinal canal segmentation from the received scan data to generate a binary three-dimensional array corresponding to the scanned spine image; characterized in that the process further comprises: generating (306) an inverse density three-dimensional array corresponding to the generated binary three-dimensional array having a size the same as the generated binary three-dimensional array; down sampling (308) the inverse density three-dimensional array to a predetermined size; performing (310) programming based optimal spinal canal path finding to generate an optimal spinal canal path; sampling (312) the inverse three-dimensional array around the generated optimal spinal canal path with higher resolution and limited size; performing programming based optimal spinal canal path finding (314) to generate a revised optimal spinal canal path; repeating the sampling and the programming based optimal spinal canal path finding by incrementally increasing the resolution and decreasing the size; determining (316) whether an accuracy of the optimal spinal canal path satisfies a predetermined threshold; and generating (318) a final optimal spinal canal path as a spinal canal center line, wherein a reformatted image is generated according to the final optimal spinal canal path as a spinal canal center line.

Description

FIELD OF THE INVENTION The invention relates to the field of medical imaging, in particular to a method and apparatus of extracting a spinal canal center line that is not sensitive to segmentation errors. BACKGROUND OF THE INVENTION Medical imaging is often used to visualize a patient's anatomy where the visualization data is used to diagnose diseases or injuries. In a trauma setting, medical personnel will often perform an imaging scan such as a Computer Tomography (CT) scan to diagnose fractures of the ribs and spine. To help a radiologist read the CT spine image, the spine image is often reformatted in a straightened curved planar reformat (CPR). The reformatted image may be a straightened view in 2D or 3D according to the spinal canal curve. The first step in generating the reformatted image is to extract the center line of the spinal canal. This is typically accomplished by executing a three-dimensional (3D) spinal canal segmentation and then extracting the center line of the segmentation. If the segmentation result is good, extracting the center line may not be difficult. Frequently, however, there are some errors in the segmentation. On occasion, part of the spinal canal is missed (gaps in between slices). On other occasions, there are false segmentations where the predicted spinal canal does not reflect that actual spinal canal. These issues create complex challenges to the center line extraction. If the extracted spinal canal center line is incorrect, the reformatted image will also be incorrect leading to unrealistic image distortions within the resulting image. Reading imaging scans, and more specifically trauma or Emergency Department scans, is a time-critical task that needs to be done with high attention to avoid overlooking critical findings. Reformatting an image scan to a straightened view in 2D or 3D according to the spinal canal curve is performed to aid medical professionals diagnosing critical rib and spine fractures. A reformatted image with unrealistic image distortions defeats the purpose of the reformatted image. Thus, the need exists for innovative techniques to extract a spinal canal center that are not sensitive to segmentation errors. WO2017/074890 discloses a segmentation method of the spine using 3D volumetric data, by detecting a spine centerline and a spine canal centerline, localizing the vertebra and intervertebral disc centers, generationg hard constraints for each vertebra digit and apply these constraints when segmenting each vertebra digit. An article "Automatic spinal canal detection in lumbar MR images in the sagittal view using dynamic programming", by Koh Jaehan et al, in Computerized Medical Imaging and Graphics, Pergamon Press, NY, US, vol. 38, no.7, 18 June 2014, p 569-579, ISSBN 0895-6111, DOI 10.1016/J.COMPMEDIMAG.2014.06.003, disclosesa computer-aided diagnosis and characterization framework using lumbar spine MRI to provide radiologists a second option. It proposes a left spinal canal boundary extraction method, fusing the absolute intensity difference of T1-weighted and T2-weighted sagittal images and the inverted gradient of the difference images into a dynamic programming scheme. An article "Automated generation of curved planar reformations from MR images of the spine; Automated generation of CPRs from MR images of the spine", by Tomaz Vrtovec et al, in Physics in Medicine and Biology, Institute of Physics Publishing, Bristol GB, vol. 52, no. 10, 21 May 2007, p 2865-2878, ISSN 0031-9155, DOI:10.1088/0031-9155/52/10/015, discloses automated curved planar reformation of MR images of the spine. The 3D spine curve and axial vertebral rotation are described by polynomial functions obtained by robust refinement of initial estimates. An article "On computerized methods for spine analysis in MRI: a systematic review", by Rak Marko et al, in International Journal of Computer Assisted Radiology and Surgery, Springer, DE, vol. 11, no. 8, 9 February 2016, p 1445-1465, ISSN 1861-6410, DOI: 10.1007/S11548-016-1350-2, provides a review of computerized methods for MR-based spine analysis. SUMMARY OF THE INVENTION The object of the present invention provides techniques for extracting a spinal canal center that are not sensitive to segmentation errors. The techniques may be applied to virtually any spine image that is reformatted to a straightened view in 2D or 3D according to the spinal canal curve. These spine images may be generated by a number of imaging systems including CT, CT-arm, Single Photon Emission Computer Tomography CT (SPECT-CT), Magnetic Resonance CT (MR-CT), Positron Emission Tomography CT (PET-CT), and Magnetic Resonance Imaging (MRI) systems. According to a first aspect of the invention a method of generating an optimal spinal canal is provided. The method comprising receiving scan data as a three-dimensional array of a scanned spine image and executing a three-dimensional voxel wise spinal canal segmentation from the received scan data to generate a bi