Granted Patent B2

US 12,456,282 · App. 17/790,524 · Granted Oct 28, 2025

Inventors: Jun Ma (Singapore, SG); Kok Kiong Tan (Singapore, SG); Ban Leong Sng (Singapore, SG); Tiong Heng Sia (Singapore, SG)

Assignees: National University of Singapore; Singapore Health Services Pte. Ltd.

G06V10/764A61B8/085A61B8/0875A61B8/4444A61B8/5223A61B8/5246G06T7/248G06T7/62G06T7/66G06V10/16G06V10/44G06V10/751A61B2017/3413G06T2207/10132G06T2207/30012

Abstract

Claims (53)

1. An ultrasound imaging method for identifying an anatomical feature of a spine, the method comprising:

receiving, by a processor, a transverse ultrasound image of a portion of the spine;

extracting, by the processor, features of the portion of the spine from the transverse ultrasound image based on a distinct pattern associated with the anatomical feature of the spine;

identifying, by the processor, a midline of the portion of the spine in the transverse ultrasound image;

extracting, by the processor, midline features using pixel intensity values of the transverse ultrasound image; and

identifying, by the processor based on a combination of the extracted features of the portion of the spine and the extracted midline features, the anatomical feature in the transverse ultrasound image,

wherein the anatomical feature of the spine includes a neuraxial space,

wherein the extracted features of the portion of the spine include one or more of: a depth of a posterior complex, a depth of an anterior complex, a normalized cross-correlation strength of the posterior complex, a normalized cross-correlation strength of the anterior complex, a location of a left articular process, a location of a right articular process, a normalized cross- correlation value of a left articular process, a normalized cross-correlation value of a right articular process, and a horizontal position of the posterior complex, and a horizontal position of the anterior complex, and

wherein the extracted midline features include a ratio of black pixels in a predefined window and a ratio of white pixels below the posterior complex.

2. The method of claim 1 , wherein identifying the midline of the portion of the spine comprises using, by the processor, a cost function associated with a presence of white pixels within the predefined window of the transverse ultrasound image, wherein the cost function includes a first term associated with a depth of the pixels and a second term associated with a position of the midline in relation to a centre of the transverse ultrasound image.

3. The method of claim 1 , further comprising pre-processing, by the processor, the transverse ultrasound image prior to extracting features of the portion of the spine and the midline.

4. The method of claim 1 , further comprising:

calculating, by the processor, a depth of the neuraxial space; or identifying, by the processor, the neuraxial space using a Support Vector Machine (SVM) classifier.

5. The method of claim 4 , wherein the neuraxial space is identified using the SVM classifier, the method further comprises training, by the processor, the SVM classifier using historical data associated with the features of the portion of the spine and the midline features.

6. A computer readable medium storing processor executable instructions which when executed on a processor cause the processor to carry out a method according to claim 1 .

7. An ultrasound imaging method for identifying an anatomical feature of a spine, the method comprising:

receiving, by a processor, a transverse ultrasound image of a portion of the spine;

extracting, by the processor, features of the portion of the spine from the transverse ultrasound image based on a distinct pattern associated with the anatomical feature of the spine;

identifying, by the processor, a midline of the portion of the spine in the transverse ultrasound image;

extracting, by the processor, midline features using pixel intensity values of the transverse ultrasound image;

identifying, by the processor based on a combination of the extracted features of the portion of the spine and the extracted midline features, the anatomical feature in the transverse ultrasound image;

receiving, by the processor, a paramedian sagittal ultrasound image of another portion of the spine;

identifying, by the processor, morphological features of the paramedian sagittal ultrasound image for classifying the another portion of the spine, the morphological features comprise one or more of: a contour linearity, a size, a centre location and an area of the paramedian sagittal ultrasound image; and

determining, by the processor based on the identified morphological features, if the another portion of the spine includes a sacrum,

wherein identifying the morphological features of the paramedian sagittal ultrasound image comprises identifying, by the processor, the morphological features using template matching, and the method further comprises:

identifying, by the processor, an interspinous space based on the identified morphological features of the paramedian sagittal ultrasound image, or wherein the another portion of the spine is determined to include the sacrum, the method further comprises:

receiving, by the processor, a further paramedian sagittal ultrasound image of the spine;

determining, by the processor based on template matching of (i) laminas of the spine in the paramedian sagittal ultrasound image and the further paramedian sagittal ultrasound image and (ii) a cross relation strength between the paramedian sagittal ultrasound image and the further paramedian sagittal ultrasound image, if the further paramedian sagittal ultrasound image is suitable to be stitched to the paramedian sagittal ultrasound image; and

stitching, by the processor, the further paramedian sagittal ultrasound image to the paramedian sagittal ultrasound image to form a portion of a panorama image of the spine if it is determined that the further paramedian sagittal ultrasound image is suitable to be stitched to the paramedian sagittal ultrasound image.

8. The method of claim 7 , wherein the portion of the panorama image is formed, and the method further comprises:

identifying, by the processor, locations of interspinous spaces of the portion of the panorama image; and

identifying, by the processor, spinal levels using the locations of the interspinous spaces.

9. The method of claim 8 , wherein identifying locations of interspinous spaces of the portion of the panorama image further comprises:

calculating, by the processor, a motion between the further paramedian sagittal ultrasound image and a leading edge of the paramedian sagittal ultrasound image using phase correlation; and

transforming, by the processor, coordinates of a centre of an ultrasound probe in respect of the portion of the panorama image.

10. The method of claim 9 , further comprising:

determining, by the processor, if the ultrasound probe is at an optimal level of the spine based on the identified spinal levels; and

transmitting, by the processor, a request for obtaining the transverse ultrasound image of the portion of the spine if the ultrasound probe is at the optimal level.

11. The method of claim 7 , further comprising pre-processing, by the processor, the paramedian sagittal ultrasound image prior to identifying the morphological features.

12. The method of claim 11 , wherein pre-processing the paramedian sagittal ultrasound image comprises performing, by the processor, a contrast enhancement on the paramedian sagittal ultrasound image using a Difference of Gaussians (DoG) local normalization filter or eliminating, by the processor, candidate contour points in the paramedian sagittal ultrasound image using intensity thresholding.

13. An ultrasound imaging system for identifying an anatomical feature of a spine, the system comprising:

an ultrasound probe arranged to scan the spine for a transverse ultrasound image of a portion of the spine; and

a computer comprising a processor and a data storage storing computer program instructions operable to cause the processor to:

receive the transverse ultrasound image of the portion of the spine;

extract features of the portion of the spine from the transverse ultrasound image based on a distinct pattern associated with the anatomical feature of the spine;

identify a midline of the portion of the spine in the transverse ultrasound image;

extract midline features using pixel intensity values of the transverse ultrasound image; and

identify, based on a combination of the extracted features of the portion of the spine and the extracted midline features, the anatomical feature in the transverse ultrasound image,

wherein the anatomical feature of the spine includes a neuraxial space,

wherein the extracted features of the portion of the spine include one or more of: a depth of a posterior complex, a depth of an anterior complex, a normalized cross-correlation strength of the posterior complex, a normalized cross-correlation strength of the anterior complex, a location of a left articular process, a location of a right articular process, a normalized cross-correlation value of a left articular process, a normalized cross-correlation value of a right articular process, and a horizontal position of the posterior complex, and a horizontal position of the anterior complex, and

wherein the extracted midline features include a ratio of black pixels in a predefined window and a ratio of white pixels below the posterior complex.

14. The system of claim 13 , wherein the ultrasound probe comprises a clamping system arranged to attach a needle to the ultrasound probe for performing a neuraxial procedure.

15. The system of claim 14 , wherein the clamping system comprises a clip for attaching the needle using a spring loading mechanism and a locking mechanism for locking a position of the clip to fix a position and an angle of the needle, or wherein the clamping system is in a form of a cover for the ultrasound probe.

Assignments (4)

Priority Claims (1)

Continuity (1)

References Cited (51)