Detecting spinal shape from optical scan

Patent metadata

Field	Value
Publication number	US-12016697-B2
Application number	US-201917271687-A
Country	US
Kind code	B2
Filing date	Aug 28, 2019
Priority date	Aug 28, 2018
Publication date	Jun 25, 2024
Grant date	Jun 25, 2024

Abstract

A method comprising: generating a parametrized three-dimensional (3D) body surface model on a training set comprising a plurality of 3D scans of subjects, wherein at least some of said 3D scans are of subjects having a skeletal deformity; receiving one or more target 3D scans of a target subject; optimizing said body surface model with respect to said one or more target 3D scans to calculate a target body surface model of said target subject; training a skeletal estimation model on a training set comprising: (i) body surface models of a plurality of subjects, and (ii) skeletal landmarks sets of said plurality of subjects; and applying said trained skeletal estimation model to said calculated target body surface model of said target subject, to estimate a skeletal shape of said target subject.

First claim

What is claimed is:

1. A method comprising:
generating a parametrized three-dimensional (3D) body surface model, based, at least in part, on a training set comprising a plurality of 3D scans of subjects, wherein at least some of said 3D scans are of subjects having a skeletal deformity;
receiving one or more target 3D scans of a target subject;
optimizing said body surface model with respect to said one or more target 3D scans, based, at least in part, on minimizing a loss function which registers said body surface model to said target 3D scans, to calculate a target body surface model of said target subject;
training a skeletal estimation model, based, at least in part, on a training set comprising:
(i) body surface models of a plurality of subjects, and
(ii) skeletal landmarks sets of said plurality of subjects; and
applying said trained skeletal estimation model to said calculated target body surface model of said target subject, to estimate a skeletal shape of said target subject,
wherein at least some of said target 3D scans are labelled with a pose of said subject associated with a respective target 3D scan.

2. The method of claim 1 , wherein said training of said 3D body surface model is configured to encode a body type parameter, based, at least in part, on a labeling of said training set with said skeletal deformity.

3. The method of claim 2 , wherein said labelling is a scoliosis classification category selected from the group consisting of: main thoracic, double thoracic, double/triple major, and thoracolumbar/lumbar.

4. The method of claim 2 , wherein said encoding is a low-dimensional encoding.

5. The method of claim 1 , wherein said target 3D scans comprise at least one of: point clouds, triangulated meshes, and splined surfaces.

6. A system comprising:
at least one hardware processor; and
a non-transitory computer-readable storage medium having stored thereon program code, the program code executable by the at least one hardware processor to:
generate a parametrized three-dimensional (3D) body surface model, based, at least in part, on a training set comprising a plurality of 3D scans of subjects, wherein at least some of said 3D scans are of subjects having a skeletal deformity,
receive one or more target 3D scans of a target subject,
optimize said body surface model with respect to said one or more target 3D scans, based, at least in part, on minimizing a loss function which registers said body surface model to said target 3D scans, to calculate a target body surface model of said target subject,
train a skeletal estimation model, based, at least in part, on a training set comprising:
(i) body surface models of a plurality of subjects, and
(ii) skeletal landmarks sets of said plurality of subjects, and
apply said trained skeletal estimation model to said calculated target body surface model of said target subject, to estimate a skeletal shape of said target subject,
wherein at least some of said target 3D scans are labelled with a pose of said subject associated with a respective target 3D scan.

7. The system of claim 6 , wherein said training of said 3D body surface model is configured to encode a body type parameter, based, at least in part, on a labeling of said training set with said skeletal deformity.

8. The system of claim 7 , wherein said labelling is a scoliosis classification category selected from the group consisting of: main thoracic, double thoracic, double/triple major, and thoracolumbar/lumbar.

9. The system of claim 7 , wherein said encoding is a low-dimensional encoding.

10. The system of claim 6 , wherein said target 3D scans comprise at least one of: point clouds, triangulated meshes, and splined surfaces.

11. A computer program product comprising a non-transitory computer-readable storage medium having program code embodied therewith, the program code executable by at least one hardware processor to:
generate a parametrized three-dimensional (3D) body surface model, based, at least in part, on a training set comprising a plurality of 3D scans of subjects, wherein at least some of said 3D scans are of subjects having a skeletal deformity;
receive one or more target 3D scans of a target subject;
optimize said body surface model with respect to said one or more target 3D scans, based, at least in part, on minimizing a loss function which registers said body surface model to said target 3D scans, to calculate a target body surface model of said target subject;
train a skeletal estimation model, based, at least in part, on a training set comprising:
(i) body surface models of a plurality of subjects, and
(ii) skeletal landmarks sets of said plurality of subjects; and
apply said trained skeletal estimation model to said calculated target body surface model of said target subject, to estimate a skeletal shape of said target subject,
wherein at least some of said target 3D scans are labelled with a pose of said subject associated with a respective target 3D scan.

12. The computer program product of claim 11 , wherein said training of said 3D body surface model is configured to encode a body type parameter, based, at least in part, on a labeling of said training set with said skeletal deformity.

13. The computer program product of claim 12 , wherein said labelling is a scoliosis classification category selected from the group consisting of: main thoracic, double thoracic, double/triple major, and thoracolumbar/lumbar.

14. The computer program product of claim 12 , wherein said encoding is a low-dimensional encoding.

15. The computer program product of claim 11 , wherein said target 3D scans comprise at least one of: point clouds, triangulated meshes, and splined surfaces.

Assignees

• Technion Res & Dev Foundation
• New York For The Relief Of The Ruptured And Crippled Maintaining The Hospital For Special Surgery
• New York Soc For The Relief Of The Ruptured And Crippled Maintaining The Hospital For Special Surger

Inventors

• Kimmel Ron
• Groisser Benjamin
• Wolf Alon
• Widmann Roger F
• Hillstrom Howard J

Classifications

• G06T2200/08

  involving all processing steps from image acquisition to 3D model generation · CPC title

• G06T2210/41

  Medical · CPC title

• G06T2207/30012

  Spine; Backbone · CPC title

• G06T2207/20101

  Interactive definition of point of interest, landmark or seed · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title