System and method for patient privacy protection in medical images

Patent metadata

Field	Value
Publication number	US-10679740-B2
Application number	US-201816006663-A
Country	US
Kind code	B2
Filing date	Jun 12, 2018
Priority date	Jun 12, 2018
Publication date	Jun 9, 2020
Grant date	Jun 9, 2020

Abstract

Systems and methods for de-identification of medical images can be applied to medical images acquired using various techniques. A 3D medical image can be analyzed to generate an image mask that partitions the image into a foreground region and a background region. From the image mask, a “skin surface” can be reconstructed based on the boundary between the foreground region and the background region. The image mask can be modified, e.g., by moving a randomly-selected subset of the voxels from the foreground region to the background region so that the shape of the skin surface is altered, thus obscuring patient-identifying features. The original medical image can be modified by changing the intensity of voxels in the background region while preserving the original intensity of voxels in the foreground region.

First claim

What is claimed is:

1. A method of de-identifying a medical image, the method comprising:
obtaining medical image data representing anatomy of a patient, the medical image data including a set of voxels defined in a three-dimensional space, each voxel having an original intensity value;
analyzing the medical image data to generate an image mask that assigns each of the voxels to either a foreground region or a background region such that a skin surface at a boundary between the foreground region and the background region corresponds to one or more surface anatomical features of the patient;
modifying the image mask by moving a randomly selected subset of voxels from the foreground region to the background region such that the skin surface is reshaped; and
modifying the medical image data by assigning a uniform background intensity value to each voxel of the medical image data that is assigned to the background region of the modified image mask while preserving the original intensity values of each voxel of the medical image data that is assigned to the foreground region of the modified image mask.

2. The method of claim 1 wherein analyzing the medical image data to generate the image mask includes:
identifying a largest connected region of low intensity voxels as belonging to a first approximate background region of a first approximate image mask; and
identifying all other voxels as belonging to a first approximate foreground region of the first approximate image mask.

3. The method of claim 2 wherein analyzing the medical image data to generate the image mask further includes:
applying a morphological correction to the first approximate background region; and
modifying the first approximate image mask based on the morphological correction to produce a second approximate image mask that divides the voxels into a second approximate foreground region and a second approximate background region.

4. The method of claim 3 wherein analyzing the medical image to generate the image mask further includes:
reconstructing an approximate skin surface from the second approximate image mask;
identifying a set of super-voxels around the approximate skin surface; and
for each super-voxel:
determining whether the super-voxel includes at least a threshold number of voxels that are within the second approximate foreground region;
assigning all voxels within the super-voxel to a refined foreground region of a refined image mask in the event that the super-voxel includes at least the threshold number of voxels that are within the second approximate foreground region; and
assigning all voxels within the super-voxel to a refined background region of the refined image mask in the event that the super-voxel does not include at least the threshold number of voxels that are within the second approximate foreground region.

5. The method of claim 1 wherein modifying the image mask includes:
randomly selecting a plurality of seed locations on the skin surface of the image mask; and
for each seed location, applying a kernel at the seed location to select one or more voxels near the seed location to be moved from the foreground region to the background region of the image mask.

6. The method of claim 5 wherein the same kernel is applied at each seed location.

7. The method of claim 6 wherein the kernel is a spherical kernel.

8. The method of claim 5 wherein modifying the image mask further includes:
applying iterative Gaussian smoothing to propagate a deformation at each seed location.

9. The method of claim 1 wherein the medical image data is data produced from a magnetic resonance imaging (MM) scan of the patient.

10. A computer system comprising:
a storage medium to store medical image data representing anatomy of a patient, the medical image data including a set of voxels defined in a three-dimensional space, each voxel having an original intensity value; and
a processor coupled to the storage medium and configured to:
analyze the medical image data to generate an image mask that assigns each of the voxels to either a foreground region or a background region such that a skin surface at a boundary between the foreground region and the background region corresponds to one or more surface anatomical features of the patient;
modify the image mask by moving a randomly selected subset of voxels from the foreground region to the background region such that the skin surface is reshaped; and
modify the medical image data by assigning a uniform background intensity value to each voxel of the medical image data that is assigned to the background region of the modified image mask while preserving the original intensity values of each voxel of the medical image data that is assigned to the foreground region of the modified image mask.

11. The computer system of claim 10 wherein analyzing the medical image data to generate the image mask includes:
identifying a largest connected region of low intensity voxels as belonging to a first approximate background region of a first approximate image mask; and
identifying all other voxels as belonging to a first approximate foreground region of the first approximate image mask.

12. The computer system of claim 11 wherein analyzing the medical image data to generate the image mask further includes:
applying a morphological correction to the first approximate background region; and
modifying the first approximate image mask based on the morphological correction to produce a second approximate image mask that divides the voxels into a second approximate foreground region and a second approximate background region.

13. The computer system of claim 12 wherein analyzing the medical image to generate the image mask further includes:
reconstructing an approximate skin surface from the second approximate image mask;
identifying a set of super-voxels around the approximate skin surface; and
for each super-voxel:
determining whether the super-voxel includes at least a threshold number of voxels that are within the second approximate foreground region;
assigning all voxels within the super-voxel to a refined foreground region of a refined image mask in the event that the super-voxel includes at least the threshold number of voxels that are within the second approximate foreground region; and
assigning all voxels within the super-voxel to a refined background region of the refined image mask in the event that the super-voxel does not include at least the threshold number of voxels that are within the second approximate foreground region.

14. The computer system of claim 10 wherein modifying the image mask includes:
randomly selecting a plurality of seed locations on the skin surface of the image mask; and
for each seed location, applying a kernel at the seed location to select one or more voxels near the seed location to be moved from the foreground region to the background region of the image mask.

15. The computer system of claim 14 wherein the same kernel is applied at each seed location.

16. The computer system of claim 15 wherein the kernel is a spherical kernel.

17. The computer system of claim 14 wherein modifying the image mask further includes:
applying iterative Gaussian smoothing to propagate a deformation at each seed location.

18. The computer system of claim 10 wherein the medical image data is data produced from a magnetic resonance imaging (MRI) scan of the patient.

19. A computer-readable storage medium having stored therein program instructions that, when executed

Assignees

• Univ Hong Kong Chinese

Inventors

• Chen Weitian
• Gan ke

Classifications

• G16H10/60

  for patient-specific data, e.g. for electronic patient records · CPC title

• G06T7/0012

  Biomedical image inspection · CPC title

• G06T19/20

  Editing of three-dimensional [3D] images, e.g. changing shapes or colours, aligning objects or positioning parts · CPC title

• G06T5/20

  using local operators · CPC title

• G06T7/155

  involving morphological operators · CPC title