Plaque segmentation in intravascular optical coherence tomography (OCT) images using deep learning

Patent metadata

Field	Value
Publication number	US-11710238-B2
Application number	US-202017001126-A
Country	US
Kind code	B2
Filing date	Aug 24, 2020
Priority date	Oct 24, 2019
Publication date	Jul 25, 2023
Grant date	Jul 25, 2023

Abstract

Embodiments discussed herein facilitate segmentation of vascular plaque, training a deep learning model to segment vascular plaque, and/or informing clinical decision-making based on segmented vascular plaque. One example embodiment accessing vascular imaging data for a patient, wherein the vascular imaging data comprises a volume of interest; pre-process the vascular imaging data to generate pre-processed vascular imaging data; provide the pre-processed vascular imaging data to a deep learning model trained to segment a lumen and a vascular plaque; and obtain segmented vascular imaging data from the deep learning model, wherein the segmented vascular imaging data comprises a segmented lumen and a segmented vascular plaque in the volume of interest.

First claim

What is claimed is:

1. A non-transitory computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform operations, comprising:
accessing vascular imaging data for a patient, wherein the vascular imaging data comprises a volume of interest;
pre-processing the vascular imaging data to generate pre-processed vascular imaging data;
providing the pre-processed vascular imaging data to a deep learning model trained to segment a lumen and a vascular plaque;
obtaining segmented vascular imaging data from the deep learning model, wherein the segmented vascular imaging data comprises a segmented lumen and a segmented vascular plaque in the volume of interest;
computing one or more plaque attributes associated with the segmented vascular plaque, wherein the one or more plaque attributes comprise a plaque thickness, a plaque depth, and a plaque arc angle; and
generating a score based on one or more plaque attributes of the segmented vascular plaque, wherein the score is indicative of whether or not the segmented vascular plaque will limit stent expansion.

2. The non-transitory computer-readable medium of claim 1 , wherein the volume of interest is identified by an additional deep learning model trained to identify one or more potential volumes of interest.

3. The non-transitory computer-readable medium of claim 2 , wherein the operations further comprise performing at least one of:
a morphological opening operation to remove at least one isolated frame associated with the one or more potential volumes of interest, or
a morphological closing operation to include at least one missing frame associated with the one or more potential volumes of interest.

4. The non-transitory computer-readable medium of claim 1 , wherein pre-processing the vascular imaging data comprises log transforming the vascular imaging data to convert multiplicative speckle noise to additive speckle noise.

5. The non-transitory computer-readable medium of claim 4 , wherein pre-processing the vascular imaging data comprises filtering the log transformed vascular imaging data to reduce the additive speckle noise.

6. The non-transitory computer-readable medium of claim 1 , wherein the operations further comprise employing conditional random fields to the segmented vascular imaging data to reduce noise in the segmented lumen and the segmented vascular plaque.

7. The non-transitory computer-readable medium of claim 1 , wherein the segmented vascular plaque comprises one or more of a segmented lipidous plaque or a segmented calcified plaque.

8. The non-transitory computer-readable medium of claim 1 , wherein the vascular imaging data is represented as (r, θ) data.

9. The non-transitory computer-readable medium of claim 1 , wherein the operations further comprise:
operating the deep learning model to generate a vector of class probabilities at each pixel of the vascular imaging data during segmentation; and
operating a conditional random field to operate on each pixel to generate a final class ownership of the pixel, wherein the conditional random field uses the vector of class probabilities generated by the deep learning model and pixel intensities to determine the final class ownership.

10. A non-transitory computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform operations, comprising:
accessing vascular imaging data for a patient, wherein the vascular imaging data comprises a volume of interest;
pre-processing the vascular imaging data to generate pre-processed vascular imaging data;
providing the pre-processed vascular imaging data to a deep learning model trained to segment a lumen and a vascular plaque;
operating the deep learning model to generate a vector of class probabilities at each pixel of the vascular imaging data during segmentation;
operating a conditional random field to operate on each pixel to generate a final class ownership of the pixel, wherein the conditional random field uses the vector of class probabilities generated by the deep learning model and pixel intensities to determine the final class ownership;
obtaining segmented vascular imaging data from the deep learning model, wherein the segmented vascular imaging data comprises a segmented lumen and a segmented vascular plaque in the volume of interest;
computing one or more plaque attributes associated with the segmented vascular plaque, wherein the one or more plaque attributes comprise a plaque thickness, a plaque depth, and a plaque arc angle; and
generating a score based on one or more plaque attributes of the segmented vascular plaque.

11. The non-transitory computer-readable medium of claim 10 , wherein pre-processing the vascular imaging data comprises generating one or more rotationally offset versions of the vascular imaging data.

12. The non-transitory computer-readable medium of claim 10 , wherein pre-processing the vascular imaging data comprises pixel shifting or converting multiplicative speckle noise to additive speckle noise.

13. The non-transitory computer-readable medium of claim 10 , wherein pre-processing the vascular imaging data comprises log transforming the vascular imaging data to convert multiplicative speckle noise to additive speckle noise.

14. The non-transitory computer-readable medium of claim 13 , wherein pre-processing the vascular imaging data comprises filtering the log transformed vascular imaging data to reduce the additive speckle noise.

15. The non-transitory computer-readable medium of claim 13 , wherein the vascular plaque comprises one or more of a lipidous plaque or a calcified plaque, and wherein the deep learning model is trained to segment the lumen and the one or more of the lipidous plaque or the calcified plaque.

16. The non-transitory computer-readable medium of claim 10 , wherein the vascular imaging data is represented as (r, θ) data.

17. The non-transitory computer-readable medium of claim 10 , wherein the vascular imaging data comprises intra-vascular optical coherence tomography (IVOCT) data.

18. A non-transitory computer-readable medium storing computer-executable instructions that, when executed, cause a processor to perform operations, comprising:
accessing vascular imaging data for a patient;
pre-processing the vascular imaging data to generate pre-processed vascular imaging data;
providing the pre-processed vascular imaging data to a deep learning model trained to generate segmented vascular imaging data comprising segmented plaque;
computing one or more plaque attributes associated with the segmented plaque, wherein the one or more plaque attributes comprise one or more of: a lumen area, a plaque thickness, a plaque depth, or a plaque arc angle;
generating a score based on the one or more plaque attributes, wherein the score is indicative of whether or not the segmented vascular plaque will limit stent expansion; and
generating a treatment recommendation based on the score.

19. The non-transitory computer-readable medium of claim 18 , further comprising:
generating initial segmentation results of the pre-processed vascular imaging data; and
generating the segmented vascular imaging data by performing refinement of the initial segmentation results via conditional random field (CRF) using information from both an image intensity and a probability map.

20. The non-transitory computer-readable medium of claim 18 , wherein the deep learning model is a convolutional neural network.

Assignees

• Univ Case Western Reserve

Inventors

• Wilson David L
• Gharaibeh Yazan
• Prabhu David
• Lee Juhwan
• Kolluru Chaitanya

Classifications

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/096

  Transfer learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/091

  Active learning · CPC title