Three-dimensional delineation of tumor boundaries via supervised machine learning algorithms and augmented reality

Patent metadata

Field	Value
Publication number	US-12100141-B2
Application number	US-202117646263-A
Country	US
Kind code	B2
Filing date	Dec 28, 2021
Priority date	Dec 28, 2021
Publication date	Sep 24, 2024
Grant date	Sep 24, 2024

Abstract

Techniques that facilitate three-dimensional (3D) delineation of tumor boundaries via one or more supervised machine learning algorithms are provided. An example embodiment includes a computer-implemented method that includes: extracting, by a computing system operatively coupled to a processor, one or more feature vectors from a time-series evolution of fluorescence distribution observed at a section of bodily tissue of interest, wherein the one or more feature vectors represent a physical model describing on-tissue dye dynamics of the section of bodily tissue; and generating, by the computing system, a classification attribute for the section of bodily tissue represented by the one or more feature vectors, wherein a pre-trained classifier designates the section of bodily tissue as a biopsy or a non-biopsy candidate through execution of the one or more supervised machine learning algorithms.

First claim

What is claimed:

1. A system, comprising:
a memory; and
a processor that executes computer-executable components stored in the memory, the computer-executable components comprising:
a feature composition component that extracts one or more feature vectors from a time-series evolution of fluorescence distribution observed at a section of bodily tissue of interest, wherein the one or more feature vectors represent a physical model describing on-tissue dye dynamics of the section of bodily tissue; and
a classification component of a classifier, wherein the classification component generates a classification attribute for the section of bodily tissue represented by the one or more feature vectors which allows the classifier to designate the section of bodily tissue as a biopsy or a non-biopsy candidate through execution of one or more supervised machine learning algorithms.

2. The system of claim 1 , wherein the computer-executable components further comprise:
a time-series generation component that generates one or more spatially distributed time-series of fluorescence intensities, from multispectral video streams generated by one or more medical imaging devices, as a result of on-tissue dye dynamics at the section of bodily tissue.

3. The system of claim 2 , wherein the computer-executable components further comprise:
an absorption coefficient estimation component that estimates one or more absorption coefficients from the one or more spatially distributed time-series of fluorescence intensities as an inverse problem for one or more Diffuse Optical Tomography (DOT) equations.

4. The system of claim 3 , wherein the absorption coefficient estimation component employs the one or more Diffuse Optical Tomography (DOT) equations as a choice of parametrization to generate the on-tissue dye dynamics in three-dimensions.

5. The system of claim 4 , wherein the computer-executable components further comprise:
a feature vector extraction component wherein the absorption coefficient estimation component analyzes one or more parameters in conjunction with additional inputs of patient metadata to generate the one or more feature vectors for time-series of fluorescence distribution at one or more voxels in three-dimensions, wherein one voxel of the one or more voxels is a data point defined on a regular grid in three-dimensions (3D) such that several voxels combine to define physical characteristics of a volume such as tissue.

6. The system of claim 1 , wherein the classification component of the classifier also assigns labels indicative of medical information to the one or more feature vectors through the one or more supervised machine learning algorithms, by using inputs in a form of clinical data, non-clinical data or patient metadata.

7. The system of claim 6 , wherein the classification component of the classifier further assigns a probabilistic score to indicate a measure of uncertainty of one or more of the labels indicative of medical information, wherein the one or more of the labels are applied using the one or more supervised machine learning algorithms.

8. A computer-implemented method, comprising:
extracting, by a computing system operatively coupled to a processor, one or more feature vectors from a time-series evolution of fluorescence distribution observed at a section of bodily tissue of interest, wherein the one or more feature vectors represent a physical model describing on-tissue dye dynamics of the section of bodily tissue;
generating, by the computing system, a classification attribute for the section of bodily tissue represented by the one or more feature vectors, wherein a classifier designates the section of bodily tissue as a biopsy or a non-biopsy candidate through execution of one or more supervised machine learning algorithms; and
training, by the computing system, a classification component of the computing system through supervised machine learning to classify a particular group of one or more feature vectors of the section of bodily tissue of interest as biopsy candidates or non-biopsy candidates.

9. The computer-implemented method of claim 8 , further comprising:
generating, by the computing system, one or more spatially distributed time-series of fluorescence intensities, from multispectral video streams representative of the on-tissue dye dynamics observed at the section of bodily tissue; and
recording, by the computing system, the time-series evolution of fluorescence distribution observed at the section of bodily tissue of interest, as a result of the on-tissue dye dynamics.

10. The computer-implemented method of claim 9 , further comprising:
estimating, by the computing system, one or more absorption coefficients from the one or more spatially distributed time-series of fluorescence intensities, wherein the one or more absorption coefficients are one or more parameters for reconstruction of unseen physical characteristics of the section of bodily tissue.

11. The computer-implemented method of claim 8 , further comprising generating, by the computing system, a model representing the on-tissue dye dynamics in three dimensions based on one or more estimated absorption coefficients.

12. The computer-implemented method of claim 10 , further comprising:
generating, by the computing system, the one or more feature vectors based on the one or more parameters and patient metadata, wherein the generating comprises generating the one or more feature vectors for the time-series evolution of fluorescence distribution.

13. The computer-implemented method of claim 8 , wherein the classification component is trained and re-trained using information from the one or more feature vectors and a corpus of labels via the one or more supervised machine learning algorithms to improve classification accuracy.

14. The computer-implemented method of claim 8 , wherein the classification component is trained and re-trained via information feedback between the classification component and a training logic component of a pre-trained classifier of the computing system, and wherein the classification component is trained and re-trained based on information provided via a device associated with one or more experts in a medical community.

15. The computer-implemented method of claim 8 , wherein the computing system outputs a description of geometric boundaries of the section of bodily tissue via a graphical user interface of a display of the computing system in a form of augmented reality.

16. A computer-implemented method, comprising:
analyzing, by a computing system operatively coupled to a processor, one or more inputs in form of multispectral videos to track one or more time-series of fluorescence profiles in real-time representative of on-tissue dye dynamics at a section of bodily tissue of interest; and
generating, by the computing system, outputs in form of a mesh representing one or more absorption coefficients at different times and a description of one or more geometric boundaries of the section of bodily tissue, based on the analyzing; and
generating, by the computing system, based on the outputs, a bio-physical model of the section of bodily tissue, wherein the bio-physical model is employable to diagnose one or more medical conditions.

17. The computer-implemented method of claim 16 , wherein the one or more inputs comprise one or more multispectral videos captured from one or more medical imaging devices and wherein the one or more time-series of fluorescence profiles represent dye dynamics at the section of bodily tissue.

18. The computer-impl

Assignees

• IBM

Inventors

• Tirupathi Seshu
• Epperlein Jonathan Peter
• Macaonghusa Pol
• Nair Rahul
• Tchrakian Tigran Tigran
• Zayats Mykhaylo
• Zhuk Sergiy

Classifications

• G16H30/40

  for processing medical images, e.g. editing · CPC title

• G06T19/006

  Mixed reality (object pose determination, tracking or camera calibration for mixed reality G06T7/00) · CPC title

• G16H50/20

  for computer-aided diagnosis, e.g. based on medical expert systems · CPC title

• G06T2207/30024

  Cell structures in vitro; Tissue sections in vitro · CPC title

• A61B5/0073

  by tomography, i.e. reconstruction of 3D images from 2D projections (A61B5/0066 takes precedence) · CPC title