Tissue Characterization: Status, Challenges and Opportunities - - PowerPoint PPT Presentation

▶
tissue characterization status challenges and
SMART_READER_LITE
LIVE PREVIEW

Tissue Characterization: Status, Challenges and Opportunities - - PowerPoint PPT Presentation

Mar 15, 2023 258 likes •482 views

AI IV-OCT Image Analysis for Tissue Characterization: Status, Challenges and Opportunities Thomas Milner Biomedical Engineering Department The University of Texas at Austin Yin-and-Yang of AI IV-OCT Tissue Classification Yang Yin Mostly

slide-1
SLIDE 1

AI IV-OCT Image Analysis for Tissue Characterization: Status, Challenges and Opportunities

Thomas Milner Biomedical Engineering Department The University of Texas at Austin

slide-2
SLIDE 2

Yin-and-Yang of AI IV-OCT Tissue Classification

AI Neural Network Mostly Unknown

Yin

Mostly Known

Yang

slide-3
SLIDE 3

Why AI I IV IV-OCT Tissue Classification?

Mostly Unknown

Yin

Mostly Known

Yang Expert IV-OCT Users

Inexperienced IV-OCT Users/Readers

Expert IV-OCT Users/Readers

  • Training Tool
  • Second Opinion
  • IVOCT Knowledge

Registry

MANY FEW

slide-4
SLIDE 4

Motivation: VH IV IVUS

VH IVUS Introduced

Clinical Trial Enrollment

  • Before VH-IVUS: 550 people/year
  • After VH-IVUS: 1800 people for year

Before VH-IVUS After VH-IVUS

slide-5
SLIDE 5

Neural Network Development Work-Flow

IV-OCT Imaging

slide-6
SLIDE 6

Status: Texas Human Heart Data Library ry

  • 42 human hearts collected
  • 79 arteries imaged
  • 63 arteries processed with histology
  • Histopathological analysis conducted in

collaboration with Deborah Vela, M.D. and L. Maximilian Buja, M.D. at the Texas Heart Institute, Cardiovascular Pathology Department

  • 54 arteries co-registered to OCT
  • 3229 histology sections matched to OCT

frames

  • Corresponding OCT frames encompass

approximately 300 million data points

  • 54 arteries fully tabulated to 8 plaque types
slide-7
SLIDE 7

IV IV-OCT Im Imaging

  • St Jude Illumien: 56 arteries (28 LAD, 28 RCA)
  • Hearts from 28 men and 14 women, average age at death 56 +/- 10 years
  • Volcano CorVue: 20 arteries (10 LAD, 10 RCA)

Arteries/Cases Pixels A-Scans Training 28 1.5 million 100,800 Testing 28 1.5 million 100,800 TCFA 17 N/A N/A

slide-8
SLIDE 8

Feature and Node Optimized Neural Networks

65 70 75 80 85 90 95 100

Feature Selection

Lipid Accuracy Fibrous Accuracy Calcium Accuracy Lipid Optimized Features Fibrous Optimized Features Calcium Optimized Features 70 75 80 85 90 95 100

Network Architecture Optimization

Lipid Accuracy Fibrous Accuracy Calcium Acuracy Lipid Optimized Architecture Calcium Optimized Architecture Fibrous Optimized Architecture

Features and network architecture must be tailored for tissue type for highest accuracy Solution: Multiple independently-optimized neural nets are used. Feature And Node Optimized Neural Nets (FANONN)

slide-9
SLIDE 9

Feature Ext xtraction

Statistical

  • Mean Value
  • Variance
  • Skewness
  • Kurtosis
  • Energy

GLCM

  • Contrast
  • Energy
  • Correlation
  • Homogeneity
  • Entropy
  • Max Probability

A-Scan Analysis

  • Attenuation
slide-10
SLIDE 10

Feature and Node Optimized Neural Networks

slide-11
SLIDE 11

Multi-Layer Neural Network Performance

OCT Histology Guided Expert Demarcation Neural Network Colorization

Histology

Fibrous Calcium Lipid

  • Virtual Histology-OCT neural

network classifier was tasked to colorize entire B-scans

  • Three clinically relevant pathological

plaque conditions were evaluated

  • Neural network colorizations

match histology

Thin cap fibroatheroma Fibrocalcific plaque Thick cap fibroatheroma

Tissue Type Sensitivity Specificity Fibrous 87% 80% Calcium 85% 87% Lipid 86% 80% TCFA 82% 76%

Sc Scal ale bar bars s 1 1 mm Samples 5L, 4L, 4R

slide-12
SLIDE 12

Macrophage Neural Network Layer

slide-13
SLIDE 13

Expanded Tissue Categories

Scale bars 1 mm

slide-14
SLIDE 14

Next xt-Genration Neural Networks

  • Sequential samplings/convolutions for automatic feature

generation

  • U-nets for rapid pixel classification
slide-15
SLIDE 15

Challenge: Resource Requirements

Source of Human Hearts Histopathology/Registration Network Optimization

Time

slide-16
SLIDE 16

Challenge/Opportunity: : Next xtGen In Instrumentation

  • Minimize Requirements for Retraining?
  • Downward-compatible instrument-

independent neural networks?

Existing Instrument Next Generation

  • PS-OCT
  • PT OCT

. . .

slide-17
SLIDE 17

1210nm Photothermal (P (PT) IV IV-OCT

  • Lipid has distinct absorption at 1210nm
  • Photothermal Contrast detect arterial lipid
  • Tested successfully on 3 arteries
  • Needs to be integrated into classification workflow

500µm

slide-18
SLIDE 18

Opportunities

  • Better Patient Outcomes
  • Enable and empower better therapies
  • Increased IVOCT use
  • Hybrid opto-electronic computing

technologies

slide-19
SLIDE 19

Summary ry and Conclusions

  • Multilayer histology-validated neural networks constructed and tested

to detect fibrous, lipid, calcium and plaque type

  • Sensitivity and specificity for fibrous, lipid, and calcium tested on non-

trained hearts is 80-87%

  • Each tissue type requires an individualized feature and node optimized

neural network

  • Approaches needed to develop downward compatible instrument-

independent neural nets for next-generation IVOCT imaging instrumentation

  • Opportunities to provide better patient outcomes to PCI
slide-20
SLIDE 20

University of Texas at Austin UT Health Sciences Center San Antonio Texas Heart Institute Vikram Baruah, BSc Aydin Zahedivash, BS, MD Austin McElroy, BSc Arnold Estrada, PhD Taylor Hoyt, BSc Andrew Cabe, BSc Meagan Oglesby, BSc Marc Feldman, MD Deborah Vela, MD Maximilian Buja, MD Piotr Antonik, PhD

slide-21
SLIDE 21