Spine Recognition: From Medical Image Analysis to Image- Guided - - PowerPoint PPT Presentation

spine recognition from medical image analysis to image
SMART_READER_LITE
LIVE PREVIEW

Spine Recognition: From Medical Image Analysis to Image- Guided - - PowerPoint PPT Presentation

Spine Recognition: From Medical Image Analysis to Image- Guided Interventions : Yunliang Cai Background Spine structure: Contain 24 bones (vertebra)


slide-1
SLIDE 1

Spine Recognition: From Medical Image Analysis to Image- Guided Interventions

脊柱骨骼识别:从医学图像分析到图像引导介入治疗

Yunliang Cai

slide-2
SLIDE 2

Background

  • Spine structure:
  • Contain 24 bones (vertebra)
  • Strong but highly deformable
  • Combined w/ bone, soft tissue, nerve …

胸椎 腰椎 头椎 椎体 棘突 橫突 脊髓

slide-3
SLIDE 3

Background

  • Some common spine problems
  • Spondylolysis 腰椎脱离
  • Fractures 椎体骨折
  • Deformity 畸形
  • Degenerative disc 椎间盘退化
  • Some spinal surgeries
  • Spine fusion 融合手术
  • Discectomy 椎间盘切除
  • Laminectomy 椎弓切除

Spine Fusion

slide-4
SLIDE 4

Background

  • How to automate diagnosis?
  • How to automate surgery?
slide-5
SLIDE 5

Computational Spine Model

脊骨计算模型

slide-6
SLIDE 6

Computational Spine Model

  • Our Approach --- Computational Anatomy

Parameterized Computation Model

Multi-Modality/View/Subject

Optimal Geometric Parameters

For Diagnosis For Surgery Guidance

Deform & Match Extract

slide-7
SLIDE 7

Computational Spine Model

  • Hierarchical Model Design
slide-8
SLIDE 8

Computational Spine Model

  • Spine Image Analysis: Basic Workflow
  • Y. Cai et al, TMI 2015
slide-9
SLIDE 9

Spine Image Analysis

脊骨图像分析

slide-10
SLIDE 10

Spine Image Analysis

  • Local Appearance Module
  • Use Multi-Modal Deep Net for Cross-Modality Detection

Cross Modality

slide-11
SLIDE 11

Spine Image Analysis

  • Global Geometry Module
  • Use universal 3D CAD model for spine shape analysis
  • Register by point registration method (Coherent Point Drift,

Myronenko et al PAMI 2010)

Spine Shape Analysis

slide-12
SLIDE 12

Spine Image Analysis

  • Local Geometry Module
  • Use Lie-Group-based Congealing for vertebra pose

estimation (Y. Cai et al TIP 2013)

  • Use tri-planar estimation for 3D pose

Vertebra Pose

slide-13
SLIDE 13

Spine Image Analysis

  • Multi-view, multi-modal, and whole spine recognition
  • For diagnosis of spondylolysis, deformity, …
slide-14
SLIDE 14

Image-Guided Spinal Surgery

图像引导脊骨手术

slide-15
SLIDE 15

Image-Guided Spinal Surgery

  • Why need it?
  • Precision!!
  • How to?
  • Patient Registration:

Pre-Operative Images (手术前) register with Intra-Operative Structure (手术进行时)

?

Pre-Operative CT Surgery on going…

slide-16
SLIDE 16

Image-Guided Spinal Surgery

  • Pre-Operative Modalities:
  • CT, MR
  • Intra-Operative Modalities:
  • Ultrasound 超声, Fluoroscopy 透视, Cone Beam CT 锥束CT
slide-17
SLIDE 17

Image-Guided Spinal Surgery

  • Our Pre-Op Images: CT
  • Our Intra-Op Images: Stereovision(SV)

CT SV images SV surface 立体视觉 Microscopy

(拍摄用)

Stealth Station

(实时跟踪用) Pre-Op CT

  • S. Ji et al, TBME 2015

Radiation Free

slide-18
SLIDE 18

Image-Guided Spinal Surgery

  • Our method: Rectify the spine to a “deformation-free” pose

NLPCA Flattening Inverse NLPCA

CT

NLPCA Flattening Inverse NLPCA

SV

Non-linear PCA: Auto-associate Network (Auto-Encoder Network)

  • M. Scholz et al LNCS 2007
slide-19
SLIDE 19

Image-Guided Spinal Surgery

  • Y. Cai et al, SPIE 2016
  • 2D image reg. (faster!)
  • Dense point-point correspondence

(more accurate)

slide-20
SLIDE 20

Demo

Accuracy Test: Err < 1.5 mm ! Image-Guided Drilling

slide-21
SLIDE 21

Future Directions: Spine Image Analysis

  • Explore on Soft Tissue
  • Explore on Spinal Cord

PET/CT SPECT/CT Diffusion Tensor Imaging Analysis for: Cancer Infectious Disease Trauma … Analysis for: Neural Trauma Vascular Function … CT Angiography PET/MR

slide-22
SLIDE 22

Future Directions: Image- Guided Interventions

  • Multimodality Integration
  • Towards Robotic Surgery

CT + Ultrasound + SV + ….?

Mazor Robotics (Technion)

微创手术机器人 (第三军医大学新桥医院 中科院渖阳自动化研究所)

Surgical Robot Assistant (Wright State Univ.) AQRate Robotic System (EPFL)

slide-23
SLIDE 23

Conclusions

  • Spine Image Analysis
  • Computational spinal model
  • Cross-modality vertebra detection
  • Spine shape analysis
  • Vertebra pose estimation
  • Image guidance spinal surgery
  • CT-SV image guidance system
  • Geometric rectification for deformed spines
  • Image-guided spinal fusion surgery
  • Future work
  • Towards hybrid modality analysis
  • Towards robotic spinal surgery
slide-24
SLIDE 24

Acknowledgment

  • Research led by
  • Surgical team
  • Projects funded by

PI: Prof. Shuo LI PI: Prof. Songbai JI

  • Dr. Scott S. Lollis, MD

(Primary Surgeon)

R21 NS078607

slide-25
SLIDE 25

Thank You!

  • Q & A