navigation in robotic radiosurgery achim schweikard

Navigation in Robotic

Radiosurgery

Achim Schweikard

beam

Stereotactic Radiosurgery

Conventional Procedure

Stereotaxic Frame

Arc motion

Local Anesthesia

Concept

Image-Guidance

Camera A Camera B

BKS

LCS

Source B Source A

T

Robot System

Robot System

Tumors in the chest or abdomen move due to respiration

Respiratory Motion

Respriatory Motion

Safety Margin

Tumor

Beam

Robotic Arm

Robotic Arm

Volume of Sphere:

4/3 r3

Hysteresis

Previous Work

HD Kubo, BC Hill: Respiration gated radiotherapy treatment: a technical study. Phys. Med. Biol. 1996; 41: 83-91.

T Tada et. al.:Lung cancer: Intermittent radiation synchronized with respiratory motion. Radiology, 1998; 207:779-783.

MR Sonntag, ZW Lai: Characterization of respiratory motion for pediatric conformal 3D therapy. Med. Phys. 1996: 23: 1082.

Image-Guidance

Camera A Camera B

BKS

LCS

Source B Source A

T

Stereo X-Ray Camera System

Camera A Camera B

Gold fiducial

Infrared Tracking

Correlation

Correlation

DART

System Overview

Correlation external chest vs. external abdomen

Motion Table Phantom Study(Overall Accuracy)

Robot Tracking

Prediction

Phantom Study

First Patient Treatment

Measured Target Excursionvs. Correlation Error

Follow-up Study

Four months follow-up

Treatment

Synchrony

Major Challenges

Fiducial-less tracking

Implanting of Gold Fiducials

Respiration

Courtesy Paul Sellars, JHU

Fiducial-Less Tracking: Principle

Compute or acquire 4D CT

Principle

Compute or acquire 4D CT

Compute DRRs (2D Projections)

Principle

Compute or acquire 4D CT

Compute DRRs (2D Projections)

Live Shots (2D x-ray pair)

Principle

Compute or acquire 4D CT

Compute DRRs (2D Projections)

Live Shots (2D x-ray pair)

Find the DRR that best matches live shots

Principle

„Ground truth“

Principle

i.e. Tumor position with time stamp

Principle

Thus: same ground truth information as in fiducial-based tracking is available

Principle

Correlation + Prediction

real-time target position

Principle

CT-ScanInhalation

CT-ScanExhalation

IntermediateCT-Scan 1

IntermediateCT-Scan n...

„4D CT“

Best match determines respiratory state and therefore tumor position

Results Registration

Implementation Uses Intensity-based rigid registration 2D/3D Optimization with the Hooke-Jeeves Pattern-

Search Uses hardware-accelerated volume-rendering

based on texture mapping as DRR generation technique

Used metrics are correlation coefficient and mutual information

Results 7D RegistrationRegistration results with DRRs from synth. scan 40 % inhaled

0,00

0,20

0,40

0,60

0,80

1,00

1,20

1,40

1,60

1,80

CT Scans (real and synthetic)

MI

Val

ues

Results 7D RegistrationRegistration results with DRRs from original scan max. inhaled

0,00

0,20

0,40

0,60

0,80

1,00

1,20

CT Scans (real and synthetic)

MI

Val

ues

Results single 2D/3D Registration

Computing time

10 seconds per registration

Conclusion

Feasible w.r.t. computing time

Feasibility w.r.t accuracy and reliability not yet established