Home Synopsis Contact_us

           

Image guided radiotherapy with control engineering

 

Introduction

The European integrated project MAESTRO (Methods and Advanced Equipment for Simulation and Treatment in Radiation Oncology) aims to provide a European alternative to American funded research on image guided radiation therapy (IGRT). Unique to the project is the combination of radiotherapy physics, medical imaging and control engineering to improve the behaviour of current radiotherapy treatment devices.

Overall aims and objectives

To develop new control techniques to enable radiotherapy treatment equipment to move accurately and adapt to detectable organ and patient motion

Organ motion prediction

Facts:

There are delays between organ motion and machine reaction:

 MLC, gantry, PSS are electromechanical devices with limited response time (0.12s to 0.33s)
 Lags are introduced by imaging systems (0.03s to 0.09s for marker tracking, and larger for image or volume imaging)

Problem:

By the time the MLC or PSS reaches a desired position, the organs have moved.

Solutions:

- Identify system limitations, delays/lags,
- Predict organ position
- Re-design the control system to account for system’s delays and dynamics.

 

Approach to organ motion compensation

i) Analyse the performance of Elekta, Varian and Siemens equipment through a series of test campaigns involving the University Hospitals Coventry and Warwickshire NHS Trust (UHCW), Coventry, the Newcastle General Hospital, Newcastle, the Western General Hospital, Edinburgh and the Addenbrookes Hospital in Cambridge

ii) Develop representative models of the system to be controlled.

iii) Utilise the models developed to design new model based control strategies to improve the speed of response and accuracy of motion.

iv) Using predicted as well as measured organ motion, calculate the ‘reference’ that must be followed by the device controlled (e.g. MLC, PSS, gantry) to minimise influence of organ motion on treatment quality.

   

IGRT Phantoms

i)   Identify type, range, speed, acceleration of organ motion.

ii)   Design phantoms for various treatment sites.

iii)  Program control system to move phantoms.

iv)  Assess accuracy of phantoms.

v)  Use phantoms to assess IGRT and generate test images.

 

 

Organ motion modelling and prediction


i)    Understand the ‘system’ via organ motion analysis

ii)   Develop models to understand the system and generate organ motion trajectories for the phantoms

iii)  Develop predictors

     - Neural Networks, a powerful ‘black box’ predictor if used appropriately

    - Model based prediction such as Kalman filter, Interactive Multiple Models, bilinear filters and polynomials.

iv)  Evaluate the predictors

 

 

Results and Conclusions

The PSS hysteresis was found to be insignificant as opposed to the table top deflection with some of the newest carbon fibre units being the worse performers.

The gantry motion circularity was found to be adequate (within 0.3mm RMS), with the size of the mechanical isocentre being between 0.8mm and 1.4mm diameter.

The PSS model developed includes the effect of deflection and nonlinearities such as friction.

The best average predictor for 0.2s ahead was found to be the artificial neural network (ANN) followed by the interactive multiple model (IMM) filter and the Kalman filter with RMS error prediction of 0.6, 0.98 and 1.1mm respectively. IMM is much simpler to implement than the ANN developed and does not lead to large maximal error that may occur when extreme breathing patterns occur.

A preliminary control system was developed and gave rise to promising results.

 

Acknowledgements

The Authors are thankful to the Physics staff at Addenbrookes, Western General and Newcastle General Hospitals for their help in realising the experimental work presented in this poster and to the Virginia Commonwealth University, USA and the Hokkaido University Hospital, Japan for providing organ motion data.

For more details, please use the following link : Coventry publications