Msc Projects List From O Haas 2011

8/6/2019 Msc Projects List From O Haas 2011

1/26

O Haas projects

* indicate projects that have priority for my research

*Projects 1 to 4: MAESTRO phantom (LabVIEW, Simulink/dSPACE, micro-controller)

Program an electromechanical system to replicate breathing motion of human torso

Project 5: Organ motion modelling and prediction

Develop a program to combine existing motion models and motion prediction to improve accuracy of realistic irregular breathing motion patterns.

*Project 6: Simulink based neural network for motion prediction

Implement in Simulink a neural network algorithm developed in Matlab and try to improve itsperformance

Project 7: Template matching algorithm for motion prediction

Predict future position based on previously seen behaviours.

*Project 8:correlation model for external internal breathing motion

Review and implement correlation models that are used to infer internal organ motion from external

motion measurement for patient lying on a patient support system undergoing radiotherapytreatments.

*Project 9: Betatron treatment evaluation

Review existing models developed to simulate the delivery of radiation using a prototype machinedeveloped at the University Hospital with the support of CTAC. Develop a new improved model able toreplicate measurements carried out at the University hospital.

Project 10: Target tracking with Simulink - Vehicle velocity identification using on boardcameras

Use Simulink together with a camera to perform image processing tasks.

* Project 11 Agent based Controls in Service Oriented Architectures (SOA): Traffic FlowSimulation Studies in Intelligent Transportation Systems

investigate various traffic scenarios that can be applied in the simulated environment of WD- iTS. The

aim is to replicate current congestion at selected points in the city and simulate them before trying tofind the means to reduce congestion.


2/26

Control of radiotherapy phantom using a LabVIEW or an electronics interface

The aim of the phantom is to evaluate the tracking error of radiotherapy equipment that take intoaccount the patient motion and react to it. Typical patient breathing motion are of frequency 0.3Hz butcan vary from 0.2 to 0.7 Hz with velocity in the range 0.5mm/s or 65 mm/s. The phantom needs toreplicate these possible velocities.

The aim of the project is to design using a microcontroller or a PC based LabVIEW system a program to

drive four stepper motors to be able to realise any arbitrary motion of amplitude longitudinal=80mm,lateral=50mm, vertical = 80mm and ribs = 20mm.

The four motors need to be synchronised such that their relative motion can be specified (there couldbe a phase shift between each motor).

The trajectory to be programmed include:

patient data, sinusoidal motion with different frequencies ranging from 5 breaths-per-minute (bpm)

(0.085Hz) to 30 bpm (0.5Hz), and a fixed amplitude between 5 mm and 30 mm. The 30 bpm incombination with a 20 mm amplitude would lead to a speed of 60 mm/s.

(Lujan et al, 1999) Phys Med Biol. 2008 September 21; 53(18): 4855 4873. t o model respiratorymotion:

Where

A 0 is the amplitude in mm

t is the time in second

is the period to be selected by the user between 2.5s and 5.5s: [2.5,5.5]

is the starting phase angle to be selected by the user between - 90 and +90 (be carefulwith degree radian conversion required)

n is An integer parameter n , n = [2,6].

Bilinear model (CTAC)

There are currently four distinct projects associated with the phantom. Some of the information

provided below may only be useful for some of the project.

Material:

The MAESTRO Phantom

The thorax phantom is part of the MAESTRO project: Methods and Advanced Equipment for Simulationand Treatment in Radio Oncology. The project groups together 25 institutions (Laboratories,Companies) from several European countries. In Coventry, the CTAC and the university hospital takepart in the project.The present project, MAESTRO, proposes innovative research to develop and validate in clinicalconditions the advanced methods and equipment needed in cancer treatment for new modalities inhigh conformal external radiotherapy employing electrons, photons and protons beams.
http://www.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&retmode=ref&cmd=prlinks&id=18711250http://www.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&retmode=ref&cmd=prlinks&id=18711250http://www.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&retmode=ref&cmd=prlinks&id=18711250http://www.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&retmode=ref&cmd=prlinks&id=18711250http://www.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&retmode=ref&cmd=prlinks&id=18711250http://www.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&retmode=ref&cmd=prlinks&id=18711250


3/26

The project includes different work packages on research and development activities, and of trainingand management. The thorax phantom takes place in the first work package, which is: AdaptiveRadiation delivery, tracking and control for radiotherapy.In this first chapter, we will describe the Phantom, the stepper motors, and explicit the problemdefinition.

The Phantom is constituted of two parts:- A three dimensional (3D) drive system for tumour motion- A rib cage motion system

Figure 1: Picture of the whole Phantom; 1: 3D drive system, 2: Rib cage and 3: LabVIEWDevice

Figure 2 Unislide, LG Motion, Basingstoke, UK

The phantom includes two static lungs and the anthropomorphic part includes the trachea, spine andrib cage.

Tumours of various sizes can be mounted on a rod that passes into the lungs. This rod is connected toa linear drive system positioned outside the anatomical phantom part that can move the tumour in

1

3

2

1

3

2

Limit switch

Limit switch

Centre position switch

Limit switch

Limit switch


4/26


5/26

Accuracy of current systemFor each motor, the distance travelled for a motion of one step is different. We will call this variableStepInc and we will see after how we calculate it.The values of StepInc in millimetres are given in the following table:

X motor Y motor Z motor Rib motorStepInc (mm) 0,017 0,005 0,015 0,016

Table 1: The theoretical values of Stec Increment

Thanks to the previous plot, we can know what is the maximal slope for each motor :

Maximum speed of current system (obtained for sampling of 1ms)

Motor X Motor Y Motor Z Motor ribMax Speed(mm.s-1)

16.638 4.975 15.028 16.000

Table 2 : Maximum Speed for each motor, Ts = 1ms

The different speeds are linked to the value of the Step Increment for each motor:

MaxSpeedTs is the new value of the maximal speed of the motor with the sampling time Ts in seconds.

MaxSpeed0.001 is the previously calculated value of the maximal speed of the motor with Ts =1ms

Limits SwitchesOn each axis, they are three limit switches: two at both ends and one in the middle. The two switchesplaced at both ends enable us to know if the motor reaches the end. The one in the middle can be used

to initialise the motor position.

The following diagram describes how to connect the different switches.

Figure 4 : Limit switches connections

We need a voltage to detect. For example normally the input would float at a voltage V (here 5V) andthen when the switch is made it will take that voltage down to 0V through a resistor.

5V

10k

0V

InputSwitch


6/26

Issues with current design

Software improvements and redesign:

Calculate current expected position and display it for each axis (in LabVIEW). Use sensor to feedback position of drive Use limit switches (see Limit switches Section that follows) (in LabVIEW). Simplify user interface and improve its design(in LabVIEW).

Investigate Matlab/Simulink + micro-autobox to perform control of motors Use a micro controller programmed in C or auto-coded from Matlab / LabVIEW

Limit switches are not used in the software

Use limit switches to stop motor when reaches the limit Use middle position switch to initialise motor position Use limit switches to compensate for drift by comparing expected position with current possition

Positional Drift when steps are missed due to mechanical assembly

new mechanical assembly of 3 drives X, Y and Z Control Solution: use feedback to compensate for drift.

Motion of Y axis too slow similarly all other axes are also too slow

New gearing of motor drives New drive systems able to move motors at high velocity more consistently.

Long period of operation without user intervention: The phantom should operate without failure anduser intervention for 40 minutes and with typical duration of 15 min. Investigate duty cycles of components currently used.

Improve robustness of mechanical design and software Develop a technique to remote control the phantom (using remote desktop or something similar

to replicate user interface) or longer wires between DAQ and motor. Add a video signal Add sensor feedback (remotely observable) to observe what the phantom is doing

Mechanical design

Move power supply to rib motor such that the cable is not in the field of view of the scanner Check rib motion mechanism and increase its range as well as anterior posterior

displacement (vertical motion)

Problem Definition

Control the different motors to follow a specific trajectory (sinewave or real patient breathing data).The implementation of the motor control will be done either using LabVIEW or Matlab Simulink dSPACEor a microcontroller. If a microcontroller solution is selected it should be possibly to program differentmathematical expressions as trajectories to follow.


7/26

The requirements are:- The respect of the sampling time between each samples sent to the motors- The respect of the trajectory (amplitude)- The synchronisation of the motors

The range of motion in the different direction to be achieve are- 32mm in superior-inferior direction- 29mm in anterior-posterior direction- 25mm in lateral direction

The maximal speed should be about 30 mm/s but ideally peak at 60mm.s -1 for short period of operation to be able to simulate fast and irregular breathing motion.

Operation:

The phantom must move to repeat the given trajectory until the user decides to stop it.

At this stage, it is necessary to remember the position of the phantom such that it can be restartedfrom where it stopped or moved back to the start position.

A number of switches are used to indicate the maximum travel of the slide moving the phantom andtransforming the motor rotational motion into a translational motion.

These switches can be used to initialise the start position. If a switch is activated, the phantom shouldstop immediately.

The trajectory to be given to the phantom should be checked (possibly using another program) toensure that they are feasible and will not lead to the activation of a limit switch.

Phantom Project 1: Improvement of LabVIEW software for current phantom design

Literature search

Review and critical evaluation (from publicly available information) of existing phantom on themarket

Review methods to drive motors to achieve high duty cycle and good accurtacy andrepeatability

Review method to control mechatronic systems

Review existing code and evaluate components that can be reused

User interface redesign

Design a user interface on paper based on required user input and outputs/displays

Control of motors and limit switches

Calculate current axes position and feed the information back to the user. Calibrate phantom to verify position calculated with position measured Implement software limits using limit switches Use limit switches to compensate for drift Use encoder (e.g. HEDS-5701 G00 AVAGO) to monitor drift and compensate for it

Evaluate phantom controller in CTAC (bench test) prior to its demonstration at University Hospital.

If these tests are successful it is planned to travel with the phantom to Brussels in the summer.


8/26

Phantom Project 2: Move PC+DAQ solution to a PC+FPGA solution using LabVIEW forcurrent phantom design

Similar literature searches than Project 1 with in addition research on FPGA.

Replicate the software developed to be used with USB DAQ systems such that it can exploit the realtime capabilities of LabVIEW FPGA.

Program the 4 motors using LabVIEW fpga and reuse components from existing program for otherfunctionalities.



Phantom Project 3: Electronic drive system design for 4 stepper motors

Similar literature searches than Project 1 with in addition an evaluation of different methods andassociated costs to drive slides using steppers or AC motors. Investigate the means/cost to implementposition feedback.

Electronic component realisation on breadboard and PCB implementation

Programming of microcontroller to move stepper motors



Report describing the programming

Support: A technician may be available at the hospital to help with the controller programming and theselection of the appropriate components. Changes to the design should be discussed and impact forproduction of several units should be evaluated. It is aim to develop a commercial product. Thereforewhilst academic software licences can be used for this project work it would be useful to know theimpact of licensing on using it to develop a commercial product.

Phantom Project 4: Mechanical redesign of phantom

Change design of 3D slider and phantom rib cages.


9/26

Project 5: Organ motion modelling and prediction

Develop a program to combine existing motion models and motion prediction to improve accuracy of realistic irregular breathing motion patterns.

Project 6: Simulink based neural network for motion prediction

Aim:

implement an algorithm developed for motion prediction in Simulink based on previous MSc project.Adapt the neural network algorithm to improve its performance to enable it to work using real timeworkshop. Investigate different prediction horizon and training window to establish the mostappropriate for systems with 10 Hz and 30 Hz sampling.

Material

Matlab and Simulink neural network toolbox

Previously developed neural networks programs in Matlab

Deliverable

Review of current motion prediction used in radiotherapy

Report describing the algorithm and analysing the results obtained

Software that can run on dSPACE to perform motion prediction based on 1D position against time or

3D position against time (optional).


10/26

Project 7: Template matching algorithm for motion prediction

The aim of the project is to develop an efficient algorithm, inspired from neural network, to predict futuresurrogate motion using current as well previous surrogate positions. The general idea is to find a patternof position stored in a look up table that would match the pattern observed between time t= t now and t =t now t window_size and assume that if the pattern is matched then the evolution of the motion will be similarto what has been observed in the past. If no match is found simpler algorithms based on linear orpolynomial extrapolation should be used. Finally the prediction provided to the user should be a

combination between the individual motion prediction algorithm (your new algorithm, linear extrapolationand polynomial extrapolation as well as Kalman filter). Algorithm developed in CTAC will be provided andused for comparison.

Aims:

Develop an algorithm to predict future motion based on the previous and current motion.

Objectives:

Generate look up tables. Using a moving window strategy, generate two look up tables, thefirst, referred to as INPUT,to store pattern of past surrogate motion data, the second, referred toas OUTPUT to store the corresponding predicted position

Detect patterns and ignore duplicate patterns . Look for set of duplicate patterns(considering both INPUT and OUTPUT tables at the same time) and remove them from the lookup table. This means that only data with similar pattern than stored in the INPUT look up table aswell as the same corresponding predicted position in the OUTPUT look up table will be removedor not added. Data with the same INPUT and different OUTPUT will need to be flagged for furtheranalysis. You should investigate what happen if the window size or data sampling is changed.

Implement pattern matching algorithm . Use a pattern matching prediction algorithm to findsimilar pattern in the look up table corresponding to the training data.

o If a match is found use as prediction the position in the OUTPUT look up tablecorresponding to the pattern matched in the INPUT look up table.

o If a match not found, add the most recent pattern to the INPUT look up table and then the position that

the surrogate will reach after Hp samples to the OUTPUT look up table Fit a linear or polynomial to the pattern that could not be matched . Use the resulting model

(linear or polynomial) to predict future surrogate position. Compare performance of prediction algorithms implemented . Compare the performance of

pattern matching, linear extrapolation, and polynomial extrapolation algorithms.

Automatically select most appropriate algorithm online . Create a voting system to selectthe most appropriate prediction algorithm based on the previous prediction performed as well asthe shape of trajectory.

Deliverable

Review of current motion prediction used in radiotherapy

Report describing the algorithm and analysing the results obtained

Software that can run on dSPACE to perform motion prediction based on 1D position against time or

3D position against time (optional).


11/26

Project 8: correlation model for external internal breathing motion

Review and implement correlation models that are used to infer internal organ motion from externalmotion measurement for patient lying on a patient support system undergoing radiotherapytreatments.

Deliverable: literature survey

Implement one of the model for example the one described in the following paper


12/26

Project 9: Betatron treatment evaluation

Review existing models developed to simulate the delivery of radiation using a prototype machinedeveloped at the University Hospital with the support of CTAC. Develop a new improved model tocalculate the dose received by the patient or a phantom such that they replicate measurements carriedout at the University hospital.

Total Skin Electron Betatron Treatment Unit.at

The Arden Cancer Centre, University Hospital, COVENTRY, UK

John MILLS PhD 1 , Gareth BAUGH MSc 1 , John MACLEOD MBIDT 2 ,

Vladimir CHAKLOV DSc 3 , Olivier HAAS 4 PhD

and Robert GRIEVE FRCR 5

1 Department of Clinical Physics and Bioengineering, University Hospital, COVENTRY

2. JME Engineering Ltd, Electron House, LOWESTOFT

3. Institute of Introscopy, Tomsk Polytechnical University, TOMSK.

4. Control Theory Applications Centre, Coventry University, COVENTRY.

5. Radiotherapy Department, University Hospital, COVENTRY.

1. Introduction

Total Skin Electron Therapy (TSET) has become a widely recognised treatment method for Mycosis Fungoides. Thetechnique aims to treat the superficial layer of the dermis, up to 5mm with local boosts at appropriate energy toaccommodate thicker lesions. Additional boosts may also be added to regions which have received a low dose whenthe patient was treated in totality. Other cutaneous diseases have been treated with TSET but none other has seenthe success and patient numbers as Mycosis Fungoides.

All TSET techniques combine beams in order to achieve adequate coverage of the patient surface with sufficientpenetration. Inevitably these result in non-uniform dose due to the patient surface as well as the variation in beamcharacteristics. Nevertheless, despite these shortcomings the technique has acquired widespread use.

All current treatment techniques are based upon linear accelerators with typically source to surface distances greater than 2m. The initial Stanford technique is a six patient position, 12 field technique. It is extensively used with smallmodifications usually involving degradation of the beam with a sheet of material such as PMMA. Other adaptationshave involved an arcing beam and a patchwork field matching technique which can be tailored to the variations of thedisease along the patient. Another technique which rotates the patient is also used to produce a very uniform dosedistribution around the patient. Only the translational technique and the arcing technique, both from the ChristieHospital in Manchester permit the patient to lie on a couch for the treatment. All the other Stanford based techniquesrequire the patient to be standing upright which can prove difficult for some patients.

The Compact Medical Betatron has provided an accelerator which can be translated along the length of the patientand this note describes the Total Skin Electron Therapy system developed with it in Coventry.

2. Betatron

The Betatron used is a 10MeV Clinical Medical Betatron and was supplied by JME Ltd of Lowestoft in the UK. Themachine was constructed by the Institute of Introscopy of the Polytechnical University of Tomsk in Russia. The


13/26

machine was originally supplied for Intra- operative use at the St Bartholomews Hospital in London but due tounforeseen reasons it was never used for that purpose. At a setting of 10MeV the machine produced a doserate of 200cGy/min at 1m from the source and could provide flattened field sizes up to 20 x 20cm square. The energy of thebeam at the patients surface was 8.5MeV. JME Ltd of Lowestoft provided the machine to Coventry in 1998 for thepurpose of developing a dedicated TSET treatment machine.

For use in TSET at Coventry it was found that the energy required was approximately 5MeV to give the requiredpenetration for four oblique fields around the patient, Fig 1.

Figure 1 The configuration of beams used for TSET in Coventry.

For TSET use there are two conflicting aims for the machine in order to reduce the treatment time. The first is tomaximise the field size while the second is to maximise the doserate. These conflict as the former requires the sourceto patient distance to be increased while the second requires it to be decreased. The field size was also increased bythe introduction of a scattering system. However this required careful design in order to maximise the spread of thebeam and hence the field size while minimising the reduction in the doserate due to losses in the scatterer.

Finally after extensive experimental work a compromise was found for the system involving scattering system, beamenergy, initial doserate, distance to the patient and patient coverage. As part of this the machine was specially tunedin Tomsk to provide the required 5MeV energy with a maximum dose rate. The final doserate at the patient surfacewas 500cGy/min

The control unit and operational system for the Betatron was designed by the Institute of Introscopy to comply with thesafety requirements of the International Electrotechnical Commission document IEC60601-2-1:1998.

In 2009 a new accelerating tube was delivered and fitted by an engineer from JME. Following the install, the beamwas adjusted and steering altered to restore the initial characteristics. This was the first occasion the tube had beenchanged on the machine since its delivery in 1998.

3. Gantry and Motion System

A Gantry support system with a traversing unit was designed and built by Apprentices from the local Rolls Royce(Industrial and Marine) Factory. This contribution was of enormous benefit to the project as well as providing aninteresting application for the apprentice engineers and craftsmen. The initial plan is shown in Figure 2 with the gantryunder construction in Figure 3.

Beam 1 Beam 2

Beam 3Beam 4

Beam 1 Beam 2

Beam 3Beam 4


14/26

Figure 2 The Gantry and Traversing Unit design.

Figure 3 Gantry and Traversing Unit under construction.

The traversing unit which held the Betatron was moved by a toothed belt driven by a stepper motor. The stepper motor was controlled by a TrioMotion controller system programmed from a PC.

The programme for the controller was developed by the Control Theory Applications Centre of Coventry University. Of paramount importance was the safety aspect of the patient treatment and the software was written with this in mind atall times.

An interlock was built into the system in order that the machine could be run for physics and engineering tests withoutthe full motion of the unit. This interlock system was built with interaction with the controller software so that for treatment it could only be operated with motion of the traversing unit.

4. Safety Interlocks

The patient safety was considered with respect to the software development of the controller and a watchdog systemwas built in to monitor for unexpected events However an independent hard wired safety interlock was also developedand put into place to protect the patient from machine failure. Two failure modes were considered; Failure to move


15/26

with continuing radiation and failure to irradiate with continuing movement. An independent detector system was builtfrom discrete components which constantly monitored for motion and radiation. Failure of either separately or bothtogether were capable of interrupting the treatment. Treatment interruption required the operator to take action toinvestigate the fault and to restart treatment if it was considered safe to proceed.

5. Noise

During irradiation it was felt that the noise from the Betatron was significant. This was confirmed by measurementsundertaken by the Sound and Vibration service at the local Jaguar Cars design centre. Following these measurementsa sound reducing housing was designed and built at Jaguar and this brought the sound level down to that of astandard commercial radiotherapy treatment machine of 65dBA.

The initial measurement was 84dBA and the final level measured in the treatment room was 69dBA. This was asignificant contribution to the project and improved the treatment conditions for the patients.

6. Dose Characteristics

Measurements of dose penetration have been undertaken in Coventry in the past for a TSET method using a linear accelerator. Always of great concern has been the variation of the penetration around the patient and how this varied.Figure 4 shows the phantom used for assessing the penetration and this allows measurements to be made at several

positions around the phantom for the entire treatment delivery.

Figure 4 Phantom to assess dose penetration around patient.

The variation in penetration is shown in Figure 5 and for comparison that for a typical Stanford technique is shown inFigure 6. It can be seen that the Stanford technique provides a more consistent depth of penetration. This is due tothe greater field intensity across the patient which could be achieved with a higher output from the Betatron.

Figure 5 Dose penetration for the Betatron.


16/26


17/26

Treatment recommenced in January 2006 and since then a further 14 patients have been treated. Surface dosemeasurements have been made on every patient and found to be very consistent. Initially the treatment was given as24Gy in three fractions. Four quadrants were treated in order to cover the patient entirely, Figure 1. Patients weretreated 4 times per week and each treatment covered a different quadrant. From June 2007 the prescription wasbrought into line with the UK Lymphoma Group dose of 30Gy in 20 fractions over 5 weeks. The entire patient surfacewas treated on each fraction. However this extended the overall treatment time by two weeks.

Patients have their eyes and nail-beds shielded throughout the treatment as routine. However in addition a clinical

decision is made about the other sites to be shielded and for what proportion of the treatment. For example inCoventry, the high dose down the side of the patient is reduced by shielding with custom lead during the last week of treatment.

8. Conclusion

The Compact Betatron has been successfully used to provide a dedicated TSET machine. There remains room for theimprovement of coverage and the uniformity of penetration with a Betatron designed for 5MeV with a greater dose-rate. The dedicated provision of this unit enables the standard treatments on the linear accelerators to continuewithout the interruption of a TSET treatment and this maintains patient throughput.

The dose distribution around the patient is comparable to the Stanford technique and all the clinical outcomes havebeen successful. The Betatron has proved itself to be a very reliable machine providing consistent dose delivery andbeam characteristics with minimal adjustment.

TRAJECTORY CONTROL FOR A NEW

TOTAL SKIN ELECTRON TREATMENT MACHINE

Olivier C L Haas*, Keith J Burnham*, John A Mills+

*Control Theory and Applications Centre (CTAC), Coventry University, Priory Street, Coventry CV1 5FB,UK, http://www.ctac.mis.coventry.ac.uk/

+ University Hospitals Coventry and Warwickshire NHS Trust, Coventry, UK

Abstract: The paper focuses attention on the development of a control system for a newradiotherapy treatment machine that makes use of a betatron to deliver electron beams over thewhole surface of a patient. The betatron moves above a patient at a fixed distance from the patientsupport system. This paper presents a method to predict the betatron trajectory, in terms of speedand position, according to variation in the patients geometry and is concerned with a simulation

study of a PID control strategy implemented to control the speed of the betatron. Copyright 2002IFAC

Keywords: conventional control, medical systems, medical applications, three term control,optimisation.

1. INTRODUCTION


18/26

Total skin electron (TSE) treatment is an effective procedure to treat skin cancer such as Mycosis Fungoides (Klevenhagen, 1985;Shank, 1998). TSE makes use of electron beams to treat the skin over the whole surface of the patient whilst, at the same time,sparing tissues underneath the skin. Indeed, electron beams interacting with human tissues only penetrate 10 to 15 mm before

being significantly attenuated. At the U n i v e r s i t y H o s p i t a l sC o v e n t r y a n d W a r w i c k s h i r eN H S T r u s t , C o v e n t r y , UK, TSE is currently deliveredusing dual energy linear accelerators. Linear accelerators are machines capable of producing X-ray photon or electron beams. Thecurrent treatment technique is, however, time and resource consuming, requiring patient specific immobilisation devices to bemanufactured and used to set-up the patients. In addition, TSE takes longer to administer than standard radiotherapy treatments.

This means that each patient treated for TSE on linear accelerators can potentially increase the number of patients waiting for themore routine radiotherapy treatment. Given the significant pressures imposed on the NHS to reduce waiting lists, this problemneeds to be addressed. It was therefore decided to build a specialised and dedicated TSE treatment machine capable of irradiatingthe whole body with an electron beam more efficiently and cost effectively than using linear accelerators.

This paper introduces the conception and design of the new TSE treatment machine, referred in the remainder of the paper as theTSE unit, and focuses on a method to predict and control its trajectory, in terms of speed and position. Section 2 presents the

background to the work and the design of the TSE unit. Section 3 describes the electron beam model developed to predict the penetration of the electron beam depending on the distance from the source, the angle of incidence of radiation and the dose rate.Section 3 also presents a derivation of the expression used to predict the speed of the trolley supporting the betatron. Section 4

presents a study on the effect of the electron beam field size and dose uniformity on the treatment time. The last section presentsthe control approach and optimal tuning of a three term proportional + integral + derivative (PID) controller.

2. DESIGN OF THE TSE UNIT

A number of elements were considered during the design of the TSE unit. These include the method of generating electron beams,the means to deliver a uniform dose over the whole body of the patient and the cost effectiveness of the new machine.

The first element selected was a machine, namely a betatron, capable of generating an electron beam at the appropriate power. A betatron is a cylindrical electron accelerator that was first used in medical applications in 1948 (Klevenhagen, 1985), Figure 1.The betatrons weight and small size together with the ability to produce 6 Mega electron volts ( MeV) electron beams at a fractionof the cost of a linear accelerator, makes the betatron an ideal tool for TSE.

Having selected the means to deliver an electron beam of sufficient power, the second element was to design a system thatcould irradiate the whole body of the patient. The emphasis was placed on robustness, safety and comfort. Following a call toenter into an open competition, organised by the CTAC, Coventry University, UK, (Hall, 1998), two alternative designs emerged.The first design, that offered the most flexibility, involved attaching the betatron to a six degrees of freedom industrial robot.However, due to the potential risk of collision with the patient and the additional stress put on the betatron, it was decided toimplement a solution based on a fixed gantry construction, see Figures 1 and 2. As opposed to a solution published in (Chretienet al ., 2000) where a variable speed translation couch was developed to move the patient under a fixed treatment machine, itwas decided that the patient support system would remain fixed. The betatron, positioned onto a trolley 140 cm above thetreatment couch, is allowed to move along the patient's head to toe axis (denoted x axis) (Guerin, 1999; Haas et al ., 1999), seeFigures 1 and 2. Such a solution was adopted to minimise the patients discomfort during treatment. The authors believe thatthis should ensure better reproducibility, minimise patients movement and reduce patient stress.

Other aspects to consider when developing a treatment machine are aesthetic and comfort factors. The betatron is noisy. Toattenuate the decibel level a special casing with noise absorbing material was constructed. A false wall was also fitted in front of the gantry and betatron so that the patient would only see the electron beam aperture.

Finally, the most relevant aspect to a control engineer is the need to design and develop a control system to control thedisplacement of the trolley supporting the betatron. To drive the trolley, the SA28 (Smartdrive Ltd 2002), a complete steppermotor system, was purchased. It includes the controller Euro205 that was chosen for its expansion capabilities, allowingsophisticated controllers to be implemented.

TSE unit during construction


19/26

Fig. 1. Illustrating the betatron and beam control system mounted on a moveable trolley positioned on rails on a purpose builtgantry

Fig. 2. Schematic representation of the set-up used to move the electron beam produced by the betatron over the whole lengthof the patient.

The trolley (1.1 m long by 0.75 m wide) is positioned onto rails (3.5 m long and 0.75 m wide) and its four nylon wheels (0.075 mdiameter) are designed to minimise the amount of friction with the rail. This means that a relatively small force needs to beapplied to move the trolley. However, care should be taken as the moving system composed of the trolley, the betatron and the

betatron control unit weights in the order of 300 kg. This is achieved by constraining the control input to maintain themomentum of the machine within acceptable bounds. Such an arrangement should prevent unnecessary strain on the motorgearbox and transmission as well as within the drive belt.It can be observed from Figure 1 that the height of the betatron cannot be modified. The only means of changing the distancefrom the beam to the patient is by modifying the height of the patient support system. To minimise the movement of thepatient during treatment, the position of the couch is adjusted once before the commencement of the treatment and kept fixedwhilst the betatron mounted onto the trolley irradiates the patient. Therefore, in this work, the distance from the couch thatsupports the patient to the betatron is assumed constant. The energy reaching the patient depends on the distance between thesource of the electron beam and the surface of the patient. Naturally, this distance between the source and the skin will varydepending on which part of the body is being irradiated and the geometry of the body contour in the ( y, z ) plane. To deliver aconstant dose, it is therefore necessary to modify the output of the betatron, referred to as dose rate, and/or to vary the time of exposition to the electron beam. Previous work in the area of beam intensity modulation has demonstrated that modulating beam intensity in real time is not atrivial problem. It requires state of the art radiotherapy machines, such as linear accelerators used with multileaf collimators or

patient specific compensators (Haas, 1999). Whilst the latter is applicable to standard radiotherapy treatment using X-rays, theneed to adapt the beam intensity profile to the patient geometry, along the y axis, across the full length of the patient makes itimpractical for TSE. Further, the betatron is designed for simplicity to ensure better robustness than complex radiotherapy machines. Such robust design does not permit modification of the beam output in real-time. Inserting compensators in the beam isnot, in this case, a realistic option as it would attenuate the electron beam too significantly. Therefore, the solution adopted in thiswork, to ensure that the patient receives a uniform dose over the whole body, is to vary the speed of the trolley for a fixed fieldsize. When the speed is low the time of exposition to radiation is large and the dose delivered to the patient is correspondinglyhigh. As the speed of the trolley increases, the exposure time decreases together with the dose received by the patient. In thefuture, it should be possible to move the betatron at various angles along the ( x, z ) plane, however the current TSE unit has a fixed

beam position. The orientation of the electron beam (30 degree from the z axis) has been calculated to offer the best compromise

between the area covered by the radiation and the accuracy of the dose delive red. Using such a set up, the treatment is delivered infour passes. During the first pass, the betatron irradiates the front right hand-side of the patient. During the second pass, the couchis rotated by 180 degrees and the left hand side of the patient is then irradiated. Then the patient is turned over and again the rightand left hand-side are irradiated.

Betatron

Betatron control unit

Gantry built to support the

Trolley

x

Betatron Gantry supportingthe betatron

Sensor to detect thesource to skin distance

Ionisation fieldArbitrary body

Patient supports stem


20/26

3. SYSTEM MODELLING

To facilitate the understanding of the challenges to be solved both in terms of modelling and control, it is assumed, in this section,that the electron beam is directly above the patient in the y axis direction. It is also assumed that the beam can be decomposed intoa number of elemental pencil beams i.e. i th pencil beam represented by a white dashed line in Figure _ 3.

3.1 Electron beam modelling

Factors considered when modelling the effect of electron beams on human body structures include field size, dose rate, distance of

the patient from the source of radiation and angle of incidence of the beam.

Fig. 3. Illustrating the concept of pencil beams and the spherical distribution of the beams energy into an assumed flat por tion of skin.

The field size is defined as the size of the field exposed to radiation. As the field size increases, the energy lost due to multiplescattering between the electrons and the air decreases. This, leads to an increase in the energy delivered to the patient. To accountfor this effect, the original energy is multiplied by a so-called output factor; denoted OF. In this work, the field size can beadjusted from 6 _ cm 2 to 40 cm 2 corresponding to output factors of 1 and 1.043 respectively (Guerin, 1999).

The dose rate is the quantity of energy delivered per unit of time. In this work, it is assumed that the betatron delivers 1.4 Gy/min.The dose rate, denoted I ref , is normalised and given at a distance of 1 meter from the source. The distance from the source of radiation to the patient is referred to as source to skin distance, denoted SSD. In this work, the SSD has been selected to be 140 cmto offer the best compromise between sufficiently large field size and dose rate at the patient surface. The dose rate I for a fieldsize associated with an output factor OF at a distance r=SSD from the source is given by

2.5

ref I OF I r (1)

The angle of incidence of the beam is the angle between each pencil beam and the normal to the patient. In this work,consideration is given to the patient head to toe axis (or longitudinal direction), therefore, the angle of incidence in the lateraldirection is not considered. The angle of incidence in the longitudinal direction, denoted is taken between the line made by the

pencil beam central axis and the slope of the patient surface in the longitudinal direction, see Figure 3. The angle of incidenceinfluences the penetration of the electron beam, with the highest penetration achie ved for an angle normal to the patients surface,i.e . = 90.

Penetration of electron beam

Distance in cm

Fig. 4. Illustrating the beam penetration with respect to the patient geometry, patient contour , dose penetration - - .

Distribution of uniform dose

skin

Air

Humantissues

Beam path

x

y

xi

SSDi

r

y0 y

Elemental pencil beam i

Electron beam path


21/26

The penetration is given by the following expression:

0 sin y y (2)

where y0 and y are, respectively, the penetration with an angle of incidence normal to the patient and a beam incident at an angle from the horizontal axis, see Figure 3. An arbitrary profile given in Figure 4 shows that for most of the body the penetration of

the beam is adequate with the exception of the area under the chin, the top of the chest and the feet.

To improve the dose uniformity, it has been suggested by (Guerin 1999) that i) patients could stretch their feet and ii) the patientsupport system could be tilted resulting in a patient surface orthogonal to the beam central axis. However, as the latter twosolutions involve moving the patients, it was proposed that the betatron should be mounted on a gantry that can be tilted such thatthe electron beam can be orthogonal to the patients surfaces.

3.2 Determining the speed of the betatron

Radiotherapy treatments such as TSE are in general delivered in a number of fractions, each fraction accounting for a part of theoverall dose prescribed. In this work, it is assumed that a dose of 24 Gy is delivered in 12 fractions of 2 Gy. A two-step processhas been adopted to calculate the trajectory of the beam that is assumed to comprise n elemental pencil beams. The first step is tocalculate the dose rate at the skin level at the point of intersection between each elemental pencil beam, denoted i =1... n, and the

patients surface. The second step is to determine the speed of the betatron assuming that the motor follows a continuousmovement with a fixed field size. As the patient is non-flat, the dose rate at the surface of the patient changes. Regions such as theneck or the legs are further away from the beam source than the chest or the abdomen and will, therefore, require longer exposition to receive the same amount of dose. In order to take this variation into account, the average dose rate for a given fieldsize, denoted I avg , due to each elemental pencil beam i for the points within the open field is calculated as follows:

1

2.52 21

1

n

i nref i

avg i

i i

I OF I I

n n SSD x (3)

with i=1.. n, is the index corresponding to the pencil beams considered, n is the number of pencil beams, OF is the output factor, I ref is the dose rate at 1 m, SSDi is the source to skin distance along the y axis, and xi is the distance from the point considered tothe beam central axis. As the trolley moves, each point in the patient receives a dose from each pencil beam (that are oriented atdifferent angles due to the beam divergence). The speed of the betatron can therefore be obtained from the average dose rate suchas:

iavg

p p

avg

x I x x x

Dt D I

(4)

where x is the distance along the x axis between two consecutive points on the body, D p is the prescribed dose and, I avg is theaverage dose rate.

4. EFFECTS OF SPEED MODULATION AND FIELD SIZE OVER TREATMENT ACCURACYHaving established a model to predict the speed of the betatron from the source to skin distance, this sub-section investigatesthe effect of speed modulation and field size on the accuracy of the dose delivered to the patient.

Given a required dose per fraction of D p=2.0 Gy delivered using a 20 cm2 field size, and knowing the length of the patient and the

average distance from the source, it is possible to deduce an average speed for the betatron Unit (here 8.8 cm/min). It can beobserved in Figure 5 that such a scheme lead to a dose received by the patient varying between 1.6 Gy and 2.3 Gy. By contrast,when variable speed is used the dose received by the patient is much more uniform, with some small areas receiving dosearound 1.8 Gy and 2.2 Gy. The improvement realised by varying the speed is significant, however, the feet together with the

head are over-dosed whilst the ankles and the neck under-dosed. Using a field size of 20 20 cm means that at any one time anarea of 20 cm 2 is irradiated. The speed is calculated from the average distance between the beam source and the regionirradiated. If the region irradiated contains body surfaces that change significantly in the y direction, then body surfaces that arecloser to the source will receive higher dose than those further away.

In an attempt to further improve the accuracy of the dose delivered to the patient, simulation studies have been performed todetermine the influence of field size and the SSD to the dose uniformity. Figure 6 illustrates the variation in terms of the dosereceived


22/26

Distance (cm) along the patient head to toe ( x ) axis

Fig. 5. Illustrating the dose received by a typical patients profile with constant (black solid line) and variable (grey dotted line)TSE unit speed.

by the skin of the patient for a fixed field size and a speed varying according to the moving average of the dose rate at thepatient level (dose represented by o, corresponding to continuous motion). It can be observed that large field sizes are no t ableto adapt to the patients profile with the dose distribution being inversely proportional to the SSD. Smaller field sizes are able toproduce a more homogeneous dose. The latter improvement is however achieved at the expense of a longer treatment time; it

is increased by a factor of more than 3 when a 6 cm2

field is used as opposed to a 20 cm2

field, see Table _1. Note that, in Table1, the numbers in brackets represent the dose distribution obtained without the additional boost.

Table 1: Illustrating the trade off between treatment time and accuracy of dose delivered

Field size (cm 2) 6 20 40

Mean Dose (Gy) +++

ooo

2.01

(1.99)

2.02

(1.99)

2.04

(2.00)

Variance +++

ooo

0.014

(0.019)

0.035

(0.051)

0.047

(0.074)

Standard deviation +++

ooo

0.025

(0.035)

0.057

(0.076)

0.082

(0.11)

Treatment time (min) 110 32 16

The foreseen increase in treatment time for small field sizes makes the latter unpractical therefore, an alternative strategy wasinvestigated. It involved delivering the dose with the largest field available and then adding dose boosts using a step and shootapproach with a smaller field size. The latter could be delivered semi continuously by stopping the betatron and changing thefield size. However, as such operation cannot be carried out instantly, the following approach is being considered forimplementation. Once the first pass has been preformed with a large field size, the machine field size is reduced to a sufficientlysmall field and the betatron unit is moved to the required locations where a dose boost is delivered.

The plots represented by + in Figure 6 show the improvement made from the additional boost method. It can also be observedthat such method is mainly useful when large field are used.

50 100 150 200

1.8

2

2.2

D o s e

i n G y

Influence of speed on dose uniformity


23/26

Dose delivered to the patient along head to toe axis

Distance (cm) along the patient head to toe ( x) axisFig. 6. Illustrating the deposition of dose for fixed field size and variable speed o and with additional boost + using a 6 cm 2 field

size.

As it is not possible to remove dose, when calculating the initial speed it may be of advantage to select the maximum speed asopposed to the mean speed. This would ensure that no region is overdosed. The latter approach has however the disadvantage of requiring greater dose boosts, which in turn slows the overall process.

5. CONTROL OF THE BETATRON UNIT

The previous sections have highlighted the complexity of the TSE unit control problem in terms of trade off between accuracy of the dose delivered to the patient and the time necessary to deliver such a dose. In this section, it is assumed that the set point (therequired speed of the betatron) has been calculated, taking into account relevant clinical requirements together with practicalconsiderations. In the actual system, additional safety constraints are considered to accommodate for possible machine

breakdown. The most serious event would be for the betatron to malfunction. When such an event is detected, the power required by the betatron to generate an electron beam is automatically halted and the trolley stopped. To detect such malfunction, the doserate is monitored constantly. When the dose rate varies within predefined limits, the speed can be adjusted to take into accountthose variations. However, when such variation becomes unacceptable, the machine is stopped and the treatment interrupted. Suchunacceptable variations together with the general health of the machine could be detected using techniques similar to those

investigated in (Haas et. al., 2001).

A Simulink TM model of the betatron and its motion has been developed. The control scheme is based on a proportional integralderivative (PID) scheme with a feedforward gain. Such a controller has been selected as it is widely available in industry. Asystem model was developed from a number of step responses of the system. A recursive least squares algorithm was used toidentify the coefficients of a second order transfer function. The latter was implemented within Simulink TM .

An optimisation scheme has been developed in-house to tune controllers according to requirements in terms of: maximum peakovershoot (MPO), settling time (ST), number of oscillations (NO), integral of absolute error (IAE) and the sum of the square of the error (SSE). The optimisation algorithm is based on Nelder-Mead simplex method implemented in the MATLAB TM 6.1(MATLAB,2001) function fminsearch.m. In this work, the following objective function was used to optimise the PID gains (Kp, Kiand Kd) together with the feed-forward gain Kf:

C = (100*MPO) 2, + (SSE)2 +(100*ST) 2+(PEN)2 (5)

where PEN is a penalty function set to an arbitrarily large value. The latter is required as the selected optimisation routine does notconstrain its solutions to be positive. The gains obtained for Kp, Ki, Kd and Kf are 0.318, 0.997, 0.193, 0.445, respectively.

The overall system was simulated in Simulink TM , with the set-point calculated to produce a uniform dose at 200 cGy using a 6 cm 2 field size, see Figure _ 6. Figure 7 shows the required and actual speed achieved by the betatron unit. To minimise mechanicalstress, the speed of the betatron unit is ramped up to the required speed with the beam turned off. Before passing over the patient,the beam is switched on and once it becomes stable the treatment starts. Once that the whole patient has been irradiated in thehead to toe direction, the speed of the betatron unit is ramped down. It can be observed that, as expected, the speed of the betatronfollows the same pattern than the patient profile, the smaller the source to skin distance, the faster the betatron is and the smaller the dose delivered.

6. CONCLUSIONS

60 80 100 120 140 160 180 200 220

1.8

2

2.2

60 80 100 120 140 160 180 200 220

1.8

2

2.2

60 80 100 120 140 160 180 200 220

1.8

2

2.2

D

o s e

i n G y

6cm 6 cm

20cm 20 cm

40cm 40 cm


24/26

This paper has presented on-going work in the area of systems modelling and control engineering applied to skin cancer treatmentusing total skin electron. The design of a new skin cancer treatment unit based around a betatron and a fixed gantry constructionwas described. A model was developed to simulate the effect of electrons in patients in terms of source to skin distance and angleof incidence of the electron beam. Based on such a model, simulation studies carried out with Matlab/Simulink have highlightedthe difficulty in determining a set point for the betatron control unit. It was shown that the set point (in terms of speed of the

betatron unit, is a trade off between the accuracy and the time required to deliver the treatment. The simulation studies have alsodemonstrated the suitability of PID strategy to control the total skin electron unit. The speed calculation algorithm developed todetermine the trajectory of the betatron has illustrated the ability to deliver a uniform dose of radiation over a non flat surface.

Fig. 7. Comparison between normalised predicted - and simulated + speed of the betatron unit to deliver 200cGy using a 6 cm 2 treatment field.

ACKNOWLEDGEMENT

This work was supported by the TRW Foundation.

REFERENCES

Chrtien, M., C. Ct, R. Blais, L. Brouard, L. Roy-Lacroix, M. Larochelle, R. Roy and J.A. Pouliot (2000) Variable speed

translation couch technique for total body irradiation, Medical Physics , 27(5) , pp 1127-1130Guerin, F. (1999) Dose optimisation and controller design for betatron treatment machine , BEng Dissertation, CoventryUniversity, UK

Haas, O.C.L. (1999) Radiotherapy Treatment Planning: New System Approaches , Springer Verlag London, Advances in Industrial ControlHaas, O.C.L., K. J. Burnham, J. A. Mills, R. Crichton and P. Sharpe (2001) Trend and health monitoring in safety critical machines:

examples of linear accelerators, Proc 14th Int. Conf. On Systems Science, 3, 11-14 Sept Wroclaw (Poland), pp 251-258Hall P. (1998) Cancer team in plea to students: Challenge to adapt Russian built radiotherapy machine for use at Walsgrave,

Coventry Evening Telegraph, Thursday October 1Klevenhagen S.C. (1985) Physics of electron beam, Bristol: Hilger with the Hospital Physicists' AssociationMATLAB, The MathWorks (accessed 18 9 2001) available from http://www.mathworks.com/ Shank, B., (1998). Total body irradiation, Textbook of Radiation Oncology. (Leibel S.A. and Phillips T.L., (Ed)), WB Saunders

Company, Philadelphia, pp 253-276SmartDrive Ltd, (accessed 11 2 2002) available from http://www.smartdrive.co.uk

Simulation of the speed of the betatron


25/26

Project 10: Target tracking with Simulink - Vehicle velocity identification using on boardcameras

The aim of this project is to track using Simulink two targets (regions of interest with different shapeand grey levels) filmed using web cam or other video cameras (live or recorded images).

Methods to be investigated:

Mathematical morphology for shape recognition on binary images

Hough transform for line and circle detection to recognise an object with simple geometric shapes.

Material

Matlab image processing toolbox and Simulink image acquisition toolbox, web cam then ieee1394camera.

Deliverable

Review of image processing techniques to derive the speed of vehicle from fixed landmarks on theroad.

Report describing the capabilities Simulink for real time image processing

Report comparing Hough transform based algorithm on mathematical morphology for real time regionsof interest tracking in video sequences


26/26

Project 11 Agent based Controls in Service Oriented Architectures (SOA): Traffic FlowSimulation Studies in Intelligent Transportation Systems

The project aim is to investigate and apply different traffic scenarios in order to analyse a simulationsystem ( WD- iTS) being developed as part of PhD research project. WD- iTS is a framework toprovide efficient communication and integration between highly distributed ITS components. Thesystem is based on the concepts of Semantic Agent-based Controls and uses SOA as an underlying

architecture for overall services based integration. It is developed in .NET based technologies.

The student will be required to investigate various traffic flow scenarios that can be applied in thesimulated environment of WD- iTS. The results of the simulation will then be analysed and comparedagainst other similar approaches.

The project does not require any programming but basic understanding of XML and programminglanguages will be of great advantage. You will be provided with relevant resources and expert help.

Project Benefits:

Learning and industry driven experience Working in team or individual Working with industry specialist and support The project could lead to conference publications Your opportunity to improve your CV.

Documents

Msc Projects List From O Haas 2011