RadioBiology: The view from the trenches'

1
RadioBiology The view from the trenches.
Andrew W. Beavis PhD and Jaap Vaarkamp PhD

• Department of Radiation Physics,
• Princess Royal Hospital
• Hull and East Yorkshire Hospitals (NHS) Trust and
  University of Hull

2
I am not a RadioBiologist!
answers??
More questions than
!!
3
I am a nice guy though

• Not trying to rubbish anybodies work here!

• Times are a changin
• (self appointed) task is to play devils advocate
  and raise some issues

4
Princess Royal Hospital, Hull

• Mr. Viv. Whitton
• Head of Rad. Physics
• 2 Principal Physicists
• (1 Principal in Protection)
• 3 x Junior (B8-B13)
• 2.5 Dosimetrists
• Reasonable sized research group (4 students,
  research assistant, web of collaborations)
• Clinical routine IMRT service

• 3 Varian Linacs (600C, 600CD, 2100C)
• (6, 10 MV 6-20 MeV)
• Odelft Simulator
• Dedicated Picker PQ5000 CT scanner
• GE Signa 1.5T MRI
• Philips Intera 1.5T MRI
• GE Signa 3.0T MRI
• 3 CMS FOCUS w/s networked (PC) Focal products
• SIITP, Focal fusion
• 100 patients (total) per day

5
Dose distribution is a surrogate for treatment
outcome.

• Currently compute dose distributions in treatment
  planning system
• Seek to gain plans that satisfy some/ many
  requirements
• Consult DVHs etc etc
• ACTUALLY want to predict treatment outcome/
  efficacy/ estimate therapeutic ratio
• TCP/ NTCP
• In this presentation, highlight that much of the
  data we may try and base these on could be
  considered questionable in the contemporary
  context

6
e.g. NTCP Normal tissue compromise

• Data/parameters used for NTCP models.
• Mainly, collected retrospectively, Emami data
• Most of us will rely on the literature for input
  data TD50, n, m

Burman, C., Kutcher, G. J., Emami, B., and
Goitein, M. Fitting of normal tissue tolerance
data to an analytic function. Int. J. Radiat.
Oncol. Biol. Phys. 21, pp. 123-135 (1991).
7
The NTCP Workspace
Enter the organ/structure for which to model NTCP
Enter uniform dose which results in 50
complication
Enter the n parameter describing the volume
dependency of the organ
Enter the m parameter modeling the slope of the
NTCP curve
Enter the clinical end point
Enter the reference volume corresponding to the
TD50
For documentation purposes only not needed for
NTCP calculation.
Mouse can be used interactively in the TCP window
to automatically increase or reduce the total
dose delivered by all external beams.
Typical RTPS interface that encourages user to
put the numbers in!
8
XiO/ FOCUS Online reference to published NTCP
parameters
So, its EASY to use these tools!!!
9
Probability, statistics and .

• Tumour control PROBABILITY
• Normal Tissue Control PROBABILITY
• NTCP of 1
• Acceptable? Means 1 in 100 of those who walk
  through our door
• Beavis winning the National lottery?
• 1 in 14,000,000

10
Publication Emami et al. 1991Based on 80s
data/ experienceTreatment given Conventional /
large fieldsWant to apply it Today, 10 July
2003Treatments conformal/ IMRT
11
Summary of how to use

• Compute dose distribution
• Characterise it into some parameters
• Run some calculation using these and the
  suggested biological parameters
• Produce some figures..
• Actual predictions?
• Figures of merit use to compare plans?

12
Conventional treatments calculations?- from 20
years ago

• Point calculation and assumed distribution
• Or a planned distribution

• Depth,
• d Sep/2

Sep

• Simple algorithms used
• Uniform Bulk density generally assumed

13
Outcome delivered dose correlation

• Clinical follow up/ observed outcomes and
  complications
• Related to a distribution that was probably
  reported as a point dose to some reference point
  (isocentre)
• No spatial information regarding the proximal
  organs at risk
• i.e. in prostate treatments the rectum may have
  been bathed in same dose.

14
Contemporary treatment planning

• Using more accurate planning algorithms
• Typically accounting for tissue density
  heterogeneities
• Aim to spare organs at risk
• Calculations/ delivery/ philosophy all different!

15
(No Transcript)
16
Even in contemporary data can we characterise the
distribution with a DVH??

• 2000 ?
• Assume this represents the delivery
• Organs move through out the treatment
• 1980s..
• Perfect delivery was assumed in the fitting of
  the Emami parameters
• But motion was convolved into the observations.

Courtesy Di Yan, William Beaumont and Marcel Van
Herk, NKI Amsterdam
17
Question.

• So, is it OK to apply the historical parameters
  into models/ fitted estimators to risk predictor/
  outcomes predictor in contemporary treatments
• How was distribution calculated?
• What field arrangement was used?
• What organ sparing, if any, was applied?

18
Care needed!

• We are taking data generated by fitting outcomes/
  opinion (?) to dose distributions reported in
  very simple terms
• Attempting to use these models to predict
  complications/ outcomes prospectively from
  distributions computed in quite different ways!

19
How can we address these concerns?- motion issues

• Even with careful contouring etc etc. need to be
  wary about how we SCORE delivered dose
  distributions
• Planning scan is a static SNAP shot
• Should really be adding subsequent fractional
  deliveries dynamically
• Addressing the temporal change in volumes/
  geometry

20
Deformable Registration
CT Fraction 9
CT Fraction 1

• Cannot just add distributions when underlying
  anatomy is fundamentally different!

Courtesy Rock Mackie
21
Deformable Registration
CT Fraction 9
CT Fraction 1

• Need to deform the assumed dose distribution
  computed on the planning image/ anatomy to that
  of the treated anatomy dose transformation
  matrix equals the image transformation matrix.

Courtesy Rock Mackie
22
How can we address these concerns?-
calculational issues

• Modern trials collect much more data
• Store imaging data etc etc.
• QA control of participants dose calcs etc.
• USA RTOG data store is vast
• Located at Wash. Univ. in St. Louis, USA
• How to analyse and use this data from
  multi-platform sources?

23
CERR A Computational Environment for
Radiotherapy Research

• Joe Deasy, PhD
• Washington University

Courtesy Joe Deasy
24
CERR A Computational Environment for
Radiotherapy Research

• Matlab-based
• Cross-platform
• RTOG format-based
• Self-describing format
• Freely available via webpage http//deasylab.info

Courtesy Joe Deasy
25
Courtesy Joe Deasy
26
Courtesy Joe Deasy
27
CERR components partially completed.

• Include
• Monte Carlo beamlet calculations (Iwan Kawrakow,
  Zakarian)
• Fast pencil beam computations based on Ahjneso
  formulation (Deasy, Kalinin)
• IMRT optimization module (Deasy, Eva Lee)
• Biological plan evaluations (TCP, NTCP)

Courtesy Joe Deasy
28
Uses for CERR

• Data-accessing for outcomes analysis
• Convenient plan archiving
• Data-mining for outcomes analysis
• Treatment planning prototyping tool
• IMRT treatment planning
• Monte Carlo denoising
• Sharing and comparing treatment planning results
• Building inter-institutional databases
• Any kind of RTP research

Courtesy Joe Deasy
29
Using this kind of tool

• The oncology community can start building AND
  analysing databases created from recent/ future
  trials
• Using these we can begin to produce data for TCP,
  NTCP models that are consistent with contemporary
  treatment philosophies

30
Finally, cold spots in IMRT plans?

• With the optimisers in current use in IMRT
  systems can get cold spots around the edge of the
  PTV
• A feature of dose difference cost functions

31
PTV is an envelope within which the CTV is
present with 100 certainty

• Cannot assume any edge of the PTV is less at risk
  from being under-irradiated
• (May reduce the risk.)

32
Without specific Biological target information

• Must assume that the tumour clonogen cell density
  is homogeneous
• Again most models assume this?
• In collaboration with Mark Wiessmeyer at CMS we
  are investigating a EUD based optimisation
  process that uses biological imaging data as
  input
• Poster at AAPM

33
Geometric Modulation
Courtesy Mark Wiessmeyer
Tumor region more inside may need higher dose
Smooth transitions
GTV
CTV
PTV
Phantom Dimensions(cm) Anatomy Radius Height
Origin Skin 20 40 0,0,0 Tumor
5 2 0,0,0 A hockey puck in a cylinder
Control transition in overlap region
Tumor
Normal
34
Dynamic Contrast-Enhancement

• Acquire baseline images pre-contrast
• Multiphase post-contrast
• Uptake of tumour can be analysed
• Parameter map produced

• Hold an 86.5k (YCR) grant to investigate
  application in head and neck treatments

35
Functional MRI for conformal avoidance

• Left handed 21 year old with a glioma in the left
  frontal lobe
• T2-weighted fast spin-echo (TE/TR100/3640 ms) 3D
  T1-weighted fast spoiled gradient-echo
  (TE/TR/flip 4.2/13.4ms/200)
• fMRI acquisition (self-paced finger-thumb
  opposition with the right hand).
  T2-weighted
  (TE/TR/flip50/300 ms/900)

Activation plot
36
The whole body 3T cometh!MR Spectroscopy 3DCSI
3D-CSI (3D Chemical shift imaging)
37
Thanks !

• Dr. Rock Mackie
• Dr. Joe Deasy
• Mr. Mark Wiessmeyer
• Dr. Gary Liney and our students
• Physics and Planning staff back home.
• CMS for their support, interest and financial
  help with our research programs

Lemmy Inspiration, loud noises, art of tact and
subtly