Biocomputing: is there useful to compute radiotherapy margin

1
Biocomputing is there useful to compute
radiotherapy margin
Bondiau, Pierre-Yves (1, 2) Marcy, Pierre-Yves
(1) Clatz , Olivier (2, 3) Malandain,
Gregoire (2) Delingette , Herve (2) Sermesant
, Maxime (2) Warfield, Simon (2, 3) .
1. Centre A. LACASSAGNE. 33 avenue de valombrose,
06189 Nice cedex 2, France 2. INRIA, Epidaure
Project. 2004 route des lucioles, Boîte postale
93, 06 902, Sophia Antipolis, France Surgical
Planning Laboratory, Brigham and Women's
Hospital, Boston, MA, USA.
2
Brain tumors diversity
3
Tumor simulation (I)
Tumor volume growth
Expansion
Infiltration
4
Tumor simulation (II)

• Tumor volume growth expansion and/or infiltration

Expansion speed
Glioma
Metastasis
Hemangioblastomas
Sarcoma
Lymphoma
Meningiomas
Infiltration speed
5
Glioblastoma
IRM T1 T1i T2
6
IRM T2
Oedema Tumour cells Tumour cells
infiltrating Necrosis
Intensity
Number cells
7
Tumor simulation model

• Infiltration
• Murray 89
• Expansion
• Correlation

Tumor rate
elasticity
external forces
External forces
Tumor rate
8
Atlas
1 Voxel 0,6 mm3 Extrapolation of brainweb
images
9
Atlas
10
Brain simulation

• White matter

• Grey matter

• Ventricles

• Falx

• Skull

Labeling Elasticity Infiltration
11
White fibers integration use of the DTI
12
Adding DTI in the atlas
13
Principe
September
March (simulation)
Matching of GTV into atlas
Matching of new GTV Into MRI
March (real)
Tumour growth simulation
14
Results

• 1 - Mechanic
• 2 - Diffusion
• 3 Both
• 4 microscopic invasion

15
Mechanic
September
March (simulation)
March (real)
16
Mechanic
September
March (simulation)
March (real)
17
Diffusion
September
March (simulation)
March (real)
18
Both
September
March (simulation)
March (real)
September
March (simulation)
March (real)
19
Both
20
MicroscopicInvasion
5-10 ?
21
Microscopic invasion (2)
22
Conclusion
23
Growth tumour simulation (1)

• The simulation of GBM is complex, associating
  mechanical and diffusion components.
• The model does not require specific imaging
  protocols, and routinely acquired images are
  sufficient for our purposes
• The model do an estimation of the microscopic
  invasion (better definition of margins ?)
• An additional advantage is the estimation of
  hidden parameters (e.g. aggressiveness) for
  classification purposes, education, etc.

24
Growth tumour simulation (2)

• The individualization of the model can be
  improved by bringing more individual information
  for instance, the patient DT image, or the
  tumoural growth rate. These two components are
  complementary, and can be tuned independently
  this makes the model generic and should enable to
  simulate other tumour growths (lung, prostate,
  breast, bone, etc.)
• Next work validation with microscopic
  pathological information on a large number of
  patients.

25
Cocosco CA, Kollokian V, Kwan RKS, Evans AC
BrainWeb Online Interface to a 3D MRI
Simulated Brain Database. NeuroImage 1997 5 4,
425