Walk‐Through of CERR Capabilities
CERR: Computational Environment for Radiological Research
• CERR (pronounced 'sir') provides a platform to prototype algorithms for Radiotherapy and Radiomics. (e.g. IMRT, Outcomes research, di i (
h
Image features etc)
• Open source, free: GNU GPL copyleft
• Developers: James Alaly, Aditya Apte, Divya Khullar, Yu Wu and Joseph O. Deasy
Aditya P. Apte, Ph.D.
Department of Medical Physics
Memorial Sloan Kettering Cancer Center
New York
aptea@mskcc.org
AAPM 2015, July 15, 2015
1
CERR: Introduction
2
Documentation |Support | Community
• Wiki:
• Written in Matlab. – http://cerr.info/wiki
– Source and Compiled versions are available
• Forum:
• Import and display treatment plans from a wide variety of commercial or academic
wide variety of commercial or academic treatment planning systems. – http://groups.google.com/group/cerr‐forum
p //g p g g
/g p/
• Download:
– http://cerr.info/download.php (GitHub)
• Provides a common filetype for Research
– E.g: creation of multi‐institutional treatment plan databases for various types of research studies, including dose‐volume‐outcomes analyses ,IMRT treatment planning comparisons and Radiomics.
• Citations – 261 Web of Science, 379 Google Scholar (June 2015)
3
• CERR Viewer
Outline
Getting CERR: Control Panel
– Dose Distributions
– Images
– Structures
– DVHs
Source version:
Add all the sub-folders
to Matlab path
• CERR data structure
• Use Cases
Compiled version:
Install Matlab runtime
libraries
– IMRTP Research
– Radiotherapy outcomes modeling
– Radiomics
• Advanced visualization and analysis tools
4
5
6
1
CERR Viewer
CERR Viewer
CERR Viewer: Help
7
CERR Viewer
8
CERR Viewer
3D visualization
Courtesy: Rachael Hachadorian
9
CERR Viewer
10
CERR Viewer
Dose Management
11
Colorbar / Isodose
12
2
CERR Viewer
CERR Viewer
Dose / CT Profile
Dose Projection
Right-click options
13
CERR Viewer
Image Registration via command line:Wrapper around Plastimatch
14
CERR Viewer
Image Registration / QA
Before Registration
After Registration
[basePlanC, movPlanC] =
register_scans(basePlanC, movPlanC,
baseScanNum, movScanNum, algorithm)
15
CERR Viewer
16
CERR Viewer
Image Registration / QA
Mirror Checkerboard
CheckerBoard
Mirror Scope
Image Mirror
17
18
3
CERR Viewer
CERR Viewer
Contouring
Contouring on Fused Images
19
CERR Viewer
20
CERR Viewer
Contouring – Region of Interest
Deriving new structures
21
CERR Viewer
22
CERR Viewer
Structure Transformation Examples
Dose volume Histograms
• divideStructureAntPost(structNum)
• divideStructureAntPostLeftRight(structNum)
23
24
4
CERR Viewer
CERR Viewer
Dose volume Histograms
Dose volume Histograms
Fractional
Absolute
25
26
CERR Viewer
CERR Data structure
Default Settings: CERROptions.m
• Sinc upsampling
• PET SUVs
• Plastimatch command file
command file
• Sagittal/Coronal display of Structures
• All DICOM objects are stored under a “global” planC variable.
• planC: CellArray containing a particular type of DICOM object. For example:
For example:
Transverse slice Sinc-upsampled slice
of a PET scan at at resolution of
original resolution 512x512
of 128x128
– 3rd element (planC{3}) contains scan metadata
– Last element of planC (planC{end})contains mapping of these DICOM objects.
27
28
CERR Data structure
Unit Testing
Scripting Examples
Scan coordinates
runCERRtests
Dose within Structure
29
30
5
Use Case
Use Case
Radiomics features supported
Radiomics Pipeline
Import of Image data
ROI
Aggregation
Shape1
•Eccentricity, Circularity, Solidity, Extent, Euler Number Texture1
•Homogeneity, Entropy, Contrast, Coherence
Zone Size3
•Homogeneity, Entropy, C t t C h
Contrast, Coherence
Stats4
•Kurtosis, Slope, Skewness, Min/Max/Median/
Analysis
MTV
31
Use Case
1. El Naqa et al, Pattern Recognition, 2009; 42(6) 1162-1171.
2. Haralick et al, IEEE Transactions on Systems, Man, and Cybernetics, 1973; SMC-3(6):610-621.
3. Yu-Hua Dean Fang et al., BioMed Research International, vol. 2014, Article ID 248505, 2014.
32
4. Philippe Lambin, Hugo J.W.L. Aerts et al, European Journal of Cancer, Volume 48, Issue 4, March 2012, Pages 441-446
Use Case
Prototyping an IMRTP Eclipse
algorithm
•Metabolic Tumor Volume is the volume with SUV above a set threshold
IMRTP GUI
CERR
Planner’s Tx Plan
3rd Party
Extract Beam parameters
Create beamlets
Optimize Beamlet
weights
Calculate Influence of each beamlet at each voxel
each voxel
Output the beamlet weights (fluence) to Eclipse readable files
Generate volumetric dose in
Eclipse reading the fluence
profiles
33
Use Case
34
Use Case
IMRTP Influence Matrix
• Get influence matrix with QIB or VMC++ dose calculation algorithms
35
Getting the volumetric Dose: (Influence) (beamlet weights)

36
6
Use Case
Optimize Beamlet weights and write to Eclipse‐readable fluence
Use Case
Eclipse vs PriOpt Plans
courtesy: Paras Tiwari
Implemented
by user
Provided by
CERR
Beamlet
Optimizer
p
PBweights_to_fluence
37
Use Case
38
Use Case
Toy Example IMRTP solver
CERR Extractor
Makes it easy to batch-Extract and combine
Dosimetric & Clinical data (ETL process)
runOptimExample.m
Treatmement
data
http://cerr.info/wiki/index.php/ORART_Optimiz
ation_example?w=CERRWiKi
i
l ?
i i
Outcomes and
Clinical data
39
Use Case
40
Use Case
CERR Extractor
STEP 1: Choose directory containing Tx
plans
STEP 2: Choose structures to model
STEP 3: Manually correct unconventional
names
STEP 4: COHORT REVIEW
STEP 5: Choose Excel file containing
outcomes / clinical data
STEP 6: Export Tx, outcomes and clinical
information to DREES
STEP 7: Model
41
Cohort Viewer
42
7
Advanced tools
Dose Location Histogram (DLH)
Advanced tools
Plan Robustness Analysis
DLH gives an idea
about the location of
cold / hot spots within
a structure relative to
its surface.
Eg: 85cc of cold
spots (Dose <=
70Gy) lie within 1.4
cm from the surface
43
Advanced tools
Generating Consensus Contour
Graph of Agreement volume v/s Confidence level for the three methods: Apparent, kappa‐
corrected and STAPLE. User can interactively select the confidence level by dragging the red line and consensus regions will be updated accordingly. This process would change the estimated volume to match desired confidence level.
44
Advanced tools
 Based on "A fast
algorithm for
gamma evaluation
in 3D”, Wendling et
al, Medical
physics, 34
(5), pp. 1647,
2007.
 Works with
doses of unequal
dimensions and
having associated
transformations.
Computation time
for regular doses is
~2 seconds.
Fast 3D Gamma Metric
1
0
Warfield, Zou, and Wells 2004
45
46
Thank you
47
8