Advanced Workshop - Medikal Fizik Derneğimedikalfizik.org/uploads/kurs/sunum/Monaco-TPS... · Monaco 1.0 IMRT First TPS with Monte Carlo Monaco 3.0–3.3 SSO for DCAT FFF Support

Monaco TPS

Advanced Workshop

Istanbul, November 2019Dr. Dirk Wolff

2 | Focus where it matters

2000 – 2004: Studies of Clinical Engineering

2003 - 2009: Universital Hospital Mannheim

• Diploma Thesis (Dipl.-Ing)

• Master of science in medical physics (M.Sc.)

• Dissertation (Dr. sc. hum.)

2009 -2014: University Hospital Gießen

Since 2014: Application Specialist at Elekta (Monaco, ABAS, Oncentra, AQUA)

Dr. Dirk Wolff – A little about me


AgendaMonaco Advanced Workshop

3. Hands-on learning from template-based planning and advanced use of Multicriterial Optimization

1. Introduction to Monaco treatment planning strategies

2. Monaco concepts & tips and tricks

Monaco TPS

Intro and Concepts for Treatment Planning

Dr. Dirk WolffApplication Specialist, Team DACH


2007 2010 2011 2013 2015 2016

Monaco 1.0

IMRT

First TPS with

Monte Carlo

Monaco 3.0–3.3

SSO for DCAT

FFF Support

Monaco 2.03–2.04

VMAT

Supports Elekta

VMAT Delivery

Monaco 5.0

Basic 3D

SRS Cones

New GUI

4D Tools

SSO for dMLC

Siemens mARC

Monaco 5.11

Speed Improvements

Up to 4x faster

Present

day

5 | Focus where it matters.

Monaco 5.10

Advanced 3D

Monte Carlo

Recalculation

MRI

4D Specialty

Images

Frozen Dose

Template

Sharing

Monaco evolution

2019

Monaco 5.4/5.5

PlanScorecards,

DeformableI image

Registration, Registration

Object

Locate 2 with Angio

2017

Monaco 5.3:

Carbon Ion

Planning

Automation Toolkit,

Automated

Planning, Adaptive

Workflow


Monaco 5.11

Accuracy Speed Flexibility



dMLCS&SIMRT VMAT dMLC

S&SIMRT VMAT dMLC

S&SIMRT VMAT dMLC

S&SIMRT VMAT

Head & Neck Prostate 1 Prostate 2 Lung / Esophagus

0:18:53 0:14:53 0:27:38 0:17:47 0:10:19 0:28:21 0:12:40 0:14:25 0:19:53 0:46:10 0:31:39 0:26:22

0:05:55 0:04:33 0:08:24 0:04:47 0:03:35 0:10:19 0:04:36 0:04:51 0:05:07 0:10:48 0:07:30 0:09:03

0:00:00

0:07:12

0:14:24

0:21:36

0:28:48

0:36:00

0:43:12

0:50:24

Monaco 5.10.02

on Z840 (hh:mm:ss)

Monaco 5.11

on Z840 28c (hh:mm:ss)


Monaco 5.11 vs. 5.10.02 speed comparisons on Z840 28 core workstation

Z840 28C workstation

LP

TM

ON

171212


Monaco Software Features

Monte Carlo Algorithm

Smart Sequencing

Virtual Leaf Width


Dynamic SRS/SBRT: the Elekta solution

LP

TM

ON

171212


Dynamically position jaws inside the MLC

leaves during delivery to effectively reduce

leaf width. (Virtual Leaf Width)Smart sequencer provides high modulation when

needed and fast delivery when less modulation is

required, Versa HD is built to deliver these plans

in high dose rate mode (FFF).

9 | Focus where it matters.9 | Focus where it matters.

Monaco’s unique features for dynamic delivery


Versa HD can move the dynamic jaws

from segment to segment

The MLCs and the jaws can be

placed in 1 mm increments

For small fields this mimics the

effects of smaller MLC leaves


Take a closer look: Jaw tracking

Virtual leaf width

LP

TM

ON

171212


Monaco Monte Carlo algorithm

Arc calculation from

static gantry positions

Continuous Monte Carlo

arc calculation

Ideal for…

Small fields, high doses and anatomy with

varying densities (SRS/SBRT)

Source of truth

Continuous arc calculation

vs.

Discrete gantry positions


Multicriterial Optimization (MCO)

• Automatically achieve better normal tissue sparing without sacrificing target coverage

• Select OAR’s for MCO

• Choose the best plan possible!

Achieve the best plan with MCO


Automation with templates


12

3

4

5

SBRT Lung

5 Mouse clicks

2m 42s

Monaco Concepts

Tips & Tricks


Monaco ConceptsConstrained Optimization

OAR

OAR

PTV

PTV

Compromise


Monaco Concepts

Constrained Optimization

PTV

PTV

OAR

OAR


Monaco Concepts

• Constrained Optimization is logical way to plan.

• Monaco shows where conflicts are and highlight which cost functions are affecting the dose to targets.

• There is no guess work in what to change to achieve the target objective.

• This is a more structured approach to planning and leads to less iterations.



Monaco Concepts

1st Order Constraints• Goal will always be met.

• Serial, Parallel, Quadratic Overdose, Max Dose

2nd Order Constraints• Goal will be met UNLESS there is a 1st Order constraint.

• Quadratic Under Dose, Under Dose DVH

1st Order Objective• Goal will be met unless a 1st or 2nd Order Constraints prevents this.

• Target EUD, Target Penalty

2nd Order Objective• Goal will be met or succeeded unless Constraints prevent and UNLESS 1st order

objectives are not met.• Cost functions that have “Multi Criterial” option



DS: Voxel based toolsMonaco tells you which voxels are impacted

Monaco Unique Feature, not only DVH and dose based assessment, but also Spatial information

CF Occupancy Variation Relax Response CF Sensitivity

Voxels used by the

constraint

Voxels most impacted by

the constraint

Voxels impacted when

relaxing the constraint

Target’s Voxels impacted

by the constraint


Biological OptimizationThe Power Law Exponent - The K Value


Biological OptimizationThe Serial K values range from 1 to 20

A K value of 1 will apply

itself evenly across the

whole curve giving you a

mean dose effect


Biological OptimizationSerial K A K value of 10 will apply itself more

towards the hotter end of the curve

giving more of a max dose effect


Biological Optimization – Serial KSerial K

A K value of 20 will apply itself more heavily towards the hotter

end of the curve nearing similar behaviour to a max dose CF


Monaco Concepts

The ‘Variation’ tool shows where the cost function is applied and the effect upon the DVH.

Cost Functions – Serial


Biological OptimizationThe Parallel K values range from 1 to 4

In this Example we have asked for 30Gy to 50%

with a K value of 1. The low K value applies

broader penalty across the entire structure

50%

30Gy


Biological OptimizationParallel K

A med K, will apply more toward the

value you have selected but still have

some control over doses above and

below the selected value

50%

30Gy



A high K, will apply more directly

on the value you have selected

and wont have as much control

over doses above and below the

selected value

50%

30Gy

It then becomes more like

an Overdose DVH CF



As the low K value penalizes the dose above and below

the set value it can be too harsh and take dose from the

Target

50%

30Gy


Biological Optimization – Parallel K 4

A high K parallel cost function would look something like this, for a

OAR Ref Dose of 35Gy, mean Organ damage of 40%, and K = 4

Note that the intersection is what we asked for when K=4


Biological Optimization – Parallel K 1A lower K Value for the same parameters might look like this. Notice how the

whole curve has been pulled down and the target dose has been affected.


Monaco ConceptsCost Functions – Parallel

The ‘Variation’ tool shows where the cost function is applied and the effect upon the DVH.


What is EUD?

EUD = Equivalent Uniform Dose

• For Targets: EUD represents a homogeneous dose that when

applied to a target, has the same clinical effect as any given

inhomogeneous dose distribution within that target.

• For OARs: EUD represents a uniform dose in an OAR which

leads to the same probability of injury as a corresponding

inhomogeneous dose distribution in an OAR.


Target EUD Cost FunctionTarget EUD

0

5

10

15

20

25

65 67 69 71 73 75 77 79

Dose (cGy)

Incre

asin

g P

en

alt

y f

or

Co

ld S

po

ts

Cell Sen.

=0.1

Cell Sen.

=0.25

Cell Sen.

=0.5


Monaco Concepts

• Target Penalty is a physical cost function.

• It is an objective version of the Quadratic Underdose.

• It is a quadratic penalty constraint which starts at the threshold dose.

Cost Functions – Target Penalty

• The iso effect is a DVH-based physical parameter.

• Results in a steeper Target DVH than an EUD based cost function.


Target Penalty Cost Function


Monaco Concepts

• The Quadratic Overdose allows a Max dose and an RMS excess to be set.

• The RMS is really just a dose tolerance.

• It is much more flexible than a hard max and gives the system room to breath.

Cost Functions – Quadratic Overdose


Quadratic OverdoseControl of the DVH tail of the structure (Target or OAR)

Reduce the

RMS will reduce

the tail


Maximum Dose Cost Function

Hard Constraint- Use Wisely!Maximum Dose = 70 Gy

-50

0

50

100

150

200

250

300

350

400

50 55 60 65 70 75 80

Maximum Dose

Pe

na

lty


Monaco Concepts

• To use the maximum dose to control the global max dose, apply it to the external contour.

• Use the ‘Optimize over all voxels in volume’ option to apply it to all voxels in the study set.

• This applies a global ceiling to the optimization.

Cost Functions – Controlling Hot Spots


Monaco Concepts

• The Shrink Margin removes voxels in a structure away from adjacent targets. These voxels will not be used by the cost function for optimization.

• Extremely useful when multiple targets are being optimized to transition between a high dose and a low dose.

• Can be used in place of Optimization rings.

• Allows a transition zone between a high dose target and an overlapping or adjacent OAR.

Cost Functions – The Shrink Margin


Monaco Concepts Cost Functions – The Shrink Margin – Multiple Targets


Monaco Concepts Cost Functions – The Shrink Margin – Multiple Targets


Transitioning Dose between Targets

Head & Neck Planning Recipe


3 Targets case

• PTV66- TargetPenalty(66Gy)

- QOD68Gy (RMS=0.5,Margin=0)


- QOD66 (RMS=0.5,Margin=0)

- QOD62Gy (RMS=0.5,Margin=0.3)



- QOD55.6 (RMS=0.5,Margin=0.3)

Example of 3 levels H&N

• BODY- QOD54(RMS=1.5,Margin=0)- QOD40(RMS=1.5,Margin=1)- QOD35(RMS=1.5,Margin=2) or

maybe consider using Conformality- MAXDOSE(72.6, Opt Over All Vox)

BODY

PTV66

PTV60

PTV54


3 Targets case


- QOD68Gy (RMS=0.5,Margin=0)




- QOD55.6 (RMS=1.0,Margin=0)

Nested Targets

• BODY- QOD54(RMS=0,Margin=0)- QOD45(RMS=1.5,Margin=1)- QOD35(RMS=1.5,Margin=2) or

maybe consider using Conformality- MAXDOSE(72.6, Opt Over All Vox)

BODY

PTV66

PTV60

PTV54

Other Tips & Tricks for Monaco

High Fluence Smoothing

Smooths the fluence in stage one

Reducing the complexity of segments

It puts segments together more thereby reducing the number of MU’s and increasing the speed of delivery

Fluence smoothing enables an easier stage two calculation and more consistent result

600MU’s vs 900MU’s

Modulation degree …


Beamlet Width with High Smoothing

1mm Beamlet Width

Monitor Units: 348

Segments: 128

Calc Time: 3.31

2mm Beamlet Width

MU’s: 360.8

Segments: 121

Calc Time: 2.35

3mm Beamlet Width

MU’s: 387.7

Segments: 129

Calc Time: 2.30


Beamlet Width

1mm Beamlet Width

Monitor Units: 348

Segments: 128

Calc Time: 3.31

3mm Beamlet Width

MU’s: 387.7

Segments: 129

Calc Time: 2.30

2mm Beamlet Width

MU’s: 360.8

Segments: 121

Calc Time: 2.35


Beamlet Width

1mm Beamlet Width

Monitor Units: 348

Segments: 128

Calc Time: 3.31

2mm Beamlet Width

MU’s: 360.8

Segments: 121

Calc Time: 2.35

3mm Beamlet Width

MU’s: 387.7

Segments: 129

Calc Time: 2.30


Monaco Concepts

• The cost function can optimize over a 4cm radius or an 8cm.- This is adjusted by selecting the ‘Optimize over all voxels’ option.

• Values can be set from 0.1 – 1.0, with a lower value giving a more conformal distribution. Start at about 0.8 - 0.9.

• The value can then be lowered until the desired conformity is achieved- Great to apply when all other constraints have been met

- Great to use with MCO

Using Conformality Cost Function


Monaco Concepts

• Conformality works well for single target volumes and stereo volumes.

• It can be less effective with concave targets and complex cases.- This is due to multiple dose volumes as well as additional structures and

large changes in the patient geometry (remember if conformality is applied to the patient, higher structures will own the voxels).

Cost Functions – Conformality


Monaco Concepts

• You can control the dose to the patient by using the quadratic overdose cost functions and a series of stepped shrink margins.

• The shrink margins are used instead of optimization contours.

• Start with a value close to or just lower than the PTV target dose with a small RMS value.

Cost Functions – Controlling conformity with Quadratic Overdose


Monaco Concepts

• The first quadratic overdose is set to the same value as the outer target dose with no shrink margin.

• The cost function is applying directly against the PTV keeping the target dose inside the target.



Monaco Concepts

• The second quadratic overdose is set to a smaller dose value and has a shrink margin of 0.9cm applied.

• This is not applying in the voxels 0.9cm to the target. It only applies its penalty in the colored voxels.



Monaco Concepts

• The third quadratic overdose is set to an even smaller dose value and has a shrink margin of 2.4cm applied.

• This is not applying in the voxels 2.4cm to the target. Again, it only applies its penalty in the colored voxels.



Monaco Planning

• Sequencing parameters and IMRT Parameters will affect the quality AND deliverability of your plan just as much as the constraints will.

• The following slides will help to explain and review the parameters as well as give some tips.

• Should not be one size fits all

Sequencing & IMRT Parameters


Monaco Planning

• The Surface Margin →Target cost functions.

• Clipping contours at the patient surface.

• Ignores low doses in the build up region.

• Do not try to force dose into the build up region.

Cost Functions – The Surface Margin


Monaco Planning Cost Functions – The Surface Margin


Monaco Planning

• Targets are drawn out to the patient surface

• Decrease in MU

• Decrease in the global max dose.

• The Physician should try to avoid drawing targets to the skin surface.

Cost Functions – The Surface Margin


Monaco PlanningCost Functions – The Surface Margin


Surface Margin

• Clipping vs Surface Margin of 5mm – CF Occupancy View

Understand its application and use it wisely or avoid it!


Impact to Fluence Map






Delivery ComparisonImpact to MU and # of Segments


Monaco Planning

• More or less of the surrounding voxels to the target in optimization.

• This is comparable optimization margin in XiO

Sequencing Parameters – Target Margin


Monaco Planning

• For ‘normal’ planning, try not to restrict the margin too tight, this can affect the sequencing result if Stage 1 Fluence is poor.

Sequencing Parameters – Target Margin

68

Normal Tight Very Tight


Reduce Modulation Potential

• Consider Increasing Smoothing for simple plans

• Increase Beamlet Width (step increment in XiO) –large reduction in calc time – SSO does a great Job in delivering quality

• This also reduces segment # and presence of small undesirable segments (Modulation Degree i)

Keep it smooth for efficient results

Low Smoothing & 0.3mm BW High Smoothing & 0.5mm BW


Monaco Planning

• Filter “shapes changed” in the Optimization Console.

• Do not make Stage 2 changes until 2-3 loops have occurred.

• This will give it a chance to converge prior to altering its optimization pathway.

• Skipping Forward will bypass SSO loops continuing, which is ok if the plan quality is met.

Sequencing and Parameters – Segment Shape Optimization


When you add one beam with two rotations, Monaco

enhances the segmentation process. It essentially

splits the fluence through the BEV central X axis.

On one rotation, Monaco will optimize one half of the

volume. With the second rotation, Monaco will

optimize the other half.

Let’s look at an example.

Sequencing parameters—number of rotations

Monaco planning large volumes

72 | Focus where it matters72 | Focus where it matters.

Monaco optimizes

and segments this

half during the first

rotation.

Example: prostate with nodes and a unilateral nodal volume




Monaco optimizes

and segments this

half during the

second rotation.





74 | Focus where it matters74 | Focus where it matters.

If we look at the

segment for the first

rotation we can see

the segments

optimizing the right

side of the volume.

The same gantry angle

on the return rotation is

segmenting the other

side of the volume.




2 beams each

with 1 rotation

1 beam

2 rotations

Note how the dose between the unilateral volumes is better.




The benefits of Monaco multi-arc per beam

planning for pelvic disease sites

Background

The Monaco multiple arc-per-beam option allows radiation dose delivery to be optimized across multiple arcs, using a single collimator angle, without stopping radiation delivery.

Aim

To compare the use of 2 arc-per-beam (2APB) optimized plans with 1 arc-per-beam (1APB) plans for pelvic cancer patients.

Method

• Retrospective analysis of 17 previously treated pelvic cancer patients: o 9 prostate, 1 bladder, 3 uterus, 3 rectum, and 1

cervix (with and without involved lymph nodes)

• 2 plans generated for each patient: o one containing 2 beams using an arc-per-beam

setting of “1” (1APB)o another with a single beam using an arc-per-beam

setting of “2” (2APB)

• Elekta Infinity linac with Agility

• Plans were evaluated for PTV conformity, homogeneity, total MU, number of control points (CP), planning time and beam delivery time.

Kalet, A.M., Richardson, H.L., Nikolaisen, D.A. et al.

(2017) Dosimetric comparison of single-beam multi-

arc and 2-beam multi-arc VMAT optimization in the

Monaco treatment planning system. Medical

Dosimetry, 42(2):122-125.

Kalet, A.M. et al. (2017) Medical Dosimetry 42(2): 122-125

Figure shows comparison of dose distributions in 2 patients with complex PTV shapes

using 2APB (bottom) and 1APB (top) VMAT optimization

Monaco 5.11 Template Based

MultiCriterial Optimization

Automated Planning


What Are Templates?

What does this mean?• Templates store beam geometries, calculation parameters, calculation

settings, physician’s intent, IMRT constraints, …..

• Few clicks → ready for calculation.

• Monaco Biological cost functions → robust approach when used with anatomical volumes

What are the benefits of template-based planning?• Provides efficient ways to standardize the planning approach.

• With consistent templates, planning VMAT / IMRT is much easier.

• Decreases time to build plans.

Monaco is a template-based planning system


What is this all based on?

• Target dose → meet protocols

• OAR doses based on Quantec, ENZRAD or department

• MCO will ensure in most cases the plan result is customized over the patient-specific anatomy

OAR’s, AAPM, Quantec, TROG ENZARADENZARAD Protocol (78 Gy /39#)

PTV 78Gy ICRU 50/62/83

Rectum V50Gy < 50%

V60Gy < 35%

V65Gy < 25%

V70Gy < 20%

V75Gy < 15%

Bladder V65Gy < 50%

V70Gy < 35%

V75Gy < 25%

V80Gy < 15%

Femoral Heads V50Gy < 5%

Penile Bulb Mean dose < 52.5 Gy

https://www.eviq.org.au/radiation-oncology/urogenital/prostate


The Template

• Template contains the Target/s, Rectum and external contour- Prostate/GTV is optional

• OARs are controlled and achieved by the Body cost functions


The RectumThis can be visualized on the DVH

Serial CF

controlling

overlap region

Serial CF

controlling

high doses

Parallel

controlling low

doses


The Process

• Ensure Multicriterial is selected for all cost functions and the system is in Constrained mode

• At the end of stage 1 notice how the Isoeffect (What the system has achieved) is below the Isoconstraint (what is asked for)

Step 1: MCO over the fluence map


The Process

• When stage one is complete, uncheck ‘Multicriterial’ and switch to Pareto mode

• Enable Quadratic Underdose Cost Functions on Target/s

• Select Batch Optimise

• As everything has already been achieved it will re converge quickly and commence Stage 2

• Once stage two is complete review the plan

Step 2: PARETO optimization with SSO in Stage 2


In Summary

• Stage 1 optimisation:- Constrained Optimisation Mode

- Utilising a robust template

- Utilising Multicriterial Optimisation (MCO)

- Ensuring High Smoothing to control overmodulation

• Stage 2 optimisation- Pareto Optimisation Mode

- Disable Multicriterial

- Apply Quadratic Underdose Cost Functions to Targets

Template Based MultiCriterial Optimization Automated Planning

Thank you

Documents

Advanced Workshop - Medikal Fizik Derneğimedikalfizik.org/uploads/kurs/sunum/Monaco-TPS... · Monaco 1.0 IMRT First TPS with Monte Carlo Monaco 3.0–3.3 SSO for DCAT FFF Support