1

Adapting radiotherapy usingpatient-specific biological imaging

Mike PartridgeAssociate Professor

Translational Techniques: Up and coming techniques in Medical Physics translated into preclinical and clinical practice.IoP, 80 Portland Place, London Monday 1st December 2014

Lots of technology

• Linacs with on-board imaging• Cone beam CT• Tomotherapy, Cyberknife, Vero• Fluoroscopy• 4D CT• Surrogates for breathing / implanted fiducial

markers• Gating, tracking, ABC• Functional imaging PET, SPECT, MRI• Ultrasound

Jaffray, D. A. Nat. Rev. Clin. Oncol. 9, 688–699 (2012);

Personalized Radiotherapy Radiotherapy TodayThe fundamental aims of radiotherapy are independentof the technology we use:

– to deliver a high enough dose to control local tumour growth,– to keep normal tissue toxicity within reasonable limits.

Image Plan Treat

Improved TargetingPre-treatment we can use imaging to:

– Better identify the target volume (18FDG PET, MRI, 4D CT)– Better identify radiosensitive normal tissues

Image Plan Treat

Improved DeliveryDuring treatment we can use imaging to:

– Improve day-to-day setup and delivery accuracy– Measure (and correct for) systematic changes in anatomy

Image Plan Treat

2

Int J Radiation Oncology Biol Phys 60 1419-1424 (2004)

Fig 2, 3

CT PET Fusion

Target Definition: Head & Neck

I. J. Radiation Oncology Biol. Phys. 59 78–86 (2004)

FDG PET forvolume delineationin non-small-celllung cancer

T1N0M0 stage from CT

T1N2M0 stage from PET

PTVPlanning Target Volume

CTV

Clinical Target Volume

Volume DefinitionIn radiotherapy planning we define a number ofvolumes to account for uncertainties in the planningprocess:

GTV

Gross Tumour Volume

Where next?Use on-line imaging to reduce margins more andmore to allow dose escalation and normal tissuesparing.

– Yes, but with caution Int. J. Radiat. Oncol. Biol. Phys. 74, 388–391 (2009).

Continue to increase PTV dose with protons &carbon ions.

– Yes, but physical dose is not always the limiting factor.

Image every patient with every modality as manytimes as we can.

– Probably not …

CT PET SPECT MRI Linac ?

Oedema

Microscopicextension

Visible tumour mass

Hypoxic region

But tumours are heterogeneous

Necrotic core

Where next?Move away from delivery of a uniform population-derived dose to a (deliberately non-uniform) patient-specific prescription:

– Individualised iso-toxic dose– Biologically-guided using functional imaging– Stratified using gene expression profile

Monitor patient response during treatment to createbiologically adaptive plans:

– Boost dose to non-responding sub-regions of tumours– Reduce dose to well-responding areas– Early prediction (& reduction) of normal tissue toxicity– Intervene with radiosensitivity modulating drug

3

Imaging in Radiation Oncology

Anatomical

Before During After

Targetdefinition

Diagnosis& Staging

Biologicallyadaptive

IGRT

Early Late

Functional

Adapted from a slide by Robert Jeraj

The challenge

How could we use functional data for dose prescription?i. don’t – dose escalate to a fixed (or iso-NTCP) level

with an anatomically-defined targetii. dose escalate to a fixed level with a single functional

image defined targetiii. dose escalate to a fixed level with a multiple

functional image defined targetsiv. optimise plan using explicit radiobiological objective

function

? [Gy]

18F-FDG for boost target definition inlocally-advanced pancreatic cancer

Post80%

Post90%

Post60%

Post70%

Pre40%

GTV

Courtesy James Wilson and Maria Hawkins

MR-guided intraprostatic boost

H&E stained section Fresh slice + outline T2 MR + outline

Central gland (red)Tumour (yellow)Prostate (blue)

Diffusion coefficient map Spectroscopy voxels

Integrated area undergadolinium uptake curve

Dynamic coefficient Ktrans Dynamic coefficient Kep

Courtesy Sophie Riches, ICR

MR-guided intraprostatic boostChallenges:• Optimisation & validation of imaging (2 different

clinical trials)• Elastic matching of images pre/post hormone and

with/without endorectal probe (developed in-housesoftware)

• Developing optimum model to segment tumours(developed in-house software)

• Importing segmented template and matching withplanning CT (fiducial markers and in-house software)

MR-guided intraprostaticboost

T2W ADC T2map

Cho+Cr/Cit Ktrans IAUGC

Br J Radiol 2014;87:20130813

4

MR-guided intraprostaticboost

T2W ADC T2map

IAUGC

Br J Radiol 2014;87:20130813

FDG-PET guided boost

PTV

FDG-PET guided boostIn radiotherapy planning we define a number ofvolumes to account for uncertainties in the planningprocess:

GTV

“BTV”

PTV

GTV

“BTV”

Same mean dose to PTV

Imaging Protocol & QA

• To ensure that images are consistent, carefulprotocols should be followed for:− Patient immobilisation− Time from injection to imaging (for PET)− Image acquisition protocol− Image processing− Segmentation

• These should be regularly checked usingphantom measurements.

• Nuklearmedizin 2012; 51: 140–153

Acquisition Protocol

H. Zhuang et al J. Nucl Med 42 1412-1417 (2001)

• PET image contrastchanges as afunction of time afterinjection.

• Different types oflesion show differentkinetic behaviour.

• A well-definedprotocol is essentialfor accurateoutlining.

CT DCE MR DW MR FDG PET T2 MR

So which one is the target?

Courtesy Ceri Powell & Kate Newbold

5

Daisne Radiology (2004) 223, 93

Histopathological correlation Histopathological correlation

Eur J Nucl Med Mol Imaging (2013) 40:1828–1835

Study of 12 patients with head and neck cancer with 28 metastatic lymphnodes eligible for therapeutic neck dissection who underwent preoperativeFDG PET/CT.

Biological AdaptationUse the imaging to measure biological response totreatment and adapt individual patient treatmentsaccording to response:

Image Plan Treat Image Plan Treat

CT DCE MR DW MR FDG PET T2 MR

Early assessment of responsePre-treatment

Post-treatment

DW-MRI ADC changes at 2 weekspredicts response

Lambrecht et al.Radiotherapy and Oncology (2013)

Median ΔADC in poor responders was significantly lower than ingood responders 2 weeks into treatment (7% vs. 21%; p < 0.001).

Hoeben et al JNM 54 (2012) 532–540

baseline

baseline week 2

baseline week 2 week 4

18FLT as a predictor of response

Stratify by 45% decrease in SUVmax between baseline and week 2(patients treated with radiotherapy)

6

Zips et al Radiotherapy and Oncology 105 (2012) 21–28

18FMISO PET as a predictor of response The RHYTHM trial as an example ofimaging biomarker development

• RHYTHM: modulation of Radiotherapy according toHYpoxia: exploiting changes in the TumourMicroenvironment to improve outcome in rectalcancer.

• Objectives: To define the best method of detectinghypoxia in rectal cancer and to develop a method ofdetecting changes in rectal tumour oxygenationduring preoperative chemoradiotherapy

BiopsyBlood sampleFMISO-PET

pCTdce-MRI rectum

Day ‒7: ‒1 0 10-12 6 weeks 8 weeks

Restagingdce-MRIrectum

BiopsyBlood sampleFMISO PET

pCTdce-MRI rectum

Radiotherapy planning feasibility study

RHYTHM-I: Group B

Daily radiotherapyand capecitabinetreatment (Mon-Fri)

Resection

Histologically confirmed locally advanced adenocarcinomaof the rectum: CRM threatened/involved

MRI FMISO PET / CT FMISO PET / MRI

Images before and after 10# RT

Images courtesy of James Wilson

Compare images or doses?Picking a single fixed threshold to define a biologicaltarget volume is hazardous:• Different thresholds give different volumes!• Threshold value changes with object size• Patters of uptake vary with time

Søvik Semin Radiat Oncol 20 138–146 (2010)

Spontaneous canine head and neck tumour treated with3.0 Gy per fraction.

DCE MRI-derived pO2

Optimal dosedistribution given the

above pO2

Compare images or doses?Solution – use doses not images

Søvik Semin Radiat Oncol 20 138–146 (2010)

7

http://www.ucl.ac.uk/silva/nuclear-medicine/patients/PETMRschematic.png?hires

PET / MRI

Crijns & Raaymakers PMB 59 3241 (2014)

MRI / linac

Summary I

• Biological information already plays an importantrole in radiotherapy in staging

• It is playing an increasing role in assisting GTVdefinition

• A number of clinical trials are in progress testingfunctional image-guided “boosting” or doseredistribution.

Summary II

• Through observational trials we are learning howto use functional imaging as a quantitative andstandardised “biomarker” of response totreatment.

• This will allow patient-specific adaptation oftreatment according to response to therapy.

• The advent of the PET/MRI and the MR-LINACoffer the opportunity to get more and morepatient-specific biological information duringtreatment.

Challenges

• Knowing the best imaging modality to use andthe best time point to observe response (but stillhave time to intervene) for a particular diseaseand site needs more work.

• The link between functional image “signal” anddose-response is still largely unknown and raisesboth physical challenges (absolute imagequantification, standardisation and QA) andbiomedical challenges (clinical dose-responsestudies using heterogeneous prescriptions).

• There are safe and pragmatic ways to get startedwithout the above …

Acknowledgements

Tim MaughanMaria HawkinsJames Wilson

RHYTHM trial groupSamantha Warren