The INSIDE project: in-beam PET scanner system features ... · scanner system features and characterization V. Ferrero, on behalf of the INSIDE Collaboration ... Antonio Zampieri

The INSIDE project: in-beam PET scanner system features and

characterization

V. Ferrero, on behalf of the INSIDE Collaboration

14th Topical Seminar on Innovative Particle and Radiation Detectors, 3-6 October 2016, Siena

N. Belcari M.G. Bisogni N. Camarlinghi A. Del Guerra S. Ferretti E. Kostara A. Kraan B. Liu N. Marino M. Morrocchi

M.A. Piliero G. Pirrone V. Rosso G. Sportelli

A. Attili P. Cerello S. Coli V. Ferrero E. Fiorina G. Giraudo F. Pennazio C. Peroni A. Rivetti R. Wheadon

E. De Lucia R. Faccini P.M. Frallicciardi M. Marafini C. Morone V. Patera L. Piersanti A. Sarti A. Sciubba C. Voena

F. Ciciriello F. Corsi F. Licciulli C. Marzocca G. Matarrese

G. Battistoni M. Cecchetti F. Cappucci S. Muraro P. Sala

INSIDE coordinator: M.G. Bisogni

CNAO - TERA Foundation, Pavia Synchrotron

60-250 MeV/u protons 120-400 MeV/u carbons

Protons since 2011 Carbons since 2012

Hadrontherapy

V. Ferrero (INFN and University of Torino) - [email protected] IPDR16 Siena, 3-6 October 2016

ATREP, Trento Cyclotron

220 MeV/u protons Since 2014

CATANA, Catania Cyclotron

62 MeV/u protons Since 2002

“Radiation therapy is the medical use of ionizing radiation to treat cancer. When the irradiating beams are made of charged particles (protons and other ions, such as carbon), radiation therapy is called Hadrontherapy.”

The European Network for LIGht ion Hadron Therapy

NuPECC report “Nuclear Physics in Medicine“, 2014

mailto:[email protected]?subject=

V. Ferrero (INFN and University of Torino) - [email protected]

Hadrontherapy

What prevents particle therapy from becoming mainstream? ‣ unproven clinical advantage of lower

integral dose ‣ costs ‣ range uncertainties

Zhu X, Fakhri GE. Theranostics. 2013;3(10):731-740.

Dose release uncertainty due to: ‣ approximation of dose calculation

methods ‣ differences between treatment

preparation and delivery ‣ patient misalignment ‣ anatomical/physiological variations

among different treatment sessions ‣ internal organ motion ‣ …

http

://w

ww

.sea

ttlec

ca.o

rg/d

isea

ses/

prot

on-th

erap

y-ho

w-it

-wor

ks.c

fm

IPDR16 Siena, 3-6 October 2016

AA

PM 2

012

poll



Particle range verification

Zhu X, Fakhri GE. Theranostics. 2013;3(10):731-740.

Main contributions: 11C (T1/2 ≈ 20.3 min) 10C (T1/2 ≈ 19.3 s) 15O (T1/2 ≈ 2.0 min) 13N (T1/2 ≈ 10.0 min)

Other contributions: 14O (T1/2 ≈ 70.59 s) 15O (T1/2 ≈ 0.008 s) 12N (T1/2 ≈ 0.01 s) ….

β+ isotopes

J Pawelke et al., Proceedings IBIBAM, 26.-29.09.2007, Heidelberg


off-room PET in-beam PET


Tracker + Calorimeter = DOSE PROFILER IN-BEAM PET HEADS

‣ Dual signal operation ‣ Integrated in the

treatment room ‣ Provides immediate

feedback on the particle range




The INSIDE in-beam PET: system assembly and characteristics



FE boards (64-ch TOFPET ASIC) on cooling support

FE boards mounted and cabled in PET boxDetector blocks on glass-fibre support

256 LFS pixel crystal (3x3x20 mm3) coupled one to one to MPPCs (Multi Pixel Photon Counters, SiPMs)



Completed PET detector (running)

January 2016, Torino: ‣ Full PET heads assembled and tested ‣ First acquisition with all boards (Lu background radiation, FDG vials)

Ready for test with Tx boards and chiller connected

Xilinx SP605 FPGA boards (decode TOFPET data stream, apply energy thresholds)



The INSIDE in-beam PET

in-beam PET heads

signal β+ decay

acquisition in-spill, inter-spill, after-treatment

position heads face to face, perpendicular to the beam

axisdistance from isocenter 25 cm

output 3D activity map

The in-beam PET gives an integral

measurement of the β+ activity distribution

INSIDE @ (February 2016, Pavia)



System configuration

Switch: 24-port Gigabit + 8-port Gigabit Control PC (desktop) DAQ Server: 32 cores HT, 128 Gb RAM Monitoring PC: 4 cores, 6 Gb RAM (desktop)

CONTROL ROOM

TREATMENT ROOM

INSIDE in-beam PET system

LabView

DAQ GUI

Switch

Master board

TX

TX(….)

(….)

(….)

Crateslow control calibration monitoring

server real time coincidence finding software

monitoring LORs

online recon



In-beam PET workflowTreatment plan

Treatment

Measurement

Data acquisition

Data analysis

Image reconstructionComparison

Result

Expected value calculation

FLUKA simulation

Data analysis

Image reconstruction

on-line!




Simulation and Data



In-beam PET simulationSimulations account for: ‣ detector geometry and performances ‣ beam geometry ‣ beam delivery informations (DDS = Dose Delivery System)

Isotope production: all the statistic must be simulated

STEP 1: beam simulation

about 1/100 of primary hadrons

are simulated

time-tagged activity scoring

‣ annihilation time and position

‣ isotope production map (10C, 11C, 15O, 13N, etc….)

STEP 2: PET simulation

all positrons are simulated (x100)

data analysis and image reconstruction

same as real data (LOR, MLEM algorithm with 5

iterations)

the temporal structure of the beam is simulated

isotopes decay are simulated

time

the reconstructed image depends on the acquisition time




First beam test (2 modules)

E. Fiorina, Nuclear Instruments & Methods in Physics Research A (2015)

The first couple of PET modules was tested in May 2015

Data set: 2 monoenergetic single spot (2E+11 protons) E1=68.3 MeV/u, E2=72.03 MeV/u

Distal fall off evaluation: ‣ simulation and real data difference < 1 mm ‣ gap between 68MeV and 72 MeV compatible

with difference in proton range



Treatment Plan (data)

Beam test @ 07/06/2016

Proton Beam 15x15x20 cm3 PMMA

PTA1/PTA2, E = 83-150 MeV/u PTB1/PTB2, E = 62-129 MeV/u

Activity reconstruction: 700 s treatment + aftertreatment

The image reconstruction is made during the treatment delivery (as soon as LORs are available)

PT A1 PT A2

PT B1 PT B2

PET1

PET2

PMMAbeam

xy

z

50

600

300

200

100

400

500

50

600

300

200

100

400

500


beam

beam

beam

beam



50

600

300

200

100

400

500

Data Simulation

Data Simulation 50

600

300

200

100

400

500

50

600

300

200

100

400

500

Data Simulation

50

600

300

200

100

400

500

Data Simulation

Treatment Plan (data and simulation)


PT A1

PT A2

PT B1

PT B2



Activity profile analysis

Data Simulation

Data Simulation

Data Simulation

PT B2

Data Simulation

Data/Simulation activity profiles along beam axis (z) Mean over 0.96x0.96 cm2 from center of image


PT A1

PT A2

PT B1beam

beam beam

beam



3D Image Comparisondata (white) simulation (red)

voxelPT1 - voxelPT2

Higher difference at the edges … But the activity of interest is not in the whole volume (reconstruction artifacts), the dose delivery plan needs to be taken under consideration

activity mask

(thr 10%)

50

600

300

200

100

400

500

Data Simulation


PT A1



50

600

300

200

100

400

500

Data Simulation

Treatment Plan (data and simulation)

x [mm]0 20 40 60 80 100 120 140 160 180 200 220

y [m

m]

0

20

40

60

80

100

20−

15−

10−

5−

0

5

10

15

20

End of activity comparison

k_end3x3Entries 1610Mean 0.879RMS 1.046

/ ndf 2χ 253.7 / 4Constant 15.4± 431 Mean 0.0314± 0.8752 Sigma 0.026± 1.024

z [mm]20− 15− 10− 5− 0 5 10 15 20

coun

ts

0

100

200

300

400

500

600

k_end3x3Entries 1610Mean 0.879RMS 1.046

/ ndf 2χ 253.7 / 4Constant 15.4± 431 Mean 0.0314± 0.8752 Sigma 0.026± 1.024

End of activity distribution

PT1, data/simulation comparison:

Mean (0.87±0.03) mm RMS (1.02±0.02) mm


PT A1



Conclusions

Particle range monitoring is a key issue to improve quality assurance in particle therapy The INSIDE in-beam PET is currently being tested at the CNAO, Pavia

‣ Real-time reconstruction ‣ Simulation/data comparison ‣ Qualitative accordance ‣ Quantitative accordance within 1 mm

Knowing the expected activity distribution, the experimental distribution (strictly correlated with the Bragg Peak position) can be reconstructed and validated during the treatment session

The INSIDE in-beam PET will soon be tested with patients



Acknowledgments

Giuseppe Battistoni Vincenzo Patera Valeria Rosso

ASIC designer Angelo Rivetti Manuel Rolo

INFN – Torino Technical Staff Giuseppe Alfarone Florea Dumitrache Marco Mignone Barbara Pini Mauro Scalise Antonio Zampieri

INSIDE PET team Maria Giuseppina Bisogni Niccolò Camarlinghi Piergiorgio Cerello Alberto Del GuerraElisa FiorinaGiuseppe Giraudo Matteo Morrocchi Francesco Pennazio Cristiana PeroniMaria Antonietta Piliero Giovanni Pirrone Richard Wheadon



Back-up slides



Lu background radiation

Lutetium background spectrum Firmware-decoded events Good background rate uniformity

CTR with 5 ns window ～ 30 cm head distance ～ 2 ns separation between peaks



Calibration: ‣ Test of high rate DAQ ‣ Channel map verification ‣ Energy calibration ‣ Time calibration

Coincidence time resolutionEntries 1.031614e+07Mean 0.006961RMS 0.4569

/ ndf 2χ 1.998e+05 / 120Constant 1.31e+02± 3.15e+05 Mean 0.00013± 0.01657 Sigma 0.000± 0.417

CTR [ns]2− 1.5− 1− 0.5− 0 0.5 1 1.5 2

coun

ts

0

50

100

150

200

250

300

350310× Coincidence time resolution

Entries 1.031614e+07Mean 0.006961RMS 0.4569

/ ndf 2χ 1.998e+05 / 120Constant 1.31e+02± 3.15e+05 Mean 0.00013± 0.01657 Sigma 0.000± 0.417

Coincidence time resolution

TOT window calibration

Data set: 2 68Ge rods on the beam axis, centered in the FOV, 06/06/2016, CNAO

CTR ～ 417 ps 𝜎



Bea

m te

st 0

3/04

/201

6, C

NA

O

High Voltage (SiPM)

Filtered events

Transmitted events

FIFO max

Tx board online data

Main firmware functionalities: ‣ TOT filtering ‣ Raw to decoded data format ‣ Channel by channel application of delays ‣ Channel by channel photopeak selection

(discarded) about 10x



Requirements and issues to face Strategies adopted

On-line and high rate → fast software

TOT filter and delays performed by FPGA, C language, multi-threading, lock-

free, no memory allocation during runtime

Compatibility BOOST v. 1.54Bandwidth > 1Gbit/s Two NIC used for acquisition, TOT filter

(Inspill rate/Interspill rate) > 10 Memory buffers to address average rate

Address edge cases in packets handling Software checks

Frames fragmented in packets, packets asynchronously sent by boards

Three steps in decoding and data processing, thread-safe containers

Accurate Compton filtering vs photopeak stability vs bandwidth

Conservative TOT window on FPGA boards, coincidences TOT saved

On-line monitoringCoincidences and single-events subset

monitoring with GUI, in- and inter-spill discrimination

Data acquisition safety GUI and coincidence finding software run on different machines

Coincidence finding software



Coincidence finding software: performances

‣ Program designed to keep running even in the event of data loss (packets-frames-events)

‣ Log file implemented to monitor performance

‣ Data loss is presently about 0.05% (i.e. negligible)



Coincidence finding software: performances

Maximum single rate: 22 MHz

Maximum coincidence rate: 570 KHz

Bea

m te

st 0

7/06

/201

6, C

NA

O

What happens at higher rate? ‣ Actual in-spill particle rate on detectors

depends on - Projectile particle - Beam energy - Target position wrt PET heads

(distance, axial position) - Path length of beam inside the FOV - Dose delivery/free beam

‣ Tested with success with proton beam until 218 MeV/u (28cm in water)



GUI-based on-line monitoring‣ On-line in- and inter-spill discrimination ‣ Coincidences saved in list-mode

compatible for reconstruction ‣ Calculate EoC-SoC calibration ‣ It can be used for fine TOT windows

calibration off-line as soon as data are saved

‣ Calibrate delays & TOT window and write calibration file

‣ Can load saved histograms



Master board (to be put beneath lower PET head to minimize

radiation damage)

Tx clock distribution Tx/Master communication

Master board relay

Relays Serial port Watchdog

board

ML601 Spartan 6 evaluation board

Tx board relays control

Master board relays control

Gb ethernet

serial connection to

chiller



Image Mask

3D Image Comparison

Image

Smoothing Filter

Threshold Filter

Erosion Filter

Dilation Filter

Mask

3D Image Mask Calculation:



V. Ferrero (INFN and University of Torino) - [email protected] SIF, Padova, 26-30 September 2016

3D Image Comparison

PT A1 PT A2 50

600

300

200

100

400

500

k_end3x3Entries 1610Mean 0.06807− RMS 0.6483

/ ndf 2χ 22.98 / 2Constant 26.7± 805.7 Mean 0.01627±0.08972 − Sigma 0.0137± 0.6292

z [mm]20− 15− 10− 5− 0 5 10 15 20

coun

ts

0

100

200

300

400

500

600

700

800

k_end3x3Entries 1610Mean 0.06807− RMS 0.6483

/ ndf 2χ 22.98 / 2Constant 26.7± 805.7 Mean 0.01627±0.08972 − Sigma 0.0137± 0.6292

End of activity distribution

PT A1/PT A2 comparison:

Mean (-0.09±0.01) mm RMS (0.63±0.01) mm

x [mm]0 20 40 60 80 100 120 140 160 180 200 220

y [m

m]

0

20

40

60

80

100

15−

10−

5−

0

5

10

15

End of activity comparison


The INSIDE project: in-beam PET scanner system features and

characterization

V. Ferrero, on behalf of the INSIDE Collaboration

14th Topical Seminar on Innovative Particle and Radiation Detectors, 3-6 October 2016, Siena

N. Belcari M.G. Bisogni N. Camarlinghi A. Del Guerra S. Ferretti E. Kostara A. Kraan B. Liu N. Marino M. Morrocchi

M.A. Piliero G. Pirrone V. Rosso G. Sportelli

A. Attili P. Cerello S. Coli V. Ferrero E. Fiorina G. Giraudo F. Pennazio C. Peroni A. Rivetti R. Wheadon

E. De Lucia R. Faccini P.M. Frallicciardi M. Marafini C. Morone V. Patera L. Piersanti A. Sarti A. Sciubba C. Voena

F. Ciciriello F. Corsi F. Licciulli C. Marzocca G. Matarrese

G. Battistoni M. Cecchetti F. Cappucci S. Muraro P. Sala

INSIDE coordinator: M.G. Bisogni

Documents

The INSIDE project: in-beam PET scanner system features ... · scanner system features and characterization V. Ferrero, on behalf of the INSIDE Collaboration ... Antonio Zampieri