Wai-Hoi Wong 1, Hongdi Li 2, Yuxuan Zhang 1, Rocio Ramirez 1,

Hossain Baghaei1, Shaohui An2, Chao Wang2, Shitao Liu2, Yun Dong2

1 University of Texas MD Anderson Cancer Center, Houston, TX,

2 Shanghai United Imaging Healthcare Co, Shanghai, China.

A high-resolution time-of-flight clinical PET detection

system using the PMT-quadrant-sharing technology

The PMT-Quadrant-Sharing (PQS) detector block

• LYSO is expensive (1/5 the price of gold)

• The Objective is to get higher sensitivity per cc of LYSO

• More efficient use of LYSO by increasing the axial field of view while reducing the crystal depth

System Design

• GEANT4 MC simulation

• But large AFOV increases PMT and electronics cost

• Use PMT-quadrant-sharing reducing PMT usage by 75% for increasing AFOV cheaply

• 15mm deep LYSO, 28-cm AFOV

Detector Ring Design• To achieve ultrahigh resolution using large PMT to increase AFOV

2.35 x 2.35 mm pitch

38-mm PMT 16 x 16 LYSO block

Achieve decoding 256 crystals per PMT usage

• Adapt PMT-Quad-Sharing blocks to a gapless detector-ring geometry

Ring has 24 modules (3 x 7 blocks)

3 blocks in-plane and 7 blocks axial

The edge blocks are half-ground to fit the quadrant-sharing PMT

27.6

-cm

axi

al F

OV

A detector module

The “Slab-Sandwich-Slice” (SSS) Detector Production

These are 15 sets of inter-slab irregular reflecting masks for

a 16 x 16 array to decode 256 crystals / PMT

Each reflecting mask can be cut into any shape providing

many degrees of freedom to optimize crystal decoding

There are 4 sandwich types in this 16 x 16 block

The SSS production method is highly uniform and preciseas shown in the decoding map of these 72 detector blocks

We use the 5th-generation HYPER Pileup-event-recovery

front-end electronics

Hybrid coincidence

NEMA Resolution Expectation (mm) Measured (mm)

Transaxial (1 cm) ≤2.8 mm 2.72

Axial (1 cm) ≤3.1mm 2.76

Trans Radial (10 cm) ≤3.3mm 3.36

Trans Tangential (10cm) ≤3.2mm 3.05

Axial (10cm) ≤3.6mm 2.96

NEMA image resolution measurement

Reconstruction algorithm: FBP2D (SSRB)

Pixel Size: 0.3 x 0.3 x 1.22

FWHM 0 cm 4 cm 8 cm 12 cm 16 cm 20 cm 24 cm 28 cm

Mid plane P (V) 1.47 1.55 1.64 1.78 1.76 1.89 1.97 2.46

Mid plane P (H) 1.55 1.44 1.60 1.61 1.56 1.81 1.77 1.79

¼ axis NP (V) 1.72 1.88 2.09 2.13 2.18 2.12 2.19 2.95

¼ axis NP (H) 2.00 2.03 2.19 2.28 2.58 3.13 3.90 4.46

V: Vertical H: Horizontal

PSF reconstructed image resolution (mm)

Average time-of-flight resolution 473 ps (+ 36 ps)

Very fine axial sampling, slice-to-slice separation = 1.22 mm

3 minutes/bed

4 bed positions

Oncology

MIPTOF + PSF2 min/bed, 2 iterations

PSF3 min/bed, 3 iterations

TOF + PSF ReconSNR: TOF/PSF 2 min/bed 2 iterations = PSF 3 min/bed 3 iterations 2/3 scan time, faster recon, no loss of detectability

SNR = (Signal – Background) / SDBackground See Jakoby, et al, Phys. Med. Biol. 56

Conclusions

• With PMT-quadrant-sharing Detector design we have developed an ultrahigh resolution TOF PET/CT

• It has a resolution of 2.8 mm using FBP (1.5 mm using PSF)

• Large axial FOV 27.6 cm, ultrafine axial sampling of 1.22 mm

• This large system with ultrahigh resolution uses only 576 PMT (reducing PMT and electronics cost, while increasing reliability)

• It has 129,024 detectors with a TOF resolution of 473 ps

These developments have been supported by:

· NIH-RO1- EB001038 PHS Grant· NIH-RO1- EB001481 PHS Grant· NIH-RO1- EB004840 PHS Grant· Shanghai United Imaging Healthcare Fund