SPECT(single photon emission computed tomography) 031-8001355-5 SPECT (single photon emission computed tomography) 20 30 SPECT OS-EM (ordered subsets-expectation maximization) OS-EM SNR (signal to noise ratio) 64 16 SPECT 3
16 OS-EM SNR OS-EM OS-EM 64 3/4 48 1/2 32 OS-EM 1/2 1/2 OS-EM 3/4 3/4 OS-EM (sensitivity) (specificity) ROC(receiver operating characteristic)
1 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 1 SPECT SPECT 2 SPECT - - - - - - - - - - - - - - - - - - - - - 8 OS-EM ( ordered subsets-expectation maximization) 3 OS-EM - - - 11 SNR (signal to noise ratio) SNR 4 - - - - - - - - - - - - - - - - - - - - 21 (steak free area) OS-EM subset iteration 5 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 25 6 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 26 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 28
1 1-1 (radioisotope : RI) RI (in vivo) (in vitro) RI RI RI RI SPECT (single photon emission tomography) RI 1896 Becquerel 1913 Hevesy 1
1925 Blumgart RI RI 1931 Lawrence 1934 Joliot 1965 Segre Seaborg ( 99m Tc) RI RI X CT (computed tomography) MRI (magnetic resonance imaging) 1) 1-2 1951 Cassen GM NaI(Tl) 1960 1970 1957 Anger 2
Fig.1(a) 1962 Kuhl SPECT 3 RI 2 RI Hounsfiled X X CT SPECT 2 3 mm 1mm 3
2) Fig.1(b) GE (General Electric) Starcam4000i SPECT 1-3 SPECT SPECT Fig. 2 2 Fig.2 1 360 3) 1-4 SPECT SPECT SPECT SPECT X CT MD (multi-detector row)ct 4
X CT 10 SPECT 20 30 RI 1 1-4 SPECT SPECT 10 SPECT 5
Larsson 4) (M) M ð 2 N (1) N pixel 64 N=64 M=100 M=60 64 100 Bieszk 5) 1 SPECT Cao 6) SPECT 180 64 32 64 33 180 64 6
SPECT SPECT 180 180 Cao (filtered back projection: FBP) streak artifact Fig.3(a) (d) (line source) OS-EM ( ordered subsetsexpectation maximization ) Fig.3(e) (h) SPECT OS-EM OS-EM 7
1 2 SPECT 2-1 SPECT X CT Fig.4 RI RI Ramachandran Butterworth RI (aliasing) 8
20 SPECT 2-2 OS-EM (ordered subsets-expectation maximization) (pixel),, (ML-EM: maximum likelihood-expectation maximization ) RI OS-EM 9
( ordered subsets- expectation maximization ) ML-EM (subset) Fig.5 24 subset 6 subset 4 1 subset 2 subset 4 subset 1 1 ML-EM subset subset (iteration) OS-EM subset iteration SPECT Fig.6 subset iteration subset iteration Fig.6 subset=8 iteration=4 OS-EM subset=8 iteration=4 10
OS-EM (1) (2) SNR(signal to noise ratio) (3) (4) 3) 7) (2) (3) OS-EM 3 OS-EM 3-1 3-1-1 SPECT SPECT 11
3 X 0.01mm X CT 0.5mm SPECT 10mm 15mm (point spread function) (full width at half maximum : FWHM) 1/10 (full width at tenth maximum : FWTM) FWHM 1/2 FWTM 1/10 8) SPECT (line source) FWHM FWTM 1mm ( 99m Tc ) GE entegra 12
SPECT Fig.7(a) (d) (low energy general purpose: LEGP) 64 64 (1pixel=5.40mm), SFOV(scan field of view) Fig.7(a) (d) OS-EM Fig.7(a) (b) FWHM OS-EM Fig.7(c) (d) FWTM OS-EM 64 X Y p<0.05 p<0.01 48 X p<0.01 32 X p<0.01 OS-EM OS-EM FWTM (1) SPECT 9) Joseph 10) SPECT 13
pixel Fig.2 d D 1 D 2 (C) C = (D 2 -D 1 )/D 1 (2) SPECT IB-10 Fig.8 99m Tc 12) Fig.9(a) Fig.9(b) D 2 D 1 (3) Fig.9 16 OS-EM OS-EM 16 14
SPECT SPECT (coefficient of variation : CV) CV= /m (3) m 1) 1 projection count (reconstruction count) Fig.10 1 50(count/pixel) 64 100(count/pixel) 32 40(count/pixel) 15
CV Fig.11 CV OS-EM CV OS-EM 20 25% OS-EM 16 OS-EM CV OS-EM subset 8 1 subset 2 iteration=4 3-1-2 SNR (signal to noise ratio) 3 trade-off SNR(signal to noise ratio) 16
SNR SNR=(D 2 -D 1 )/ (4) D 1 D 2 8) SNR 20cm 8cm 99m Tc 1/2 99m Tc SPECT SPECT 1/2 (5) SNR Fig.12 OS-EM SNR SNR OS-EM Table 1 3-2 17
SPECT SNR 13) n n 2 2 nc 2 =n(n-1)/2 S j S k S j t R jt R jt t R j S k R j R j R k R j -R k 18
R j R k = Z jk + 2 j 2 k 2r jk j k (5) R j R k S j S k Z jk R j R k p(r j>k ) j k R jt R kt r jk R jt R kt (5) p(r j R k ) Z jk 6 (Table 2) 1 100count/pixel 32 48 64 OS-EM Butterworth 0.50cycle/cm OS-EM Butterworth 0.50cycle/cm 64 48 3/4 32 1/2 Fig.13 6 Fuji FM-DPL (Fuji DI-ALc) 19
9 14 SPECT Z jk 100 0 Fig.14 64 OS-EM 100 64 99 48 93 48 OS-EM 65 32 OS-EM 65 0 32 3/4 1/2 OS-EM 3-3 SNR 3-2-1 SNR 3-2 SNR SNR Fig.15 SNR SNR 20
OS-EM 64 48 SNR 48 OS-EM 4 4-1 (streak free area) Bieszk 5) (N) N=2 RV m (6) R V m SPECT R=14(cm) Butterworth V m =0.50(cycle/cm) N 44 SPECT R=20(cm) V m =0.45(cycle/cm) N 50 21
1 64 (6) Fig.3 Fig.16 16 32 48 64 (6) 4-2 64 32 16 subtraction image Fig.17 16-64 32-64 22
4-1 16 OS-EM OS-EM 4-3 2 23
14) Fig.18 32 OS-EM 16 4 4 ML-EM Fig.6 64 subset 32 iteration 1 subset 64 subset 8 subset 16 4 iteration 4-4 OS-EM subset i teration OS-EM subset iteration 24
subset iteration 10 40 subset 1 subset iteration 5 SPECT OS-EM FWHM OS-EM FWTM OS-EM OS-EM 16 OS-EM subset=8 iteration=4 OS-EM 25
20 25% SNR OS-EM 64 OS-EM 100 48 64 3/4 93 OS-EM 65 64 1/2 32 OS-EM 41 0 OS-EM 3/4 1/2 OS-EM OS-EM SPECT SPECT 6 26
OS-EM 64 ROC (receiver operating characteristic) OS-EM 27
1) (2002) 2) (2001) 3) (2001) 4) S. Larsson, A. Israelsson Consideration on system design, implementation and computer processing in SPECT, IEEE Transactions on nuclear science, 1221-1342, NS-29 (4), (1982) 5) J. A. Bieszk, E. G. Hawman Evaluation of SPECT Angular Sampling Effects - Continuous Versus Step-and-Shoot Acquisition-, The Journal of Nuclear Medicine, 28, 1308-1314, (1987). 6) Zong Jian Cao, Lawrence E. Holder and Charles C. Chen Optimal Number of Views in 360 SPECT Imaging: J Nucl Med 37:1740-1744, (1996) 28
7) OS-EM (ordered subsets-expectation maximization) 57(5) 523-529 (2001) 8) Bruce H. Hasegawa The physics of MEDICAL X-RAY IMAGING MEDICAL PHYSICS PUBLISHING Wisconsin (1991) 9) SPECT 78-79 (2001) 10) Peter M. Joseph, Raymond A. Schultz View sampling requirements in fan beam computed tomography : Med. Phys. 7(6) : 692-702, (1980) 12) 13) 271-280 (2002) 14) SPECT 7-28 (1990) 29
Table 1 (FBP) OS-EM FBP OS-EM 16 (subset=8 iteration=4 ) SNR
Table 2 6 Sample View A FPP 64 1 B FBP 48 3/4 C FBP 32 1/2 D OS-EM 64 1 E OS-EM 48 3/4 F OS-EM 32 1/2
(a) (b) Fig.1 (a) (b) SPECT
1 2 Fig.2 SPECT 1 d pixel 2 360
(a) (b) (c) (d) (e) (f) (g) (h) Fig.3 OS-EM (a) e 64 (b) (f) 48 (c) (g) 32 (d) (h) 16
吸収 散乱のない理想的系 中央部付近の計数値が大きくなりすぎる 投影 RI の無いところに計数があらわれる 逆投影 Fig.4 SPECT
S2 S1 S6 S3 S4 S5 S4 S5 S6 S3 S2 S1 S2 S3 S1 S6 S4 S5 S6 S1 S2 S3 S4 S5 初期値 逐次近似 1 回 S1 S4 S6 S5 S3 逐次近似 2 回 S1 S4 S6 S5 S3 逐次近似 3 回 S1 S4 S6 S5 S3 24 投影データを 6 つの Subset に分割 Fig.5 OS-EM subset iteration
OS-EM OS-EM Fig.6 SPECT subset iteration 64 subset=8 iteration=4 (FBP) OS-EM subset iteration
(a) (b) (a) (b) (c) (d) (c) (d) Fig.7 FBP OS-EM X Y (a) X FWHM (b) Y FWHM (c) X FWTM (d) Y FWTM
Fig.8 SPECT IB-10 204mm 152mm 70mm
D1 D1 D2 D2 (a) (b) (c) Fig.9 (a) (b) (c) 5
140 120 RECONSTRUCTION COUNT (COUNT/PIXEL) 100 80 60 40 20 64view 48view 32view 16view 0 0 50 100 150 200 250 300 350 400 PROJECTION COUNT (COUNT/PIXEL) Fig.10 (projection count) (reconstruction count)
25 CV (%) 20 15 10 OS-EM FBP FBP 64VIEW FBP 48VIEW FBP 32VIEW FBP 16VIEW OS-EM 64VIEW OS-EM 48VIEW OS-EM 32VIEW OS-EM 16VIEW 5 0 0 50 100 150 200 COUNT/PIXEL Fig.11 (FBP) OS-EM CV
18 16 14 SNR (signal to noise ratio) 12 10 8 6 OS-EM Pos. OS-EM Neg. 4 OS-EM Average FBP Pos. 2 FBP Neg. FBP Average 0 16 32 48 64 80 VIEW NUMBER Fig.12 SNR (positive) (negative) SNR 64 48 3/4 32 1/2 20cm 10cm
(A) (B) (C) (D) (E) (F) Fig.13 SPECT (A), ( B), ( C) 64 48 32 (D), (E), (F ) OS-EM 64 48 32
100 99 93 100 SCALE VALUE OF PAIRED COMPARISON 80 60 40 20 65 41 0 0 A B C D E F SAMPLE Fig.14 SPECT Sample A, B, C 64 48 32 D, E, F OS-EM 64 48 32
100 A D B 80 SCALE VALUE OF PAIRED COMPARISON 60 40 F E 20 C 0 5 7 9 11 13 SNR (signal to noise ratio) Fig.15 Fig.12 SNR Fig.13 SNR SNR A F Table 2
RLATIVE COUNT (%) 100 80 60 40 16VIEW 32VIEW 48VIEW 64VIEW 20 0 0 20 40 60 80 PIXEL NUMBER RLATIVE COUNT (%) 10 9 8 7 6 5 4 3 16VIEW 32VIEW 48VIEW 64VIEW 2 1 0 0 20 40 60 80 PIXEL NUMBER Fig.16
FBP: 64view 32view 16view (a) (b) Subtracted image: 32-64 view 16-64view (c) OS-EM: 64view 32view 16view (d) Subtracted image: 32-64 view 16-64view Fig.17 (a) 64 (a) 32 (a) 16 (b) 32 64 (b) 16 64 (c) 64 OS-EM (c) 32 (c) 16 (d) 32 64 (d) 16 64
FBP OS-EM 64view 48view 32view 16view 4view Fig.18 4 ML-EM