Professional Documents
Culture Documents
2020-6-10 MC-2 Calculation Method - YW - RG - V3
2020-6-10 MC-2 Calculation Method - YW - RG - V3
2020-6-10 MC-2 Calculation Method - YW - RG - V3
from each transmit channel. The electric field E(r) contains three vectors, which is corresponding to
three components:
Ex (r ) E x ,1 ( r ) … E x ,n ( r ) ω 1
( ) (
E ( r )= E y ( r ) =S E (r )∙ ω= E y ,1 ( r ) … E y, n ( r ) ∙ …
Ez(r ) E z , 1 ( r ) … E z ,n ( r ) ω n )( )
The contribution matrix S E (r ) Similarly, the SAR value of each cell, SAR(r ), could be represented by:
2 Ex (r )
SAR ( r )=
σ ( r ) ∙|E ( r )|
2 ρ(r )
=
σ (r )
2ρ (r) (
∙ ( E¿x ( r ) ¿ ¿
( )
E ( r ) E ( r ) )∙ E y ( r ) =
y z
Ez (r )
σ (r )
2 ρ(r )
∙( EH (r )∙ E (r ))
For a mass-averaged SAR in a certain volume, such as whole-body SAR (wbSAR) or Head SAR (HeadSAR),
the SAR M is calculated as:
1 ❑ 1 ❑ σ (r ) H
SAR M = ∫
V M
SAR ( r ) dV (r )=ω H ∙
2V ∫
M ρ ((r ) )
∙ ( S E ( r ) ∙ S E ( r ) ) dV ∙ ω=ω H ∙ Q∙ ω
Notes:
1. The absolute magnitude of transmit vector ω is determined by the normalization value. Only the
difference of magnitude and phase is given for the initial transmit vector ω . Take -15dB with 220
degrees for example. If the transmit element of I channel is fixed as 1+0i, Q channel is -15dB=0.178
refer to I channel, with 220 degrees delay, exp(-1i*220/180*pi)= -0.7660 + 0.6428i. Hence, the
transmit element of Q channel is -0.1364 + 0.1144i. The transmit vectorω is [1, -0.1364 + 0.1144i].
2. The contribution matrix S E (r ) is not the contribution from unit input power of each channel.
Hence, the absolute magnitude of transmit vector does not mean the actual input power of each
transmit channel. However, The contribution matrix S E (r ) for I channel and Q channel is
proportional to the input of each corresponding channel. Hence, the normalization of incident
electric field is based on the ratio of Q matrix (SAR values), not the absolute power.
3. The electric field component, such as E x ,1 ( r ), and the others, such as electric conductivity σ ( r ) and
density ρ ( r ) are directly extracted from SEMCAD simulation. As shown in figure below, each .mat file
contains the information and results of corresponding simulation.
Take our simulation for example. Our 1.5T and 3T simulations have two channels, I channel and Q
channel. So, the contribution matrix S E (r ) is a 3-by-2 matrix, which is represented as:
E x , I ( r ) E x ,Q ( r )
(
S E ( r )= E y , I ( r ) E y, Q ( r ) .
E z , I ( r ) E z ,Q ( r ) )
The product of S HE ( r ) ∙ S E ( r ) is a 2-by-2 matrix:
E x , I ( r ) E x, Q ( r )
H E ¿x , I ( r ) E¿y , I ( r ) E¿z , I ( r )
(
S ( r ) ∙ S E ( r )= ¿
E
E x ,Q ( r ) E ¿y ,Q ( r ) E ¿z , Q ( r ) )(
∙ E y, I ( r ) E y ,Q ( r )
E z , I ( r ) E z ,Q ( r ) )
Hence, the Q matrix is also a 2-by-2 matrix. The integral is conducted in each matrix element. Take
whole-body region for example, the Q(1,1), Q(1,2), Q(2,1) and Q(2,2) are calculated as:
1
Q(1,1)= ∑ [ σ ( r ) ∙ ( E ¿x , I ( r ) ∙ E x, I ( r ) + E¿y , I ( r ) ∙ E y, I ( r ) + E¿z , I ( r ) ∙ E z , I ( r ) ) ∙ V (r ) ]
2M M
1
Q(1,2)= ∑ [ σ ( r ) ∙ ( E ¿x , I ( r ) ∙ E x, Q ( r )+ E¿y , I ( r ) ∙ E y ,Q ( r )+ E ¿z , I ( r ) ∙ E z , Q ( r ) ) ∙ V (r )]
2M M
1
Q ( 2,1 )= ∑ [ σ ( r ) ∙ ( E ¿x ,Q ( r ) ∙ E x , I ( r )+ E¿y ,Q ( r ) ∙ E y , I ( r )+ E ¿z , Q ( r ) ∙ E z , I ( r ) ) ∙ V ( r ) ]=Q ¿ (1,2)
2M M
1
Q(2,2)= ∑ [ σ ( r ) ∙ ( E ¿x ,Q ( r ) ∙ E x ,Q ( r )+ E ¿y ,Q ( r ) ∙ E y ,Q ( r ) + E¿z ,Q ( r ) ∙ E z , Q ( r ) ) ∙V (r )]
2M M
Especially, Q(1,1) and Q(2,2) are real numbers. Q(1,2) and Q(2,1) are complex numbers. One example for
Duke model at landmark 0 is shown below:
Similar to HeadSAR and B1 calculation, each one has a corresponding 2-by-2 Q matrix.
The B1 (r ) field of each cell is also calculated by a linear superposition of the magnetic flux density field
contribution from each transmit channel. The B1 (r ) field contains two rotating scalars
B1 ( r )=¿
1❑ 1❑
B 1= ∫
S S SS ( 1 )
B1 (r )dr = ∫ S B (r )dr ∙ ( ω ω¿ )=QB ∙ ( ω ω¿ )
1
B1+¿=B (1,1 )¿
1
B1−¿=B ( 2,2) ¿
1
Our 1.5T and 3T simulations have two channels, I channel and Q channel. So the Q B matrix is also a 2-
1
by-2 matrix. The integral is conducted in each matrix element. The Q(1,1), Q(1,2), Q(2,1) and Q(2,2) are
calculated as:
❑
1
QB ( 1,1 )= ∫ S B dr ¿
1
S S 1+ ¿,Q
(r )
❑
1
QB ( 1,2 )= ∫S dr ¿
1
S S B 1+ ¿,I
(r)
❑
1
QB ( 2,1 )= ∫S dr ¿
1
SS B 1−¿,Q
(r )
❑
1
QB ( 2,2 )= ∫ S B dr ¿
1
SS 1−¿,I
(r)
B1=¿
And
B 1
❑
1
❑
1+¿=B1 ( 1,1 )= ∫S dr ∙ ω Q+ ∫S dr ∙ ωI ¿ ¿¿
S S B 1+ ¿, Q (r)
S S B1+¿ , I (r)
B 1
❑
¿ 1
❑
¿
1−¿=B 1( 2,2) = ∫S dr ∙ ωQ+ ∫S dr ∙ω I ¿¿ ¿
S S B 1−¿ ,Q (r )
S S B 1−¿ ,I (r )
Take the given source, [1, -0.1364 + 0.1144i] for example, the calculated B1 field (uT) is:
¿
Take source [1,1i], which is CW for example, the calculated B 1 field (uT) is:
¿
Take source [1,-1i], which is CCW for example, the calculated B 1 field (uT) is:
¿
Take source [1, 0.7071 - 0.7071i], which is 45deg ahead, for example, the calculated B 1 field (uT) is:
¿
Take source [1, 0], which is linear polarization, for example, the calculated B 1 field (uT) is:
¿
For circular polarization, the B1RMS value is directly calculated as B1 since the circular component of B1
field has negligible phase difference to actual field. However, for elliptical or linearly polarization field,
B1RMS value should be different from B1. The B1RMS table below is calculated as circular polarization.
With given Q matrix and the transmit vector, the scaling factor is determined based on the
normalization requirement. Take whole-body SAR for example, with given Q matrix shown above, the
calculated whole-body SAR is
ω H ∙Q ∙ ω=0.3468
Hence, the scaling factor is sqrt(2/0.3468)=2.4015. Similarly, the scaling factor for HeadSAR=3.2W/kg
can also be determined. To meet the requirement of both HeadSAR and whole-body SAR, the smaller
scaling factor is selected. This scaling factor is various from each human model and landmark. Based on
this scaling factor, the actual whole-body SAR, HeadSAR and B1+rms are re-calculated. The
normalization results for Duke model are shown in the table below:
With these scaling factor, the tangential electric field is normalized and multiplied with transfer function.
Appendix:
In this appendix, I will give the calculation of B1+¿ ¿ and B1−¿¿, based on the superposition of B1+¿ ¿, B1−¿¿
and Bx , B y , from Q and I channel, respectively. This is the proof that the calculation method for B1+¿ ¿
and B1−¿¿ in this document is correct.
B Bx + j B y
1+¿= ¿
2
B ( Bx − j B y )
¿
1−¿= ¿
2
S SB +jSB
B1+¿ = x
¿ y
S ( SB − j S B )
¿
B1−¿ =
x y
¿
2
Assign that:
❑
1
QB = ∫ S dr
x
S S B x
❑
1
QB = ∫ S B dr
SSy y
Q B ( 1,1 )=Q
1 1
❑
1 S B ,Q
❑
1 j S B ,Q QB ,Q + j QB ,Q
❑
B1+ ¿,Q= ∫S
S S B 1+¿ ,Q (r ) dr= ∫
S S 2
dr+ ∫
SS
x
2
dr=
2
¿¿ y x y
Q B ( 1,2 )=Q
1 1
❑
1 S B ,I
❑
1 j SB , I
❑
Q B ,I + j Q B , I
B1+ ¿,I = ∫S
S S B 1+ ¿ ,I (r ) dr= ∫ dr+ S ∫ 2 dr=
S S 2
x
2
¿¿ y x y
QB ( 2,1 )=Q ¿ ¿ ¿
)¿ (Q B ,Q − j QB ,Q) ¿
1 ( SB ,Q) ( j SB ,Q ) ( QB ,Q ) − ( j QB
❑ ❑ ❑
1
∫ 2 dr− 1S ∫ 2 dr=
1 ,Q
B1−¿,Q= ∫SB
S S 1−¿ ,Q
(r ) dr=
S S
x
2
y x y
=
x
2
y
¿¿
S
QB ( 2,2 )=Q ¿ ¿ ¿
)¿ (Q B ,I − j QB ,I )¿
1 ( SB ,I ) 1 ( jSB ,I) ( QB , I ) − ( j Q B
❑ ❑ ❑
1 1 ,I
B1−¿,Q= ∫SB
S S 1−¿ ,I
(r ) dr= ∫
S S 2
dr− ∫
S S
x
2
dr=
2
y x y
=
x
2
y
¿¿
B ( QB ,Q + jQ B ,Q ) (Q B ,I + jQ B , I )
1+¿=QB ( 1,1) ∙ωQ +QB ( 1,2 ) ∙ ω I= ωQ + ωI ¿
x y x y
1 1
2 2
B (Q B ,Q− j QB ,Q )
¿
(QB ,I − j QB ,I )
¿
¿ ¿ ¿ ¿
1−¿=Q B ( 2,1 ) ∙ ωQ+ QB ( 2,2) ∙ωI = ωQ + ωI ¿
x y x y
1 1
2 2
QB QB
QB =
( QB
x
y
,Q
,Q QB
x
y
,I
,I
) ,
where
❑
1
QB = ∫ SB dr
S S
x x
❑
1
QB = ∫ S dr
y
SS B y
The linear combination of Q B and ω will give the total Bx and B y field.
Bx Q ∙ ωQ +Q B , I ∙ω I
( )
By
=Q B ∙ ω= B
QB ( x
y
,Q
,Q ∙ ωQ +Q B ,I ∙ ω I
. x
y
)
The definition of B1+¿ ¿ and B1−¿¿ are given as:
B Bx + j B y
1+¿= ¿
2
B ( Bx − j B y )
¿
1−¿= ¿
2
B ¿
( j (Q B ,Q ∙ωQ +QB ,I ∙ ωI )) (Q B ,Q − j QB ,Q)
¿ ¿ ¿
( QB , Q ∙ ωQ +Q B ,I ∙ωI ) (Q B ,I − j QB ,I )
¿
( Bx − j B y ) y y ¿ ¿
1−¿= ωQ + ωI ¿
y y x y x y
= − =
2 2 2 2 2
¿
( QB ,Q − j QB ,Q ) ( 0.0097− j 0.1622) ∙ 10−5 j (−0.0051+ j0.1259 ) ∙10−5 −6
=(−0.5809− j 0.8366 ) ∙ 10
x y
= +
2 2 2
Next, let us validate the B1+¿ ¿ and B1−¿¿. Use source [1,-1j] for example.
B1+¿=Q ( 1,1) ∙ωQ +QB ( 1,2 ) ∙ ωI= ( 0.6783+ j 0.7859 ) ∙10−6 + (−0.5626+ j 0.7696 ) ∙ 10−6 ∙ ( −1 j) = ( 1.4479+ j 1.3485 ) ∙ 10−6 ¿
B1 1
B1−¿=Q ( 2,1 ) ∙ ω¿Q+ QB ( 2,2) ∙ω¿I = ( −0.5809− j 0.8366 ) ∙10 −6+ ( 0.5887− j 0.7091 ) ∙10 −6 ∙ ( 1 j ) =( 0.1282− j 0.2479 ) ∙ 10−6 ¿
B1 1
B ( QB ,Q+ j QB ,Q ) ( QB , I+ j QB ,I ) −5
( 0.0097+ j 0.1662 ) ∙10 + j ( −0.0051− j0.1259 ) ∙10
−5 −5
( 0.0026+ j 0.1479 ) ∙ 10 + j ( 0.0061+0.1151) ∙10
−5
−6
1+¿= ωQ + ωI = ∙ (− 1 j ) = ( 1.4479+ j 1.3485 ) ∙10 ¿
x y x y
+
2 2 2 2
B ( QB ,Q − j QB ,Q )
¿
(Q B ,I − j QB ,I )
¿ −5 −5
( 0.0026− j 0.1479 ) ∙ 10 + ( 0.0061− j 0.1151 ) ∙ 10
−5
( 0.0026− j 0.1479 ) ∙ 10 + j ( 0.0061− j 0.1151) ∙10
−5
¿ ¿ −6
1−¿= ωQ + ωI = ∙ ( −1 j )= ( 0.1282− j 0.2479 ) ∙10
x y x y
+
2 2 2 2
Conclusion
The method 1 and 2 are equivalent for the calculation of B1+¿ ¿ and B1−¿¿