Professional Documents
Culture Documents
On_the_Iterative_Image_Space_Reconstruction_Algorthm_for_ECT
On_the_Iterative_Image_Space_Reconstruction_Algorthm_for_ECT
On_the_Iterative_Image_Space_Reconstruction_Algorthm_for_ECT
1, MARCH 1987
D. M. TITTERINGTON
Abstract-The ISRA of [1] is shown to be an iterative algorithm that It is assumed that the (aij ) are known.
aims to converge to the least-squares estimates of emission densities. Suppose we start from an initial set of positive esti-
Convergence is established in the case where a unique least-squares
estimate exists that is, elementwise, strictly positive. It is pointed out
mates (X)(o) )
that, in terms of asymptotic theory, the resulting estimators are infe- The EM algorithm generates a sequence of sets of es-
rior to the maximum likelihood estimators, for which the EM algo- timates (X(k) ), according to the following iterative step.
rithm is a computational procedure. Potential difficulties with the be- For each k = 0, 1, * * * , and for each j = 1, . . * , J,
havior of the ISRA are illustrated using very simple examples.
E [n*a&j/( airXr)] =d
Manuscript received August 18, 1986; revised November 25, 1986.
The author is with the Department of Statistics, University of Glasgow,
Glasgow G12 8QQ, Scotland. a constant which is the Lagrange multiplier associated
IEEE Log Number 8612992. with the constraint. Multiplying by Xj and adding over j
E |n*aij/(E airXr)3
A
= 1. (4) I = E
i
(n .ajj) :. aij
I
I
Z
r
airxr (8)
(Ej aijpj ), the set of expected values of (rn). in an obvious notation, so that ATA = B and ATn * = c,
Other estimates for (pj ), and hence for (Xj ), might be then the ISRA is
obtained by minimizing other distance measures between
(r*) and (Ej aijpj). Perhaps the simplest is the least- X(k+l) X(k)Cj b(b.X(k))
squares distance
j
(9)
or, equivalently, x
i
=
xj
Cj Zr bjr x"" r (10)
2
E n~I E aiixi - Note also that although, in the EM algorithm, Sj
i X(k+l) Sj Xj(k), this is not necessarily the case in the
If we denote the vector (n *) by n *, the vector (Xj ) by ISRA.
X, and the I x J matrix (aij ) by A, then this can be written IV. CONVERGENCE PROPERTIES OF THE ISRA
as
The satisfying convergence properties of (1) are dis-
(n A-A ) T(n *- AX) cussed in detail in [3].
where "T" denotes transpose. Suppose we write (9) in the form
The least-squares (LS) estimates, X, of X are obtained X(k+l) Gj(X(k))
by solving
where
A TAx = ATn*. (5)
If ATA is of rank J, then
G1( X) = Xj C/(j bjrXr), (11)
X= (ATA) ATn* (6)
explicitly. However, if J is large (ATA) may be either ill j 1, * * J, thus emphasizing the successive-approx-
= ,
conditioned or singular and formula (6) is unusable. In imations nature of the algorithm. Then an important guide
such cases methods based on regularization may well be to the convergence properties is provided by
the matrix
employed [4] to give estimates of X. These estimates will U( X), where
no longer be LS estimates, however. aGj(X)
We shall, from now on, assume that ATA is nonsingu-
lar. (This requires that I ) J. ) axk
The link with the ISRA can be forged by noting that the for j, k = 1, , J. It is then the case that, if X repre-
jth equation in (5) can be written sents a solution of X = G( X), local convergence to )
obtains if the eigenvalues of U( X) are all, in magnitude,
E aij E
i r
airAr=
En ajj, (7) less than one [5]. Otherwise, divergence may occur in at
least some direction.
Authorized licensed use limited to: Peking University. Downloaded on July 04,2024 at 12:05:29 UTC from IEEE Xplore. Restrictions apply.
54 IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. MI-6, NO. 1, MARCH 1987
6jk -
xj jk/( brjr)r
tablished a complete theory of convergence in such cases,
or a simple illustration of the phenomenon is given in Ap-
6jk - jk(X
say. pendix II.
What can be established is the following.
In Appendix I the following Theorem is proved about Suppose X is a fixed point of the iteration in which the
the eigenvalues of V( X). first s elements are zero and the rest positive. Suppose
Theorem: If ATA is positive definite and if X > 0 (i.e., ('0) ) are chosen with the same zero-structure as X. Then
xi > 0 for j = 1. , J ) is a fixed point of (9), then
the eigenvalues of V( X) are all positive and the maximum
an argument analogous to that in Appendix I forns the
basis of a proof that the ISRA will be locally convergent
eigenvalue of V( X) is unity. at X.
The following Corollary follows at once. It is, of course, not a realistic proposition to be able to
Corollary: Under the conditions of the Theorem, it fol- prespecify which elements of X(') should be set to zero,
lows that 0 < q < 1, where q is the maximum eigenvalue but in examples like that in Appendix II, it does appear
of U( X), so that local convergence of (9) to X will hold. that convergence from a positive X occurs to a point on
It also follows that U( X) is a singular matrix, although the finite boundary of the set X > 0. (This empirical re-
this does not affect the convergence of the algorithm. mark does not constitute a proof!)
Since, as remarked in Section III, a positive X is a fixed The final remark emphasizes that, in spite of the rela-
point of (2) if and only if it is a LS estimate of X, it fol- tionship between (2) [or (9)] and the solution of (5), and
lows that, when the LS estimate is strictly positive and of the results about convergence, it does not necessarily
ATA is nonsingular, then the ISRA will be locally conver- follow that an algorithm of the form (9) can be used, as a
gent to that estimate. general rule, to solve any given set of linear equations
Convergence may, however, be slow, particularly if the
matrix ATA is not close to being diagonal. As indicated BX = c,
above, the crucial quantity is 77, the largest eigenvalue of where B is J x J and is nonsingular. A counterexample
U( X). As an illustration, Table I provides numerical re- for a matrix B that is neither symmetric nor semidefinite
sults from a couple of examples with J = 3. In one case is provided in Appendix III.
ATA has a reasonably dominant diagonal, but, in the other,
the resulting values of n , for a variety of choices of X, are V. PRECISION OF LEAST-SQUARES ESTIMATES
very close to unity. It is generally the case that, in the context of positron
Some further remarks should also be made. emission tomography, the ML estimates from (1) will be
The first backs up the comments in [3, sect. 2.3.2], in different from the LS, or projected LS, estimates obtained
which the LS method is discussed; see also [4]. There, it from (2). In comparing them one must either assess them
is emphasized that the solution of (5), even if it is unique, empirically on particular sets of data or by simulation, or
may not be nonnegative. As remarked earlier, if X(') o) investigate their statistical properties. For the latter, it is
0, then the same is true of all future iterates X(k) from (2) appropriate to look at the covariance matrices of the two
or (9). If the solution X of (5) has negative components, types of estimator.
the limit of the sequence (X(k) ) (if the sequence con- The covariance matrix of an estimator X, written cov
verges) represents some projection of the least-squares es- (X), is a J X J matrix whose diagonal elements are the
A ~ ~ ~
~ ~ ~ ~ ~~
timates onto the nonnegative orthant, X > 0. The se- variances of the elements of X and whose off-diagonal ele-
quence cannot converge to a point in X > 0 because the ments are the covariances between different elements of
discussion following the Theorem would then require that X. The variances are particularly important because large
the resulting limit point should satisfy (5), a contradic- variance means low precision.
tion. Convergence, if it occurs, must be to some X in As in [3, sect. 2.2 ] explicit formulas for the covariance
which some elements are zero. Although we have not es- matrices are available only when asymptotic results
Authorized licensed use limited to: Peking University. Downloaded on July 04,2024 at 12:05:29 UTC from IEEE Xplore. Restrictions apply.
TITTERINGTON: ITERATIVE IMAGE SPACE RECONSTRUCTION ALGORITHM ss
and ACKNOWLEDGMENT
The referees' comments on the original version of this
0.8X1 + 0.7X2 (i 2).
paper were much appreciated.
=
Authorized licensed use limited to: Peking University. Downloaded on July 04,2024 at 12:05:29 UTC from IEEE Xplore. Restrictions apply.