A New Hybrid Inversion Method For 2D Nuclear Magnetic Resonance Combining TSVD And Tikhonov Regularization

Journal of
Imaging
Article
A New Hybrid Inversion Method for 2D Nuclear Magnetic
Resonance Combining TSVD and Tikhonov Regularization
Germana Landi 1 , Fabiana Zama 1,* and Villiam Bortolotti 2

Citation: Landi, G.; Zama, F.;
Bortolotti, V. A New Hybrid
Inversion Method for 2D Nuclear
Magnetic Resonance combining
TSVD and Tikhonov Regularization. J.
Imaging 2021, 7, 18. https://doi.org/
10.3390/jimaging7020018
Academic Editor: Nicolas Papadakis
Received: 28 December 2020
Accepted: 25 January 2021
Published: 28 January 2021
Publisher’s Note: MDPI stays neu-
tral with regard to jurisdictional clai-
ms in published maps and institutio-
nal affiliations.
Copyright: © 2021 by the authors. Li-
censee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and con-
ditions of the Creative Commons At-
tribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1 Department of Mathematics, University of Bologna, 40126 Bologna, Italy; germana.landi@unibo.it
2 Department of Civil, Chemical, Environmental, and Materials Engineering, University of Bologna,
40126 Bologna, Italy; villiam.bortolotti@unibo.it
* Correspondence: fabiana.zama@unibo.it; Tel.: +39-051-209-4480
Abstract: This paper is concerned with the reconstruction of relaxation time distributions in Nuclear
Magnetic Resonance (NMR) relaxometry. This is a large-scale and ill-posed inverse problem with
many potential applications in biology, medicine, chemistry, and other disciplines. However, the
large amount of data and the consequently long inversion times, together with the high sensitivity
of the solution to the value of the regularization parameter, still represent a major issue in the
applicability of the NMR relaxometry. We present a method for two-dimensional data inversion
(2DNMR) which combines Truncated Singular Value Decomposition and Tikhonov regularization in
order to accelerate the inversion time and to reduce the sensitivity to the value of the regularization
parameter. The Discrete Picard condition is used to jointly select the SVD truncation and Tikhonov
regularization parameters. We evaluate the performance of the proposed method on both simulated
and real NMR measurements.
Keywords: hybrid regularization method; truncated singular values decomposition; tikhonov
method; Nuclear Magnetic Resonance (NMR) relaxometry
1. Introduction
Nuclear Magnetic Resonance (NMR) relaxometry has become an important tool to
study the molecular structure and properties of materials. A typical NMR experiment
consists of measuring the relaxation process due to the re-establishment of the nuclear
system into its equilibrium state, after the application of a short magnetic pulse parameter-
ized with a predefined flip angle. The relaxation process is described by longitudinal and
transversal dynamics, characterized by distributions of longitudinal (T1) and transversal
(T2) relaxation times [1]. The computation of the relaxation times distribution requires the
numerical solution of a Fredholm integral equation with separable Laplace-type kernels.
In particular, we focus on the inversion of 2D NMR relaxation data, acquired using a
conventional Inversion-Recovery (IR) experiment detected by a Carr-Purcell-Meiboom-Gill
(CPMG) pulse train [2]. Then, the evolution time t1 in IR and the evolution time t2 in
CPMG are two independent variables, and the 2D NMR relaxation data S(t1, t2) can be
expressed as:
S(t1, t2) =
∞
ZZ
0
k1(t1, T1)k2(t2, T2)F(T1, T2) dT1 dT2 + e(t1, t2) (1)
where the unknown F(T1, T2) is the distribution of T1 and T2 relaxation times, e(t1, t2) rep-
resents Gaussian additive noise and k1(t1, T1), k2(t2, T2) are Laplace-type kernels given by:
k1(t1, T1) = 1 − 2 exp(−t1/T1), k2(t2, T2) = exp(−t2/T2) (2)
J. Imaging 2021, 7, 18. https://doi.org/10.3390/jimaging7020018 https://www.mdpi.com/journal/jimaging

J. Imaging 2021, 7, 18 2 of 23
whose singular values quickly decay to zero. Since the unknown function F corresponds to
distribution of the values of the relaxation times T1 − T2, we can assume F(T1, T2) ≥ 0 for
all T1 and T2. Experimental data is usually collected at discrete values of times; therefore by
considering M1 × M2 samples of the times t1 and t2, and N1 × N2 samples of the relaxation
times T1 and T2, problem (1) is discretized as:
Kf + e = s, K = K2 ⊗ K1 (3)
where K1 ∈ RM1×N1 , K2 ∈ RM2×N2 are the discretized exponential kernels, s ∈ RM,
M = M1 · M2, is the discrete vector of the measured noisy data (which can have negative
values), f ∈ RN, N = N1 · N2, is the vector reordering of the unknown distribution and
e ∈ RM is the vector with the discretized noise.
The inversion of (3) is severely ill-conditioned and several direct and iterative methods
have been deeply discussed both in NMR [3–5] and mathematical [6–8] literature. Direct
regularization includes Truncated Singular Value Decomposition (TSVD) and Tikhonov
methods. TSVD replaces K by a matrix of reduced rank thus avoiding noise amplification
in the inversion procedure. This is usually obtained by removing the singular values
smaller than a given tolerance τ 0. Tikhonov-like regularization substitutes (3) with a
more stable optimization problem whose objective function includes a data-fitting term
and a regularization term imposing some a priori knowledge about the solution. In NMR,
Tikhonov-like regularization is usually employed. It requires the solution of a non-negative
least squares (NNLS) problem of the form
min
f≥0
n
kKf − sk2
+ αkfk2
o
(4)
where α is the regularization parameter. In the sequel, k · k will denote the Euclidean
norm of a vector or the Frobenius norm of a matrix. It is well-known that α controls the
smoothness in f and makes the inversion less ill-conditioned but, if wrongly chosen, it may
cause a bias to the result. For this reason, many parameter selection techniques have been
proposed such as Generalized Cross Validation (GCV), the L-curve and U-curve methods
and methods based on the Discrepancy Principle (DP) (see [6] and references therein). In
NMR, the Butler, Reed, and Dawson (BRD) method [9] or the S-curve method [10] are
usually applied for the selection of the regularization parameter.
Another critical issue when solving NMR problems is that the amount of data collected
for one measurement can be very large, with typical sizes M1 = 30–150 and M2 = 1000–50,000.
If a grid used for the estimated distribution has 100 × 100 elements, then the matrix K
can have up to 5 · 109 elements and the computational cost of the data inversion may be
extremely high.
The separable kernel structure of K has been widely exploited in the NMR literature
to perform data compression using independent SVDs of K1 and K2 and reduce compu-
tational complexity. The method proposed in [4], in the sequel referred to as Venkatara-
manan–Song–Hürlimann (VSH) algorithm, is considered a reference method for NMR data
inversion and it is widely used both in academia and industrial world. Given the TSVD of
K1 and K2:
Kki,i = Uki,iΣki,iV
T
ki,i, i = 1, 2 (5)
where ki is the number of considered singular values, the VSH method approximates the
original problem (4) with the better conditioned one
min
f≥0
n
kKk1,k2
f − sk2
+ αkfk2
o
(6)
where Kk1,k2
= Kk2,2 ⊗ Kk1,1. The data s is then projected onto the column subspace of
Kk1,k2
in order to obtain a NNLS problem with compressed data and kernel of significantly
smaller dimensions. A method adapted from the BRD algorithm is then used to solve the
optimization problem.

J. Imaging 2021, 7, 18 3 of 23
Other methods have been proposed in the literature for the inversion of 2DNMR
data. The algorithm of Chouzenoux et al. [11] uses a primal-dual optimization strategy
coupled with an iterative minimization in order to jointly account for the non-negativity
constraint in (4) and introduce a regularization term. A preconditioning strategy is used to
reduce the computational cost of the algorithm. In [12], an algorithm for 2DNMR inversion
from limited data is presented. Such algorithm uses compressive sensing-like techniques
to fill in missing measurements and the resulting regularized minimization problem is
solved using the method of [4]. The Discrete Picard Condition (DPC) [13] is used to choose
the truncation parameters. The 2DUPEN algorithm [14] uses multiparameter Tikhonov
regularization with automatic choice of the regularization parameters. Without any a-
priori information about the noise norm, 2DUPEN automatically computes the locally
adapted regularization parameters and the distribution of the unknown NMR parameters
by using variable smoothing. In [15], an improvement of 2DUPEN (I2DUPEN), obtained
by applying data windowing and SVD filtering, is presented. The SVD filtering is applied
to reduce the problem size as in [4].
Since the matrices K1 and K2 have typically small sizes, it is possible to compute the
exact SVD of K by exploiting its Kronecker product structure. In fact, if K = UΣVT is the
SVD of K, then
U = U2 ⊗ U1, Σ = Σ2 ⊗ Σ1, V = V2 ⊗ V1. (7)
However, the TSVD of K cannot be computed as a Kronecker product. For this reason,
different approximation of the TSVD can be found in the literature such as the matrix Kk1,k2
of the VSH method or the randomized SVD (RSVD) [16,17] where randomized algorithms
are used to approximate the dominant singular values of K. In this work, we show that it is
possible to compute the exact TSVD of K efficiently, avoiding approximations and suitably
using properties of the Kronecker product structure. Let Kk = UkΣkVT
k be the TSVD of K
where k is the number of considered singular values; we propose to solve the following
Tikhonov-like problem:
min
f≥0
n
kKkf − sk2
+ αkfk2
o
. (8)
Moreover, we propose and test an automatic rule, based on the DPC, for the automatic
selection of both the TSVD truncation index and the Tikhonov parameter. Finally, we
analyze the filtering properties of our method compared to TSVD, Tikhonov and VSH
methods. Therefore, our approach can be consideredt o be a hybrid inversion method
that combines TSVD and Tikhonov regularization: Tikhonov regularization prevents from
discontinuities and artificial peaks and TSVD acts as a preconditioning strategy and reduces
the computational cost. Actually, other approaches do exist in the literature combining
RTSVD and Tikhonov regularization [16,17]. Such techniques apply randomized algorithms
to reduce large-scale problems to much smaller-scale ones and find a regularized solution
by applying some regularization method to the small-scale problems in combination with
some parameter choice rule. The key point of these randomized algorithms is that the SVD
of the original linear operator K is never computed.
Concerning the solution of (8), in the present paper we apply the Newton Projection
(NP) method [18] where the Conjugate Gradient (CG) method is used to solve the inner
linear systems. This technique guarantees the desired high accuracy and it has been success-
fully applied in the NMR context [14,15]. Gradient projection methods with acceleration
techniques such those discussed in [19] required more computational effort to solve (8) with
the required accuracy. However, the typical sparse structure of the solution, represented
by nonzero peaks over flat regions, could possibly be taken into account in solving (8), for
example by adding L1 penalties [20].
Summarizing, the contribution of this paper is twofold; first, the paper shows that
times distribution of improved quality can be obtained by using Kk instead of the separate
TSVDs of K1 and K2 without a significant increase in the computational complexity. In
fact, the computational cost of our approach is slightly greater than the cost of VSH and
considerably smaller than the cost of RSVD. Second, the paper describes an efficient method

J. Imaging 2021, 7, 18 4 of 23
for jointly selecting the SVD truncation index k and the Tikhonov regularization parameter
α and for solving the NMR data inversion problem.
The remainder of this paper is organized as follows. In Section 2 we introduce our
method and, in Section 3, we analyze its filtering properties. In Section 4, we provide
implementation details and discuss numerical results. Some conclusions are reported in
Section 5. Finally, the VSH method is described in Appendix C.
2. The Proposed Hybrid Algorithm
In this section we illustrate the details of our approach and formalize our proposed
Hybrid Algorithm 1. We first propose to approximate problem (4) with
min
f≥0
n
kKkf − sk2
+ αkfk2
o
. (9)
where Kk = UkΣkVT
k is the TSVD of K. By projecting the data vector s onto the column
subspace of Kk and by neglecting constant terms, we obtain the equivalent formulation:
min
f≥0
n
kUkUT
k Kkf − UkUT
k sk2
+ αkfk2
o
(10)
which can be written as:
min
f≥0
n
kΣkVT
k f − UT
k sk2
+ αkfk2
o
(11)
with compressed data UT
k s ∈ Rk and kernel ΣkVT
k . We observe that the solution of (11)
lies in a subspace of the column space of Vk. In the following paragraphs we develop
the solution steps of problem (11), the rule for the choice of the TSVD truncation index k,
and of the Tikhonov parameter α. Finally, we illustrate the filtering properties of our new
method and compare it to Tikhonov, TSVD and VSH methods.
2.1. Solution of the Minimization Problem
We use the Newton Projection (NP) method [18] to solve the constrained minimization
problem (11); for a detailed description of NP we refer the reader to Appendix B. We apply
the Conjugate Gradient (CG) method to solve the inner linear systems. To this purpose, it
is necessary to perform several matrix-vector products involving the matrices Uk and Vk.
Although Uk and Vk are parts of the matrices U and V, they do not inherit the Kronecker
product structure. However, we can show that matrix-vector products can be performed
efficiently by exploiting the structure of U and V.
Given a vector x ∈ RM, let zero(x, k) be the vector obtained by zeroing the last M − k
components of x:
zero(x, k) = (x1, x2, . . . , xk, 0, . . . , 0)T
. (12)
Using the Kronecker products property:
(A ⊗ B)vec(X) = vec(BXAT
). (13)
From Equation (7) we have
UT
k s = zero(UT
s, k) = zero(vec(U
T
1 SU2), k) (14)
and
VT
k f = zero(VT
f, k) = zero(vec(V
T
1 FV2), k). (15)
where S ∈ RM1×M2 is the matrix of the measured data s.t. s = vec(S) and F ∈ RN1×N2
represents the computed distribution s.t. f = vec(F). Thus, matrix-vector products
involving Uk and Vk can be efficiently performed by using the Kronecker product structure
of U and V and by setting to zero the last components of the resulting vector.

J. Imaging 2021, 7, 18 5 of 23
2.2. Choice of the Parameters k and α
The quality of the restored distribution f strongly depends on the values of both
the truncation parameter k and the Tikhonov parameter α. We propose to choose both
these values by using the Discrete Picard Condition (DPC) [6]. This condition, for ill-
conditioned inverse problems, guarantees that a good solution is obtained by keeping in
the SVD expansion for the solution f = ∑i
uT
i s
σi
vi only the coefficients uT
i s/σi such that the
Fourier coefficients uT
i s decrease on average faster than the singular values σi. A truncation
parameter k satisfying the DPC can be selected by visual inspection of the so-called Picard
plot (i.e., a plot of the quantities uT
i s and σi versus i). Alternatively, an index k satisfying the
DPC can also be selected by using automatic techniques such as those described in [21,22].
Once the value for k has been fixed by using the DPC, the value for α is set as
α = σ2
k (16)
since, as explained in [6], the value α = σ2
i represents the breakpoint at which the ith filter
factor changes nature for the Tikhonov method [6]. Therefore this choice is motivated by
the low-pass filtering properties of both TSVD and Tikhonov methods.
Summarizing the previous considerations, we outline the steps of our proposed
Hybrid Algorithm 1.
Algorithm 1: Hybrid algorithm
1: compute the SVDs of K1 and K2
2: compute Σ = Σ2 ⊗ Σ1
3: choose k by using the DPC
4: choose α = σ2
k
5: Apply the Newton projection method to solve the constrained problem
min
f≥0
kΣkVT
k f − UT
k sk2
+ αkfk2
3. Analysis of the Filtering Properties
In this section we prove that our Hybrid algorithm acts as a low-pass filtering method,
similarly to TSVD and Tikhonov methods, and we compare it to the filtering properties of
VSH. Let us first report some basic properties of the solution of the NNLS problem:
min
f≥0
kKf − sk2
. (17)
Assume that f∗ is a solution of (17), then it satisfies [23]:
[KT
Kf∗
− KT
s]i = 0, for all i such that f∗
i 0. (18)
Once defined the active set of f∗ by
A∗
= {i | f∗
i = 0}, (19)
we can define the diagonal matrix D∗ as
[D∗
]ii =

1, i /
∈ A∗;
0, i ∈ A∗.
(20)
Then, using the relation D∗f∗ = f∗, we obtain from (18)
D∗
KT
KD∗
f∗
− D∗
KT
s = 0 (21)

J. Imaging 2021, 7, 18 6 of 23
which are the normal equations of the following LS problem
min
f
kKD∗
f − sk2
. (22)
The following result characterises the minimum norm solution of (22).
Theorem 1. Let K ∈ RN×M, s ∈ RN and let D∗ be the diagonal matrix defined as (20) where A∗
is the active set of a local solution of (17). Then, the minimum norm solution of the least squares
problem (17) has the following expression
f∗
=
N
∑
i=1
uT
i s
σi
D∗
vi. (23)
To prove this result, we first need the following lemma whose proof is given, for the
reader’s convenience, in Appendix A.
Lemma 1. Let K = UΣVT be the SVD of K ∈ RM×N and let D∗ ∈ RN×N be the diagonal
matrix defined in (20). Then, the pseudo-inverse of UΣVTD∗ ∈ RM×N is given by
(UΣVT
D∗
)†
= D∗
VΣ†
UT
. (24)
We are now in the position to prove Theorem 1.
Proof. (Theorem 1) Using the pseudo-inverse notation, we can write the solution of the LS
problem (22) as:
f∗
= (KD∗
)†
s = (UΣVT
D∗
)†
s
and, using (24) we have:
f∗
= D∗
VΣ†
UT
s
hence
f∗
= D∗
N
∑
i=1
uT
i s
σi
vi =
N
∑
i=1
uT
i s
σi
D∗
vi.
Theorem 1 shows that the solution of the constrained problem (17) can be written in
terms of the SVD of matrix K as follows:
f∗
=
N
∑
i=1
uT
i s
σi
v̂i where v̂i = D∗
vi.
Obviously, the vectors v̂i may be linear dependent if f∗ lies in a subspace of RN. It
is well known that TSVD and Tikhonov methods both compute a filtered solution ffilt of
problem (22) with different filter functions φi(·) [6]. Using the result of Theorem 1, we can
express the nonnegatively constrained TSVD solution as
fTSVD =
N
∑
i=1
φTSVD
i (k)
uT
i s
σi
v̂i, v̂i = D∗
vi (25)
where D∗ is the diagonal matrix (20) and the filter factors are
φTSVD
i (k) =

1, if i ≤ k;
0, otherwise;
with k ∈ {1, . . . , N}. (26)

J. Imaging 2021, 7, 18 7 of 23
Analogously, the non negative Tikhonov solution is given by
fTIKH =
N
∑
i=1
φTIKH
i (α)
uT
i s
σi
v̂i, v̂i = D∗
TIKHvi (27)
where D∗
TIKH is the diagonal matrix (20) defined with respect to the active set A∗ of a local
solution of (4) and the filter factors are
φTIKH
i (α) =
σ2
i + α
σ2
i
, with α ∈ R+
. (28)
Equations (25) and (27) define a low-pass filter with vectors v̂i. The solution pro-
vided by the hybrid method is still a filtered solution whose filter factors depend on two
parameters; i.e.,
fHYBRID =
N
∑
i=1
φHYBRID
i (α, k)
uT
i s
σi
v̂i, v̂i = D∗
HYBRIDvi (29)
where D∗
HYBRID is the diagonal matrix (20) defined with respect to the active set A∗ of a
local solution of (11) and the filter factors are
φHYBRID
i (α, k) =





σ2
i + α
σ2
i
, if i ≤ k;
0, otherwise;
(30)
with k ∈ {1, . . . , N} and α ∈ R+. By properly choosing the parameters k and α, the filter
factors for low-frequency components can be set close to one while filter factors for high-
frequency components can be set close to zero. Figure 1 depicts the behaviour of the filter
factors obtained for the value α = 2.5 · 105 versus the singular values σi. The σi plotted on
the abscissa axes are the singular values of the matrix K of the experiment with simulated
NMR data (see Section 4.3). We can observe that for singular values σi larger than
√
α = 500
the filter factors φHYBRID
i behave as φTIKH
i (black dashdotted line) while for smaller values
they are as φTSVD
i (red dashed line).
Figure 1. The Hybrid (blue line), TSVD (red dashed line) and Tikhonov (black dashdotted line) filter
factors φi versus σi for α = 2.5 · 105.

J. Imaging 2021, 7, 18 8 of 23
We observe that also the VSH solution can be expressed in terms of the SVD of K
(see algorithm details in Appendix C). We define the index subset I(k1, k2) including the
indices of the singular values of K which correspond to singular values of Kk2,2 ⊗ Kk1,1:
I(k1, k2) = {i | σi ∈ diag(Σk2,2 ⊗ Σk1,1)}.
Thus, we have:
fVSH = ∑
i∈I(k1,k2)
φVSH
i (α, k1, k2)
uT
i s
σ2
i
v̂i, v̂i = D∗
VSHvi (31)
where D∗
VSH is the diagonal matrix (20) defined with respect to the active set A∗ of a local
solution of (6) and the filter factors are
φVSH
i (α, k1, k2) =





σ2
i + α
σ2
i
, i ∈ I(k1, k2);
0, otherwise;
(32)
with k1 ∈ {1, . . . , N1}, k2 ∈ {1, . . . , N2} and α ∈ R+. However, it is almost impossible to
determine values of the truncation parameters k1 and k2 such that the vectors v̂i for i ∈
I(k1, k2) only correspond to low-frequency components including meaningful information
about the unknown solution. An explanatory example will be further discussed in the
numerical Section 4. For this reason, the VSH method cannot be considered a low-pass
filtering method.
4. Numerical Results
In this section, we present some results obtained by our numerical tests with both
simulated and real 2DNMR measurements. We have compared the proposed Hybrid
method with the VSH method which is a reference method for 2DNMR data inversion.
Moreover, in the case of real data, the UPEN method has been considered as a benchmark
method. The considered methods have been implemented in Matlab R2018b on Windows
10 (64-bit edition), configured with Intel Core i5 3470 CPU running at 3.2GHz.
The relative error value, computed as kFex − Fk/kFexk, is used to measure the effec-
tiveness of the compared methods, while the execution time in seconds is used to evaluate
their efficiency (here, Fex and F respectively denote the exact and restored distributions).
The reported values of the execution time are the mean over ten runs. Since both methods
are iterative, in the following, we give some details about the implemented stopping criteria
and parameters.
4.1. Hybrid Method
The NP method is used for the solution of problem (11); it is described in detail in the
Appendix B. As initial iterate the constant distribution with values equal to one is chosen.
The NP iterations have been stopped on the basis of the relative decrease in the objective
function F(f) of (11); i.e., the method is arrested when an iterate f(k) has been determined
such that
F f(k)

− F f(k−1)

F f(k)
TolNP. (33)
The inner linear system for the search direction computation is solved by the CG
method with relative tolerance TolCG. The values TolNP = 10−3 and TolCG = 10−7
have been fixed in all the numerical experiments. A maximum of KmaxNP = 500 and
KmaxCG = 104 iterations have been allowed for NP and CG, respectively.

J. Imaging 2021, 7, 18 9 of 23
4.2. VSH Method
The VSH method consists of three main steps. In the first step, the data is compressed
using SVDs of the kernels; in the second step, the constrained optimization problem is
transformed to an unconstrained one in a lower dimensional space, by using a method
adapted from the BRD algorithm. In the third step, the optimal value of α is selected by
iteratively finding a solution with fit error similar to the known noise variance. The VSH
method is described in detail in the Appendix C. Here we just report the values of the
parameters required in our Matlab implementation of the VSH method.
Each iteration of the VSH method needs, for a fixed value of α, the solution of a
reduced-size optimization problem obtained from (6) by projecting the data s onto the
column subspace of Kk1,k2
. The Newton method is used for the subproblems solution; its
iterations have been stopped on the basis of the relative decrease in the objective function.
The values TolN = 10−3 and KmaxN = 500 have been fixed. The values TolCG = 10−7 and
KmaxCG = 104 have been set for the CG method solving the inner linear system. The outer
iterations of VSH have been terminated when two sufficiently close approximations of the
unknown distribution have been determined or after a maximum of 100 iterations; the
relative tolerance value TolVSH = 0.1 has been set. As initial choice for the regularization
parameter, the value α0 = 1 has been chosen.
4.3. Simulated NMR Data
The considered simulated distribution F(T1, T2), shown in Figure 2, is a mixture
of three Gaussian functions located at (T1, T2) given by (81.86, 12.84), (8.59, 6.66) and
(433.27, 59.4) ms with standard deviations (0.1, 0.1), (0.1, 0.25) and (0.25, 0.1) ms. We
have considered the Laplace-type kernels K1 and K2 of (2), we have sampled t1 loga-
rithmically between 0.001 and 3, and t2 linearly between 0.001 and 1 with N1 = 100,
N2 = 100. The 2D data have been obtained according to the 2D observation model (3) with
M1 = 2048, M2 = 128. A white Gaussian noise has been added to get a signal-to-noise
ratio (SNR) equal to 23 dB. We remark that this environmental setting corresponds to a
realistic T1–T2 measurement.
4.3.1. Models Comparison
We first compare the proposed hybrid inversion model (9) with the VSH model (6)
and with classic Tikhonov model (4). Figure 3 depicts the singular values of K, K1 and
K2. Clearly, the singular values of K are obtained by reordering in non increasing order
the diagonal elements of Σ = Σ2 ⊗ Σ1 and they are different from the diagonal elements
of Σk1,1 ⊗ Σk2,1. That is, some unwanted small singular values of K may be included in
Kk1,1 ⊗ Kk2,2, while some large singular values may be not considered. Figure 4 shows
the singular values of the Kk obtained for τ = 1 (blue line) and the singular values of
Kk1,1 ⊗ Kk2,2 for τ1 = τ2 = 0.5 (blue dotted line), τ1 = τ2 = 1 (black dashed line) and
for τ1 = τ2 = 5 (red dashdotted line). Observe that the singular values of Kk are always
different from those of Kk1,1 ⊗ Kk2,2. Considering the threshold value τ = τ1 = τ2 = 1,
we have that the number k = 82 of singular values of Kk is larger than that of Kk1,k2
,
given by k1 × k2 = 8 × 6. Comparing, in Figure 4, the blue line and the black dashed one
(obtained for τ1 = τ2 = 1), we observe that the singular values of Kk1,k2
include a few
elements smaller than τ, and miss some larger terms that should be included. Moreover,
if τ1 = τ2 = 0.5, we have that the number of singular values of Kk1,k2
is k1 × k2 = 9 × 7
which is closer to k = 82 but there are many singular values smaller than τ. Finally,
in the case τ1 = τ2 = 5, Kk1,k2
has only a few large singular values (k1 = 5, k2 = 4)
and many relevant solution coefficients are missing. The plots in Figure 4 indicate that,
when considering problem (6), it is highly probable to include in the solution also those
components dominated by noise (if τ is slightly too small) or to discard those components
dominated by the contributions from the exact right-hand side (if τ is too large).

J. Imaging 2021, 7, 18 10 of 23
Figure 2. Top: simulated T1–T2 map and surface distribution. Bottom: projections along the T1 and
T2 dimensions.
Figure 3. The singular values of K (blue line), K1 (red dashed line) and K2 (black dashdotted
line). For easier visualization, the singular values of K1 and K2 are plotted versus the integers
1, 1001, 1002, . . . , 100100.

J. Imaging 2021, 7, 18 11 of 23
Figure 4. Singular values of Kk obtained for τ = 0.1 (blue line) and the singular values of Kk1,1 ⊗ Kk2,2
for τ1 = τ2 = 0.5 (blue dotted line), τ1 = τ2 = 1 (black dashed line) and for τ1 = τ2 = 5 (red
dashdotted line).
A visual inspection of the Picard plot (Figure 5) indicates, for the hybrid method,
the choice τ = 1 for the truncation tolerance giving the value k = 82 of the truncation
parameter. The previous considerations suggest to choose the values τ1 = τ2 = 1 for the
VSH method.
Figure 5. The Picard plot for the simulated NMR data. Singular values σi (black dashed line), Fourier
coefficients uT
i s (blue continuous line) and solution coefficients uT
i s/σi (red circles) for i = 1, . . . , 500.
Once defined the projection subspaces for VSH and Hybrid methods, we solve the
optimization problems (6) and (11) by applying the same numerical solver (Newton Pro-
jection method NP) and changing the values of the regularization parameters α. In this
way we want to investigate how the different subspace selections affect the solutions
computed by the (6) and (11) models for different values of α. Moreover, we apply the
Tikhonov method (4) in which no subspace projection is applied. For this reason, we use
NP to solve all the different optimization problems, since we aim at comparing the three
different models for data inversion, independently from the numerical method used for
their solution.

J. Imaging 2021, 7, 18 12 of 23
For each model and for ten values of α, logarithmically distributed between 10−4
and 102, Table 1 reports the relative error values (columns labeled Err), the number of
NP iterations (columns labeled It), and the time in seconds required for solving the three
inversion models (columns labeled Time). The numerical results of Table 1 are summarized
in Figure 6 where the relative error behaviour (on the left) and the computational time
in seconds (on the right) are shown versus the regularization parameter values. Figure 7
shows the distributions estimated from the three models, for α = 1 (left column), α = 10−2
(middle column) and α = 10−4 (right column). Figures 8 and 9 report the corresponding
projections along the T1 and T2 dimensions. Figures 7–9 shows the importance of the
proper choice of the subspace where the computed solution is represented. Indeed, the
distribution computed by our method lies in a subspace of the column space of Vk while
the distribution determined by the VSH method belongs to a subspace of the space spanned
by the column vectors of Vk2,2 ⊗ Vk1,1.
The results of Table 1 and the plot of Figures 6 and 7 show that the Hybrid model (9)
is less sensitive to small values of α than the Tikhonov one, because the matrix Kk is better
conditioned than K. When the value of α is properly chosen the quality of distributions
given by these two models is comparable, but the solution of (9) requires less computa-
tional effort. Also the VSH model is less sensitive to small values of α, because Kk1,k2
is
better conditioned than K. Its solution has the least computational cost but the computed
distributions exhibit artifacts which are due to subspace where they are represented (see
Figures 7–9).
Table 1. Numerical results for the models comparison on simulated NMR data. Results obtained by
applying NP method to (6), (4) and (11) with τ = τ1 = τ2 = 1.
Hybrid Tikh VSH
α Err Time It Err Time It Err Time It
100 0.536 0.90 25 0.536 0.65 20 0.542 0.73 52
21.54 0.452 1.29 31 0.451 1.11 25 0.463 0.52 33
4.6416 0.390 1.74 31 0.389 1.56 26 0.413 0.65 33
1 0.356 0.99 15 0.336 1.79 26 0.372 0.70 36
0.2154 0.300 1.57 18 0.284 1.29 15 0.338 0.97 39
0.0464 0.272 5.39 47 0.240 3.75 33 0.318 0.73 27
0.01 0.257 3.87 31 0.215 7.43 53 0.316 0.58 23
0.0022 0.255 5.55 43 0.243 11.41 68 0.314 0.35 15
0.0005 0.252 4.73 32 0.261 9.94 48 0.315 0.51 20
0.0001 0.245 2.94 25 0.304 24.91 100 0.320 0.45 18
Figure 6. Relative error behaviour (left) and the computational time in seconds (right) versus the
regularization parameter values.

J. Imaging 2021, 7, 18 13 of 23
Figure 7. Estimated distributions from the Hybrid (top row), Tikhonov (middle row) and VSH (bottom row) models for
α = 1 (left column), α = 10−2 (middle column) and α = 10−4 (right column).
4.3.2. Methods Comparison
We now compare the Hybrid and VSH algorithms on the simulated NMR data. Table 2
reports the relative error values and the required time in seconds for several values of the
truncation parameter τ (Here, τ1 = τ2 = τ). Moreover, the table shows the computed
value for the regularization parameter α, the number of performed Newton (or Newton
Projection) and CG iterations (columns labeled It N and It CG, respectively) and it reports,
only for VSH, the number of outer iterations (column labeled It). The numerical results
show that the VSH method has the lowest computational complexity while the Hybrid
method gives the most accurate distributions. The execution time of the Hybrid method
is very low, although VSH is less time consuming. Figures 10 and 11 depict the best
distributions estimated by the Hybrid and VSH methods, i.e.,: the distribution obtained
with τ = 0.05 and τ1 = τ2 = 0.1, respectively. By visually comparing figures 10 and 11, we
observe some spurious small oscillations in the VSH distribution both in the boundary and
in the flat region, while the distribution computed by the Hybrid method is less affected by
such artefacts.

J. Imaging 2021, 7, 18 14 of 23
Figure 8. Projections along the T1 dimension from the Hybrid (top row), Tikhonov (middle row) and VSH (bottom row)
models for α = 1 (left column), α = 10−2 (middle column) and α = 10−4 (right column). The red dashed and blue
continuous lines respectively represent the exact projections and the computed ones.
Table 2. Numerical results for the methods comparison on simulated NMR data.
τ Method α Err Time It It N It CG
0.001
Hybrid 1.73 10−6 2.35 10−1 0.68 / 18 1700
VSH 1.06 10−1 3.08 10−1 1.60 50 198 14,800
0.005
Hybrid 6.52 10−5 2.34 10−1 0.75 / 21 2000
VSH 9.34 10−2 3.85 10−1 1.50 50 198 14,800
0.01
Hybrid 1.66 10−4 2.32 10−1 0.62 / 18 1700
VSH 9.34 10−2 3.85 10−1 1.50 50 198 14,800
0.05
Hybrid 6.80 10−3 2.44 10−1 0.45 / 13 2200
VSH 8.52 10−2 3.15 10−1 0.65 28 112 8400
0.1
Hybrid 1.35 10−2 2.41 10−1 0.80 / 23 2200
VSH 7.96 10−2 3.02 10−1 0.11 4 20 1600
0.5
Hybrid 6.15 10−1 3.23 10−1 0.95 / 21 2000
VSH 3.93 10−2 3.12 10−1 0.19 6 35 2900
1
Hybrid 1.17 100 3.23 10−1 0.79 / 23 2200
VSH 5.07 10−2 3.43 10−1 0.25 8 53 4500
5
Hybrid 5.51 10+1 5.08 10−1 1.36 / 43 3779
VSH 9.26 10−2 6.80 10−1 0.06 2 27 1008
10
Hybrid 1.16 10+2 5.08 10−1 0.53 / 20 1496
VSH 5.99 10−2 7.03 10−1 0.10 6 50 1153

J. Imaging 2021, 7, 18 15 of 23
Figure 9. Projections along the T2 dimension from the Hybrid (top row), Tikhonov (middle row) and VSH (bottom row)
models for α = 1 (left column), α = 10−2 (middle column) and α = 10−4 (right column). The red dashed and blue
continuous lines respectively represent the exact projections and the computed ones.
Figure 10. Hybrid method: restored T1–T2 distribution and projections along the T1 and T2 dimen-
sions (τ = 0.05).

J. Imaging 2021, 7, 18 16 of 23
Figure 11. VSH method: restored T1–T2 distribution and projections along the T1 and T2 dimensions
(τ = 0.1).
4.4. Real NMR data
In this section we compare the Hybrid and VSH methods using real MR measurements
obtained from the analysis of an egg yolk sample. The sample was prepared in NMR
Laboratory at the Department of Physics and Astronomy of the University of Bologna, by
filling a 10 mm external diameter glass NMR tube with 6 mm of egg yolk. The tube was
sealed with Parafilm, and then at once measured. NMR measurements were performed
at 25 °C by a homebuilt relaxometer based on a PC-NMR portable NMR console (Stelar,
Mede, Italy) and a 0.47 T Joel electromagnet. All relaxation experimental curves were
acquired using phase-cycling procedures. The π/2 pulse width was of 3.8 µs and the
relaxation delay (RD) was set to a value greater than 4 times the maximum T1 of the sample.
In all experiments RD was equal to 3.5 s. For the 2D measurements, longitudinal-transverse
relaxation curve (T1-T2) was acquired by an IR-CPMG pulse sequence (RD-πx-TI-(π/2)x-
TE/2-[πy-TE/2-echo acquisition-TE/2]NE). The T1 relaxation signal was acquired with
128 inversion times (TI) chosen in geometrical progression from 1 ms up to 2.8 s, with
NE = 1024 (number of acquired echos, echo times TE = 500 µs) on each CPMG, and
number of scans equal to 4. All curves were acquired using phase-cycling procedures.
The data acquisition step produces an ascii structured file (in the STELAR data format)
including M1 · M2 relaxation data s in (3) where M1 = 128 and M2 = 1024, and the vectors
t1 ∈ RM1 (TI inversion times), t2 ∈ RM2 (CPMG echo times). The file is freely available upon
email request to the authors. For the data inversion, in order to respect approximately
the same ratio existing between M1 and M2, we choose the values N1 = 64, N2 = 73 and
compute the vectors T1, T2 in geometric progression in the ranges of predefined intervals
obtained from the minimum and maximum values of the vectors t1, t2 respectively. Finally,
using (2) we compute the matrices K1 and K2.
We use the times distribution restored by the 2DUPEN method [14], shown in
Figure 12 (top line), as benchmark distribution as the UPEN method uses multiparameter
regularization and it is known to provide accurate results [3]. Obviously, 2DUPEN requires

J. Imaging 2021, 7, 18 17 of 23
more computational effort since it involves the estimation of one regularization parameter
for each pixel of the distribution.
By visual inspection of the Picard plot (Figure 13) the value τ = 1 has been chosen
for the hybrid method; the same value is fixed for the VSH method. Figure 13 shows the
singular values of K, K1 and K2. For the VSH method, we report the results obtained
at the first iteration since, in this case, they worsen as the iteration number increases.
Table 3 reports the T2 − T1 coordinates (in ms) where a peak is located, its hight in a.u.
(arbitrary unit) and the required computational time in seconds. Finally, Figures 12 and 14
illustrate the distribution maps, the surfaces restored and the projections along the T1 and
T2 dimensions; the results of the Hybrid method are more similar to those obtained by
2DUPEN, while the distribution provided by the VSH method seems to exhibits larger
boundary artifacts.
Figure 12. From top to bottom: UPEN, Hybrid and VSH restored T1–T2 maps (left) and surface distributions (right).

J. Imaging 2021, 7, 18 18 of 23
Table 3. Location and height of the peak in the restored distribution and required execution time
in seconds.
Method T2 T1 Peak Height Time
Hybrid 19.85 51.90 269.52 11.06
VSH 19.85 43.70 264.48 1.24
UPEN 19.85 51.90 544.23 71.93
Figure 13. Left: singular values (black dashed line) of K and solution coefficients (red circles). Only
the first 300 singular values are depicted. Right: Singular values of K1 (red dashed line) and K2
(black dashdotted line). The horizontal line corresponds to the value τ = 1.
Figure 14. Cont.

J. Imaging 2021, 7, 18 19 of 23
Figure 14. From top to bottom: UPEN, Hybrid and VSH projections along the T1 (left) and T2 (right) dimensions.
5. Conclusions
In this paper, we have presented a hybrid approach to the inversion of 2DNMR mea-
surements. This approach combines main features of Tikhonov and TSVD regularization in
order to jointly select a suitable approximation subspace for the restored distribution and
to reduce the computational cost of the inversion. The Picard condition is used to select, at
the same time, the Tikhonov regularization parameter and the SVD truncation parameter.
The numerical results show that the proposed hybrid method is effective, efficient and
robust. In our future work, we intend to complement the presented hybrid approach in the
2DUPEN method.
Author Contributions: Conceptualization, G.L. and F.Z.; methodology, G.L. and F.Z.; software, G.L.;
validation, G.L., F.Z. and V.B.; formal analysis, G.L. investigation, G.L. and F.Z.; resources, V.B.; data
curation, V.B.; writing—original draft preparation, G.L., F.Z. and V.B.; writing—review and editing,
F.Z. and V.B. All authors have read and agreed to the published version of the manuscript.
Funding: GNCS-INdAM.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: Data Availability Statement: No new data were created or analyzed in
this study. Data sharing is not applicable to this article.
Acknowledgments: This work was partially supported by Gruppo Nazionale per il Calcolo Scientifico—
Istituto Nazionale di Alta Matematica (GNCS-INdAM). We thank Leonardo Brizi of DIFA University
of Bologna.
Conflicts of Interest: The authors declare no conflict of interest.
Appendix A.
Proof. (Lemma 1) The thesis is obtained if we verify that D∗VΣ†
UT ∈ RN×M meets the
four Moore-Penrose conditions which define the pseudo-inverse A† of A as the unique
matrix satisfying
1. AA†A = A;
2. A†AA† = A†;
3. (AA†)H = AA†;
4. (A†A)H = A†A.
By substituting UΣVTD∗ for A and D∗VΣ†
U for A† in the first condition, we have:
UΣVT
D∗
(D∗
VΣ†
UT
)UΣVT
D∗
= UΣVT
D∗
(D∗
(V(Σ†
((UT
U)Σ)VT
)D∗
) = UΣVT
D∗

J. Imaging 2021, 7, 18 20 of 23
since Σ†
Σ = IN, UTU = IM, VVT = IN and D∗D∗ = D∗. Analogously, we can prove the
second condition:
D∗
VΣ†
UT
(UΣVT
D∗
)D∗
VΣ†
UT
= D∗
VΣ†
U. (A1)
We have that
(AA†
)H
= (UΣVT
D∗
D∗
VΣ†
UT
)H
= UΣ†
VT
D∗
D∗
VΣUT
= AA†
. (A2)
The last condition follows analogously.
Appendix B. The Newton Projection Method
Let us now denote by Q(k)(f) the objective function of (11):
Q(k)
(f) = kΣkVT
k f − UT
k sk2
+ αkfk2
(A3)
and by A(f) the set of indices defined as [24]
A(f) =
n
i | 0 ≤ fi ≤ ǫ and (∇Q)i 0
o
, ǫ = min{ǫ, kf − [f − ∇Q]+
k}
where ǫ is a small positive parameter and [·]+ denotes the projection on the positive orthant.
Finally, let E and F denote the diagonal matrices [24] such that
{E(f)}ii =

1, i /
∈ A(f);
0, i ∈ A(f);
F(f) = I − E(f).
The NPCG method for the minimization of Q(k)(f) under nonnegativity constraints is
formally stated in Algorithm A1.
Algorithm A1: NP (Newton-Projection) algorithm
1: choose f(0) and set k = 0
2: for k = 1, 2, . . . do
3: compute the index subset A(k) and the matrices E(k) and F(k);
4: determine the search direction d(k) by solving, with the CG method, the linear system
E(k)
∇2
Q(k)
(f(k)
)E(k)
+ F(k)

d = −∇Q(k)
(f(k)
);
5: determine a step-length α(k) satisfying the Armijo rule along the projection arc [23];
5: compute f(k+1) = [f(k) + α(k)d(k)]+;
6: end for
Appendix C. The Venkataramanan-Song-Hürlimann Algorithm
Let Ki = UiΣiV
T
i be the SVD of Ki, i = 1, 2 and let ki be the number of consid-
ered singular values; for example, ki can be the last index of the singular values than
a given tolerance τ, i = 1, 2. Given the TSVD matrices Kk1,1 = Uk1,1Σk1,1V
T
k1,1 and
Kk2,2 = Uk2,2Σk2,2V
T
k2,2 of K1 and K2, and defined their Kroenecker product Kk1,k2
:
Kk1,k2
= Kk2,2 ⊗ Kk1,1 (A4)

J. Imaging 2021, 7, 18 21 of 23
problem (4) can be approximated as
min
f≥0
kKk1,k2
f − sk2
+ αkfk2
(A5)
or equivalently
min
F≥0
kKk1,1FKT
k2,2 − Sk2
+ αkFk2
. (A6)
The first step of the VSH algorithm consists of projecting the data s onto the column
subspace of Kk1,k2
:
P1,2(s) = Uk1,1 ⊗ Uk2,2

Uk1,1 ⊗ Uk2,2
T
s
that is
P1,2(S) = Uk1,1U
T
k1,1SUk2,2U
T
k2,2.
Hence, by neglecting constant terms, problem (A6) can now be equivalently rewrit-
ten as
min
F≥0
kUk1,1U
T
k1,1Kk1
FKT
k2
Uk2,2U
T
k2,2 − P1,2(S)k2
+ αkFk2
(A7)
which becomes
min
F≥0
k(Σk1,1V
T
k1,1)F(Σk2,2V
T
k2,2)T
− U
T
k1,1SUk2,2k2
+ αkFk2
(A8)
or, equivalently
min
f≥0
ke
Kf − s̃k2
+ αkfk2
(A9)
where
e
K = Σk2,2V
T
k2,2 ⊗ Σk1,1V
T
k1,1 and s̃ = (U
T
k2,2 ⊗ U
T
k1,1)s. (A10)
This results in compressed data s̃ and kernels Σk1,1V
T
k1,1 and Σk2,2V
T
k2,2 of signif-
icantly smaller dimensions, thus avoiding large memory requirements and reducing
computational effort.
The second step of VSH consists of solving the optimization problem (A9) for a
specific value of α. In [4], this is done by using a method adapted from the Butler-Reeds-
Dawson (BRD) algorithm [9] which maps the constrained optimization problem (A9) onto
an unconstrained one. Let us define a vector c implicitly from f by
f = max(0, K̃T
c). (A11)
Hence, problem (A9) can be reformulated as the unconstrained minimization problem
min
c

1
2
cT

G(c) + αI

c − cT
s̃

, (A12)
where
G(c) = e
K






H(e
KT
1 c), 0 . . . 0
0 H(e
KT
2 c) . . . 0
.
.
.
.
.
.
.
.
.
0 0 . . . H(e
KT
k1k2
c)






e
KT
H(x) is the Heaviside function and I is the identity matrix. The Newton method is
then used for solving problem (A12). The Conjugate Gradient (CG) method is employed
for the solution of the inner linear system.
Once (A9) has been solved for a specific α, the BRD method is used for updating the
value of α as
αnew =
√
k1k2
kck
. (A13)

J. Imaging 2021, 7, 18 22 of 23
This procedure is sketched in Algorithm A2.
Algorithm A2: VSH (Venkataramanan–Song–Hürlimann) algorithm
1: choose k1 and k2
2: compute the TSVD of K1 and K2
3: compute the kernel e
K and the projected data s̃ as in (A10)
4: choose α(0)
5: for k = 1, 2, . . . do
6: compute with the Newton method a solution c(j) to the unconstrained problem
min
c

1
2
cT

G(c) + α(j)
I

c − cT
s̃

7: compute f(j) = max(0, K̃Tc(j))
8: update α(j+1) =
√
k1k2

kc(j)k
9: end for
References
1. Ernst, R.; Bodenhausen, G.; Wokaun, A. Principles of Nuclear Magnetic Resonance in One and Two Dimensions, 2nd ed.; International
Series of Monographs on Chemistry; Oxford University Press: Oxford, NY, USA, 1997.
2. Blümich, B. Essential NMR; Springer: Berlin, Germany, 2005.
3. Borgia, G.; Brown, R.; Fantazzini, P. Uniform-Penalty Inversion of Multiexponential Decay Data. J. Magn. Reson. 1998, 132, 65–77.
[CrossRef] [PubMed]
4. Venkataramanan, L.; Song, Y.; Hurlimann, M. Solving Fredholm integrals of the first kind with tensor product structure in 2 and
2.5 dimensions. IEEE Trans. Signal Process. 2002, 50, 1017–1026. [CrossRef]
5. Kroeker, R.; Henkelman, R. Analysis of biological NMR relaxation data with continuous distributions of relaxation times. J.
Magn. Reson. 1986, 69, 218. [CrossRef]
6. Hansen, P. Rank-Deficient and Discrete Ill-Posed Problems; Society for Industrial and Applied Mathematics: Philadelphia, PA, USA,
1998; doi:10.1137/1.9780898719697. [CrossRef]
7. Engl, H.; Hanke, M.; Neubauer, A. Regularization of Inverse Problems; Mathematics and Its Applications; Springer: Dordrecht,
The Netherlands, 2000.
8. Eldèn, L. Algorithms for the regularization of ill-conditioned least squares problems. Behav. Inf. Technol. 1977, 17, 134–145.
[CrossRef]
9. Butler, J.; Reeds, J.; Dawson, S. Estimating solutions of the first kind integral equations with nonnegative constraints and optimal
smoothing. SIAM J. Numer. Anal. 1981, 18, 381. [CrossRef]
10. Fordham, E.; Sezginer, A.; Hall, L. Imaging multiexponential relaxation in the (y, logT1) plane, with application to clay filtration
in rock cores. J. Magn. Reson. Ser. A 1995, 113, 139–150. [CrossRef]
11. Chouzenoux, E.; Moussaoui, S.; Idier, J.; Mariette, F. Primal-Dual Interior Point Optimization for a Regularized Reconstruction of
NMR Relaxation Time Distributions. In Proceedings of the 38th IEEE International Conference on Acoustics, Speech, and Signal
Processing (ICASSP 2013), Vancouver, BC, Canada, 26–31 May 2013; pp. 8747–8750.
12. Cloninger, A.; Czaja, W.; Bai, R.; Basser, P. Solving 2D Fredholm Integral from Incomplete Measurements Using Compressive
Sensing. SIAM J. Imaging Sci. 2014, 7, 1775–1798. [CrossRef]
13. Hansen, P. The discrete Picard condition for discrete ill-posed problems. BIT Numer. Math. 1990, 30, 658–672. [CrossRef]
14. Bortolotti, V.; Brown, R.; Fantazzini, P.; Landi, G.; Zama, F. Uniform Penalty inversion of two-dimensional NMR relaxation data.
Inverse Probl. 2016, 33, 015003. [CrossRef]
15. Bortolotti, V.; Brown, R.; Fantazzini, P.; Landi, G.; Zama, F. I2DUPEN: Improved 2DUPEN algorithm for inversion of two-
dimensional NMR data. Microporous Mesoporous Mater. 2017, 269, 195–198. [CrossRef]
16. Xiang, H.; Zou, J. Regularization with randomized SVD for large-scale discrete inverse problems. Inverse Probl. 2013, 29, 085008.
[CrossRef]
17. Xiang, H.; Zou, J. Randomized algorithms for large-scale inverse problems with general Tikhonov regularizations. Inverse Probl.
2015, 31, 085008. [CrossRef]
18. Bertsekas, D.P. Projected Newton method for optimization with simple constraints. SIAM J. Control Optim. 1982, 20, 221–246.
[CrossRef]
19. Bonettini, S.; Landi, G.; Loli Piccolomini, E.; Zanni, L. Scaling techniques for gradient projection-type methods in astronomical
image deblurring. Int. J. Comput. Math. 2013, 90, 9–29. [CrossRef]

J. Imaging 2021, 7, 18 23 of 23
20. Bortolotti, V.; Landi, G.; Zama, F. 2DNMR data inversion using locally adapted multi-penalty regularization. arXiv 2020,
arXiv:math.NA/2007.01268.
21. Levin, E.; Meltzer, A.Y. Estimation of the Regularization Parameter in Linear Discrete Ill-Posed Problems Using the Picard
Parameter. SIAM J. Sci. Comput. 2017, 39, A2741–A2762. [CrossRef]
22. Landi, G.; Piccolomini, E.L.; Tomba, I. A stopping criterion for iterative regularization methods. Appl. Numer. Math. 2016,
106, 53–68. [CrossRef]
23. Bertsekas, D. Nonlinear Programming, 2nd ed.; Athena Scientific: Nashua, NH, USA, 1999.
24. Vogel, C.R. Computational Methods for Inverse Problems; SIAM: Philadelphia, PA, USA, 2002.

A New Hybrid Inversion Method For 2D Nuclear Magnetic Resonance Combining TSVD And Tikhonov Regularization

More Related Content

Similar to A New Hybrid Inversion Method For 2D Nuclear Magnetic Resonance Combining TSVD And Tikhonov Regularization (20)

More from Pedro Craggett (20)

Recently uploaded (20)

A New Hybrid Inversion Method For 2D Nuclear Magnetic Resonance Combining TSVD And Tikhonov Regularization