REVIEW 3 major objections 4 minor 25 references
Approximate peak time to time-domain fluorescence diffuse optical tomography for nonzero fluorescence lifetime
T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash
Pith's one-line read A single explicit equation approximates the peak time of time-domain fluorescence diffuse optical tomography, reducing target localization to a one-dimensional root-finding problem.
desk verdict Single-target peak-time equation for nonzero fluorescence lifetime is a solid new result; the multi-target boundary-scan is heuristic and needs more support, but the paper deserves serious review. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the approximate peak time equation (3.3), $P(t;\lambda)=\lambda e^{-(\sqrt{kt}-\lambda)^2/t}-\pi^{1/2}\ell^{-1}t^{3/2}=0$, whose unique positive root is the approximate peak time. Here $\lambda:=\big((|x_d-x_c|^2+|x_s-x_c|^2)/(2vD)\big)^{1/2}$ summarizes the target's location relative to each source-detector pair, and all uniqueness, order, and symmetry results are statements about how the root depends on this single number. The equation is obtained from the asymptotic expansion of Theorem 2.3, $\int_0^t u^a_m(s)\,ds \sim k^{-3/4}(\pi\lambda)^{1/2}u^a_m(\lambda k^{-1/2})$ for $\lambda\gg 1$, which converts the convolution integral in the fluorescence response into a point evaluation at the saddle time $s=\lambda k^{-1/2}$, with $u^a_m$ the large-depth profile of the zero-lifetime solution. The same machinery carries to multiple targets by selecting, for each source-detector pair, the target that dominates the squared-distance sum.
What would settle it
Compute the true peak time by numerical integration of (2.1) for two targets placed so that, for the central source-detector pair, $|x_d-x_c^{(1)}|^2+|x_s-x_c^{(1)}|^2 = |x_d-x_c^{(2)}|^2+|x_s-x_c^{(2)}|^2$ (violating (2.18)), and compare the true peak time with the roots of (3.8) for each target; if neither root approximates the true peak time, or if Algorithm 3 returns a spurious local minimum between the two targets, the approximate equation's dominance assumption is the failing link.
Extended reading notes
Core claim
The paper's central claim is that for a single fluorescent point target in a highly scattering half-space with Robin boundary condition and nonzero fluorescence lifetime, the peak time $t_{\mathrm{peak}}$ solves, to leading asymptotic order, $P(t;\lambda)=0$ with $P(t;\lambda):=\lambda e^{-(\sqrt{kt}-\lambda)^2/t}-\pi^{1/2}\ell^{-1}t^{3/2}$, where $k=\mu_av$ and $\lambda^2=(|x_d-x_c|^2+|x_s-x_c|^2)/(2vD)$. For multiple well-separated targets, the same equation holds with $\lambda$ computed from whichever target minimizes $|x_d-x_c^{(j)}|^2+|x_s-x_c^{(j)}|^2$ for that source-detector pair. The authors prove uniqueness of the root under a lower bound on $\ell$, prove that the approximate peak time is monotonically ordered by $\lambda$, and confirm numerically that the root tracks the true peak time to within a few percent across practical values of absorption, diffusion, lifetime, and depth. On the strength of these properties they assert that peak-time localization reduces to locating the minimal peak time over a boundary scan (horizontal coordinates) and then solving the one-dimensional equation for depth.
Load-bearing premise
For multiple targets, the method assumes that from every source-detector pair one target is strictly closer in the squared-distance sense $|x_d-x_c^{(j)}|^2+|x_s-x_c^{(j)}|^2$ than all others, and that the measured peak-time landscape has exactly one local minimum per target; if two targets are close enough to violate this, the approximate equation and the scanning search can misidentify or miss targets.
Editorial extensions
If this is right
- The explicit equation turns depth reconstruction into a one-dimensional root-finding problem once horizontal coordinates are known, eliminating iterative forward solves and regularization.
- For a single target, the bisection algorithm converges to the true location, with numerical reconstructions accurate to a few percent relative error even when the measured peak time is perturbed by up to 5% noise.
- For well-separated multiple targets, the boundary-scan algorithm recovers both horizontal coordinates from the local minima of the peak-time landscape and then solves each depth independently.
- The paper's numerics show the approximation error decreases with increasing fluorescence lifetime and with smaller diffusion constants, so the formula is most trustworthy for nanosecond-lifetime fluorophores in tissues with smaller diffusion coefficient.
Reading between the lines
- A consequence the authors leave implicit is that any monotone function of $\lambda$, not just the peak time, could drive the same two-stage reconstruction; the bisection algorithm is essentially minimizing $\lambda$ over the boundary, so alternative statistics of the temporal response that preserve that ordering would work.
- The separation condition behind the multiple-target result is checkable from data only indirectly; a practical safeguard would be to verify near each detected local minimum that the measured peak time satisfies the single-target equation (3.8) at the reconstructed depth, flagging pairs whose residual exceeds the noise level as unresolved.
- The saddle-point asymptotic of Theorem 2.3 is generic, so the same approximate equation should extend to other boundary conditions ($\beta=0$, $\beta=\infty$) and, as the authors note, to curved measurement surfaces via parabolic-scaling Green functions; testing the formula on those geometries is a natural next step.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. This paper studies an inverse problem for time-domain fluorescence diffuse optical tomography (FDOT) in a half-space with point fluorescent targets having nonzero fluorescence lifetime. The authors start from the coupled diffusion model (1.1)-(1.2), replace the emission solution by an asymptotic expansion in the fluorescence lifetime (2.2)-(2.3), and then use a Laplace-type asymptotic argument to derive the approximate peak time equation (3.3), whose root defines the approximate peak time. Uniqueness of this root is proved in Theorem 3.1. The equation is extended to multiple targets under a nearest-target dominance condition (2.18), and two reconstruction algorithms are proposed: a bisection algorithm for a single target and a boundary-scan algorithm for multiple well-separated targets. Numerical experiments in Section 5 test the approximation and the algorithms under noiseless and noisy data.
Significance. If the results are valid, equation (3.3) is a genuinely useful reduction: peak-time localization becomes one-dimensional root-finding once the horizontal target coordinates are known, and the derivation is parameter-free with no constants fitted to data. The paper's strengths include the explicit derivation from (2.2) through Theorem 2.3, the uniqueness theorem for the approximate peak time, the reproducible numerical verification over a range of optical parameters, and the demonstrated robustness of the single-target bisection algorithm. However, the multi-target reconstruction claim depends on an unproved equivalence between local minima of the peak-time landscape and true target positions, and the dominance condition underlying the multi-target asymptotic formula is insufficient at the peak-time scale. The single-target contribution is sound and publishable, but the multi-target claim currently needs substantial additional support.
major comments (3)
- [Corollary 2.4, Eqs. (2.19)-(2.20) and (3.8)] The dominance claim in Corollary 2.4 is not justified at the time scale used in the rest of the paper. Condition (2.18) only asserts that one target has the smallest value of |x_d-x_c^(j)|^2+|x_s-x_c^(j)|^2. But Theorem 2.3 evaluates the time integral at the saddle s = λ^(l) k^{-1/2}, so the relevant comparison time in equation (3.3) is t ≈ λ^(l)/√k. The exponential ratio controlling the contribution of a competing target j is then exp(-((λ^(j))^2-(λ^(l))^2)/t) = exp(-√k ((λ^(j))^2-(λ^(l))^2)/λ^(l)). This quantity is not necessarily o(1): if the target separation in depth is fixed while λ^(l) grows, the exponent tends to 2√k Δλ, a constant, so the competing target contributes at O(1) rather than o(1). Therefore (2.19), (2.20), and the multi-target approximate peak time equation (3.8) require a stronger separation condition, such as (λ^(j))^2-(λ^(l))^2 ≫ λ^(l), which is neither stated nor verified in Example 5.3.
- [Definition 4.4 and Algorithm 3] Definition 4.4 defines well-separated targets by assuming that the peak-time landscape P has J local minima. This is exactly the conclusion that Algorithm 3 needs in order to reconstruct J targets, not a proven premise. No theorem establishes that the number of local minima of the measured t_peak equals J, that each local minimum is attained at an S-D pair satisfying (4.17), or that the discrete scan grid Ξ contains such a pair for every target. In Example 5.3, the reconstructed two-dimensional positions are (3.00,5.00) versus the true (3.3,5.2) and (17.00,17.00) versus (17.4,16.7), showing grid quantization; at 1% noise the second target shifts to (18.00,17.00). The ad hoc 3×3 moving-average smoothing used in Example 5.3 to remove spurious local minima indicates that the peak-time landscape does not robustly encode the target count. This is the central unsupported step for the multi-target reconstruction claim.
- [Section 3.1 and Figure 3.1(d)] The statement that the approximate peak time has excellent accuracy under practical parameters is not supported for all parameter values shown. In Figure 3.1(d), with default parameters (3.7), the relative error exceeds 10% at fluorescence lifetime ℓ = 500 ps, and the error decreases only gradually as ℓ increases, consistent with the ℓ ≫ 1 assumption in expansion (2.2). The paper does not quantify the joint validity region in (ℓ, λ, x_c3) and gives no error bound for equation (3.3). Since Stage 2 of the single-target reconstruction solves (3.4) with the measured t_peak as input, a quantitative statement about the approximation error is needed to assess the bias that the algorithm can introduce.
minor comments (4)
- [Proof of Theorem 4.2, text near Eq. (4.13)] The monotonicity sentence after Eq. (4.13) has the signs reversed: for fixed λ^(1), the left-hand side t ln(λ1/λ2) decreases as λ2 increases, while the right-hand side increases. The uniqueness conclusion is correct and can be proved from this one-signed behavior, but the proof as written should be corrected.
- [Eq. (4.10)] Eq. (4.10) contains an apparent typo: the threefold equality ta,(2)_peak = ta,(2)_peak = ta,(2)_peak should presumably define the two quantities in (4.11), which are the supremum-based and infimum-based characterizations.
- [Algorithm 1, Step 3] The notation 'ROI := (xb, x_t)' and 'ROI := (x_b, x_t)' is inconsistent and ambiguous; the one-dimensional interval passed to Algorithm 2 should be written consistently, e.g. (x_b, x_t) or (x_l, x_r).
- [Eq. (5.1)] The perturbation in Eq. (5.1) is described as time jitter, but it is a relative multiplicative perturbation of t_peak by a uniform random variable. It is not a jitter in the measured temporal response function. Please clarify the noise model.
Circularity Check
The multi-target boundary-scan reconstruction is circular by definition: 'well-separated' is defined as having J peak-time local minima at the S-D pairs closest to the true targets, which is exactly what Algorithm 3 uses to reconstruct them.
-
self definitional
[Section 4.3, Definition 4.4 (near eq. (4.19)); used in Algorithm 3, Step 2 (eq. (4.20))]
"We say that these J unknown point targets are well-separated if there are J local minimums in P for J multiple point targets tmin,(j)_peak := tmin_peak(xmin,(j)_d, xmin,(j)_s; x(1)_c,···, x(J)_c), j = 1, 2,···, J, where the S-D pair {xmin,(j)_d, xmin,(j)_s}∈Ξ is the S-D pair of the smallest distance from the j-th target in Ξ."
Algorithm 3 Step 2 reconstructs the horizontal coordinates of target j as (xmin,(j)_s1+xmin,(j)_d1)/2 and (xmin,(j)_s2+xmin,(j)_d2)/2, i.e., it uses the grid S-D pair at which the measured peak-time landscape has a local minimum. The correctness of this step requires exactly that each local minimum corresponds to a true target and that its S-D pair is the one minimizing distance to that target. Definition 4.4 does not derive this correspondence from the forward model (2.1) or the approximate peak time equation (3.8); it posits it as the defining property of 'well-separated' targets.
full rationale
The single-target chain is not circular: equation (3.3) is obtained by inserting the explicit zero-lifetime asymptotic profile ua_m (Lemma 2.1, quoted from the authors' published [4]) into the peak condition ∂t Ua_m = 0 and simplifying via Theorem 2.3; no constant is fitted to data, and the cited [4] result is a parameter-free asymptotic theorem with stated assumptions, not an assumption of the target conclusion. The bisection algorithm solves the resulting equation rather than assuming the answer. The circularity is confined to the multi-target boundary-scan claim. Corollary 2.4 only justifies the per-target approximate peak time (3.8) for S-D pairs satisfying the dominance condition (2.18), and Definition 4.4 then defines 'well-separated' targets as those for which the measured peak-time landscape has J local minima at the grid S-D pairs nearest to the true targets. Since Algorithm 3 Step 2 uses those local-minimum S-D pairs directly as the reconstruction, the key premise of the multi-target algorithm is identical to its definition of the problem class. The paper also concedes that Algorithm 3 requires 3x3 moving-average smoothing to remove spurious local minima, further indicating that the local-minimum/target correspondence is an auxiliary assumption rather than a derived fact. These issues lower the score to 6 (partial circularity) while the single-target and forward-model derivations remain independent.
Assumptions & free parameters
free parameters (3)
- 3x3 moving-average smoothing window =
3x3 grid
- S-D pair separation L =
8 mm (single target), 2 mm (multiple target)
- Bisection tolerances epsilon1, epsilon2 =
0.1 mm or 1.25 mm
assumptions (8)
- domain assumption Diffusion approximation for light transport in tissue, equations (1.1)-(1.2)
- domain assumption Half-space domain R3_+ and planar boundary
- domain assumption Point-target fluorescence distribution mu(x)=sum c_j delta(x-x_c^j)
- domain assumption Single-exponential fluorescence decay f_ell(t)=ell^{-1}e^{-t/ell}
- ad hoc to paper Large fluorescence lifetime expansion, ell much greater than 1 in (2.2)
- domain assumption Large-depth and large-lambda asymptotics from [4] (Lemma 2.1) and Theorem 2.3
- ad hoc to paper Nearest-target dominance condition (2.18)
- ad hoc to paper Well-separated targets with J distinct peak-time local minima
Cite this review
Pith. "Pith review of Approximate peak time to time-domain fluorescence diffuse optical tomography for nonzero fluorescence lifetime." pith.science (2026). https://pith.science/paper/K376PVYX
@misc{pith2026241115698,
author = {Pith},
title = {Pith review of: Approximate peak time to time-domain fluorescence diffuse optical tomography for nonzero fluorescence lifetime},
year = {2026},
howpublished = {\url{https://pith.science/paper/K376PVYX}},
note = {Machine review of arXiv:2411.15698}
}
read the original abstract
This paper concerns an inverse problem for fluorescence diffuse optical tomography (FDOT) reconstructing locations of multiple point targets from the measured temporal response functions. The targets are multiple fluorescent point objects with a nonzero fluorescence lifetime at unknown locations. Peak time, when the temporal response function of the fluorescence reaches its maximum, is a robust parameter of the temporal response function in FDOT because it is most less suffered by the artifacts, such as noise, and is easily determined by experiments. We derive an approximate peak time equation based on asymptotic analysis in an explicit way in the case of nonzero fluorescence lifetime when there are single and multiple point targets. The performance of the approximation is numerically verified. Then, we develop a bisection algorithm to reconstruct the location of a single point target from the algorithm proposed in [4] for the case of zero fluorescence lifetime. Moreover, we propose a boundary-scan algorithm for the reconstruction of locations of multiple point targets. Finally, several numerical experiments are implemented to show the efficiency and robustness of the addressed algorithms.
Figures
Figures from the paper (2 more)
Reference graph
Works this paper leans on
- [1]
-
[2]
A. N. Bashkatov, E. A. Genina, V. V. Tuchin, Optical properties of skin, subcutaneous, and muscle tissues: a review, J. Innov. Opt. Heal. Sci. , 4 (2011) 9–38
work page 2011
- [3]
-
[4]
S. Chen, J. Eom, G. Nakamura, G. Nishimura, Approximate peak t ime and its application to time- domain fluorescence diffuse optical tomography, Commun. Anal. Comput. , 1 (2023) 379–406
2023
-
[5]
J. Eom, M. Machida, G. Nakamura, G. Nishimura, C. Sun, Express ions of the peak time for time-domain boundary measurements of diffuse light, J. Math. Phys. , 64 (2023) 083504
2023
-
[6]
D. Hall, G. Ma, F. Lesage, Y. Wang, Simple time–domain optical meth od for estimating the depth and concentration of a fluorescent inclusion in a turbid medium, Opt. Lett. , 29 (2004) 2258–2260
work page 2004
-
[7]
S. Han, S. Farshchi-Heydari, D. Hall, Analytical method for the f ast time–domain reconstruction of fluorescent inclusions in vitro and in vivo, Biophys. J. , 98 (2010) 350–357
work page 2010
-
[8]
L. Herv´ e, A. Koenig, J. M. Dinten, Non-uniqueness in fluoresce nce-enhanced continuous wave diffuse optical tomography, J. Opt. , 13 (2011) 015702
work page 2011
Show all 25 references
-
[9]
Hoshi, Y
Y. Hoshi, Y. Yamada, Overview of diffuse optical tomography and its clinical applications, J. Biomed. Opt., 9 (2016) 091312
2016
-
[10]
Hebden, R
J. Hebden, R. Kruger, K. Wong, Time resolved imaging through a highly scattering medium, Appl. Opt., 30 (1991) 788–794
1991
-
[11]
Hebden, S
J. Hebden, S. Arridge, D. Deply, Optical imaging in medicine: I. Ex perimental technique, Phys. Med. Biol., 42 (1997) 825–840
1997
-
[12]
Jiang, Diffuse optical tomography: principles and applications , CRC Press, Boca Raton, 2010
H. Jiang, Diffuse optical tomography: principles and applications , CRC Press, Boca Raton, 2010
2010
-
[13]
Jiang, Fluorescence Molecular Tomography: Principles and Applic ations, Springer, Cham, Switzer- land, 2022
H. Jiang, Fluorescence Molecular Tomography: Principles and Applic ations, Springer, Cham, Switzer- land, 2022
2022
-
[14]
J. R. Lakowicz, Principles of Fluorescence Spectroscopy , Second Edition, Kluwer Academic/Plenum Publishers, New York, 1999
1999
-
[15]
S. Lam, F. Lesage, X. Intes, Time domain fluorescent diffuse op tical tomography: analytical expressions, Opt. Exp. , 13 (2005) 2263–2275
2005
-
[16]
J. Liu, M. Machida, G. Nakamura, G. Nishimura, C. Sun, On fluore scence imaging: the diffusion equation model and recovery of the absorption coefficient of fluorophores , Sci. China Math. , 65 (2022) 1179–1198
2022
-
[17]
Y. Liu, W. Ren, H. Ammari, Robust reconstruction of fluoresce nce molecular tomography with an optimized illumination pattern, Inverse Probl. Imag. 14 (2020) 535–568
2020
-
[18]
Milstein, S
A. Milstein, S. Oh, K. Webb, C. Bouman, Q. Zhang, D. Boas, R. Milla ne, Fluorescence optical diffusion tomography, Appl. Opt. , 42 (2003) 3081–3094
2003
-
[19]
Mycek, B
M. Mycek, B. Pogue, Handbook of Biomedical Fluorescence , Marcel Dekker, New York, 2003
2003
-
[20]
Nitziachristos, C
V. Nitziachristos, C. Tung, C. Bremer, R. Weissleder, Fluoresc ence molecular tomography resolves protease activity in vivo, Nat. Med., 8 (2002) 757–761
2002
-
[21]
Nakamura, H
G. Nakamura, H. Wang, Solvability of interior transmission proble m for the diffusion equation by con- struting its Green function, J. Inverse Ill-Pose. Probl. , 27 (2019) 671–701
2019
-
[22]
Papadimitriou, E
K. Papadimitriou, E. Rosas, E. Zhand, R. Cooper, J. Hebden, S . Arridge, S. Powell, Dual wavelength spread-spectrum time-resolved diffuse optical instrument for th e measurement of human brain functional responses, Biomed. Opt. Express , 11 (2020) 3477–3490. 22
2020
-
[23]
Prieto, G
K. Prieto, G. Nishimura, A new scheme of the time-domain fluores cence tomography for a semi-infinite turbid medium, Opt. Rev. , 24 (2017) 242–251
2017
-
[24]
Nishimura, T
G. Nishimura, T. Suzuki, Y. Yamada, H. Niwa, T. Koike, Depth det ection limit of a fluorescent object in tissue-like medium with background emission in continuous-wave mea surements: a phantom study, J. Biomed. Opt. , 29 (2024) 097001
2024
-
[25]
Taroni, A
P. Taroni, A. Pifferi, A. Torricelli, D. Comelli, R. Cubeddu, In vivo ab sorption and scattering spec- troscopy of biological tissues, Photoch. Photobio. Sci ., 2 (2023), 124–129. 23 0 500 1000 1500 2000 2500 -5 0 5 10 10-28 Um Um a,(1) Um a,(2) Um a,(3) Um a,(4) Um a,(6) 0 500...
2023
Reviewed August 12, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.