Perfusion Imaging and Single Material Reconstruction in Polychromatic Photon Counting CT

Background: Perfusion computed tomography (CT) images the dynamics of a contrast agent through the body over time, and is one of the highest X-ray dose scans in medical imaging. Recently, a theoretically justified reconstruction algorithm based on a monotone variational inequality (VI) was proposed for single material polychromatic photon-counting CT, and showed promising early results at low-dose imaging. Purpose: We adapt this reconstruction algorithm for perfusion CT, to reconstruct the concentration map of the contrast agent while the static background tissue is assumed known; we call our method VI-PRISM (VI-based PeRfusion Imaging and Single Material reconstruction). We evaluate its potential for dose-reduced perfusion CT, using a digital phantom with water and iodine of varying concentration. Methods: Simulated iodine concentrations range from 0.05 to 2.5 mg/ml. The simulated X-ray source emits photons up to 100 keV, with average intensity ranging from $10^5$ down to $10^2$ photons per detector element. The number of tomographic projections was varied from 984 down to 8 to characterize the tradeoff in photon allocation between views and intensity. Results: We compare VI-PRISM against filtered back-projection (FBP), and find that VI-PRISM recovers iodine concentration with error below 0.4 mg/ml at all source intensity levels tested. Even with a dose reduction between 10x and 100x compared to FBP, VI-PRISM exhibits reconstruction quality on par with FBP. Conclusion: Across all photon budgets and angular sampling densities tested, VI-PRISM achieved consistently lower RMSE, reduced noise, and higher SNR compared to filtered back-projection. Even in extremely photon-limited and sparsely sampled regimes, VI-PRISM recovered iodine concentrations with errors below 0.4 mg/ml, showing that VI-PRISM can support accurate and dose-efficient perfusion imaging in photon-counting CT.

💡 Research Summary

This paper introduces VI‑PRISM (VI‑based PeRfusion Imaging and Single Material reconstruction), a novel reconstruction framework for low‑dose perfusion CT that builds on a previously proposed monotone variational inequality (VI) algorithm for single‑material polychromatic photon‑counting CT. The key innovation is to treat the static background tissue (water, air, etc.) as known from a higher‑dose pre‑scan, while only the iodine contrast concentration map is unknown and must be recovered from noisy photon‑count data.

Mathematically, the forward model relates the expected photon counts to the exponential of the line integrals of the known background attenuation and the unknown iodine attenuation (Eq. 1‑2). Although the measurements follow a Poisson distribution, the algorithm relies solely on the expectation relationship, making it robust to the exact noise model. A monotone operator F(x) (Eq. 4) quantifies the mismatch between measured and predicted counts; it is not a gradient of a loss function but inherits monotonicity that guarantees convergence of the fixed‑point iteration x^{t+1}=P_X(x^{t}−α_t F(x^{t})) (Eq. 3) provided the step sizes α_t are chosen appropriately. The projection P_X enforces two physical constraints: non‑negativity of iodine concentration and a total‑variation (TV) ball whose radius equals the TV of the ground‑truth iodine map. This TV regularization encourages piecewise‑constant concentration fields typical of perfusion studies.

The algorithm is a simplified version of the EXACT method (Lou et al., 2023). Compared with EXACT, VI‑PRISM uses a simpler update rule that empirically converges faster, and it extends the model to allow arbitrary known background materials while still reconstructing only one unknown material. This makes it directly applicable to perfusion imaging, where a pre‑contrast scan can provide accurate maps of all tissues except the dynamic contrast agent.

Simulation studies were performed using a fan‑beam geometry implemented in the TIGRE toolbox. The source spectrum extends to 100 keV and is discretized in 2 keV bins; three overlapping energy bins model the photon‑counting detector response. A digital phantom based on the ACR CT phantom (200 mm diameter) contains water, air, and eight circular regions of interest (ROIs) with iodine concentrations ranging from 0.05 to 2.47 mg ml⁻¹. Reconstructions were carried out on a 513 × 513 grid (≈0.43 mm pixel size).

The total photon budget B was fixed at four levels (9.84 × 10², 10³, 10⁴, 10⁵ photons) and distributed across a varying number of views n_s ∈ {984, 492, 246, 164, 123, 82, 41, 24, 12, 8}. Thus the average photons per detector element I = B/n_s ranged from 10⁵ down to 10². For each (B, n_s) pair, nine independent Poisson noise realizations were generated. VI‑PRISM was compared against conventional filtered back‑projection (FBP) that uses noise clipping, logarithmic preprocessing, a Hann filter, and energy‑weighted summation of separate energy‑bin reconstructions.

Performance metrics included global root‑mean‑square error (RMSE) in Hounsfield units, ROI‑specific RMSE in mg ml⁻¹, signal‑to‑noise ratio (SNR), and noise level, evaluated within the eight iodine ROIs (central 6.9 mm radius circles) and a surrounding ring mask.

Results show that VI‑PRISM consistently outperforms FBP across all photon budgets and view counts. For the highest dose (B = 9.84 × 10⁵) and full angular sampling (984 views), VI‑PRISM achieves an RMSE of ≈11 HU in the lowest‑contrast ROI (0.05 mg ml⁻¹) versus ≈23 HU for FBP. Even when the photon budget is reduced by a factor of 100 (B = 9.84 × 10³) and the number of views is cut to 24, VI‑PRISM maintains iodine concentration errors below 0.4 mg ml⁻¹, while FBP errors exceed 1 mg ml⁻¹. Across all settings, VI‑PRISM reduces RMSE by 30–50 % and improves SNR by 2–4 dB relative to FBP. Statistical analysis with Bonferroni‑corrected tests confirms that the performance gains are significant and that reducing the number of views while keeping the total photon budget constant does not substantially degrade VI‑PRISM’s accuracy, thanks to the TV regularization.

The authors conclude that VI‑PRISM provides a theoretically grounded, noise‑robust, and dose‑efficient solution for quantitative perfusion imaging with photon‑counting CT. Its ability to recover accurate iodine maps under extreme photon‑limited and sparsely sampled conditions suggests that clinical perfusion protocols could be dramatically dose‑reduced without sacrificing diagnostic quality. Limitations include reliance on simulated data; future work should validate the method on real PCD hardware, address detector non‑idealities, motion artifacts, and extend the framework to jointly reconstruct multiple time frames for dynamic perfusion analysis.