arXiv subjects

Sabee Molloi

Publications and source records attributed to Sabee Molloi.

3 recordsLinked to original sources

arXiv2026

Unified Projection-Domain Material Decomposition for Energy-Integrating and Photon-Counting CT

Objective: To develop a unified and efficient method for projection-domain material decomposition in spectral computed tomography (CT) that incorporates an arbitrary number of spectral measurements, independent of the acquisition mechanism, without reducing the measurements to two effective energies. Approach: A nested optimization method was developed for two-material basis decomposition from K>=2 spectral measurements. It extends an existing two-measurement univariate method by replacing its polynomial approximation with profile minimization of the Poisson negative log-likelihood, reducing the inversion to two one-dimensional searches that use all measurements without spectral rebinning. The method was evaluated in simulation for energy-integrating (GE Revolution Apex Elite, Siemens SOMATOM Flash) and photon-counting (Siemens NAEOTOM Alpha) CT. Main results: The method recovered line integrals with noiseless iodine errors below 3e-6 g/cm^2 and noise within 3% of the Cramer-Rao lower bound (CRLB). Virtual monoenergetic images (VMIs) agreed, at 70 keV and above, with reference CT numbers and, on all three systems, with vendor VMIs of physical scans (mean absolute error 6-22 HU). For photon counting, four measurements instead of two reduced the iodine CRLB by 19% and image noise by 8-16% at every VMI energy, and lowered the regression error at 70 keV and above. Iteration and tolerance settings changed CT numbers by less than 0.04 HU, and decomposition took 0.34-0.67 s per million rays on an A100 GPU. Significance: The method provides fast, accurate decomposition for any spectral CT acquisition mechanism and number of measurements, so one method can serve every scanner in a virtual imaging trial.

eess.IV↗

arXiv2026

An anatomically grounded human-scale steady hydraulic model of the lymphatic network

The lymphatic system returns 2 to 3 L of interstitial fluid per day to the venous circulation and drains the central nervous system. Most models of lymph transport describe one lymphangion or a short chain, so they cannot show how body-wide topology, calibre hierarchy and the venous outlet set pressure and flow. We build a passive steady hydraulic model on an anatomically derived graph of the human lymphatic system, with 1010 vertices and 1126 edges expanded to 89,060 sub-edges. Radii follow a Murray cube law, derived both from minimum work and from wall-shear-stress homeostasis. Pressure and flow follow from nodal hydraulic balance over Hagen-Poiseuille conductances, with an exponential wall law, hysteretic valves, Starling leak, node retention and a Windkessel outlet. Active contraction is off, and only the inflow scale is calibrated. Two quantities that are not fitted land in clinical range at the same operating point. Radii fall inside the vessel-class range, from 1.50 mm in the thoracic duct to 0.16 mm in the cervical collectors, and pressures stay below the 50 mmHg cuff scale on all sub-edges. About 49% of the gross load is removed before the outlet, nearly independent of the outflow scale and of the outlet parameters. On the route that drains the central nervous system, 30% of the superficial cervical load reaches the vein, at 10 to 18 mmHg along a segment that cervical-scale models replace with a prescribed outlet pressure. The model is a passive whole-body baseline for contraction-driven and clinical extensions.

physics.med-ph↗

arXiv2026

An anatomically grounded human-scale dynamic model of contraction-driven lymphatic pumping

Lymphangions, the valved muscular segments of lymphatic vessels, pump lymph against an adverse pressure gradient. Current models resolve this pumping in one vessel, a short chain or a regional network, and whole-body human models are passive. We extend the passive whole-body network of a companion static study to the active regime on the same anatomical graph, with intrinsic contraction, dynamic valves and a respiration-gated venous outlet on every vessel. Without fitting the waveform or the contraction amplitude, the network settles into a bounded pumping cycle at the lymphovenous junction, at a clinical pressure of 7.5 to 12.3 mmHg and a drainage of 35 to 50 mm^3/s. Respiration adds a swing of 1 to 2.5 mmHg at the junction and contraction about 1 mmHg inside the duct. Pressure generation is limited more by load than by drive, and no single contraction amplitude fits every vessel. Raising the subclavian venous pressure from 3 to 15 mmHg lifts the deep cervical node from 8.6 to 21.1 mmHg at the weakest drive, a transmission of 1.04 that falls to 0.78 at the strongest, so contraction buffers cervical drainage against venous pressure. The same solver gives the human cervical outlet pressure that cerebrospinal-fluid drainage models prescribe.

physics.med-ph↗