A preprint reports a three-dimensional imaging method that maps elemental fluorescence across an entire specimen while acquiring structural phase-contrast tomography in the same geometry. Called full-field fluorescence computed tomography, or F3CT, the approach uses a calibrated virtual cone-beam model and pinhole-encoded detection to turn those measurements into a volume. It was demonstrated on a silver-stained zebrafish trunk and a granite-serpentinite core stack, making this a technical proof of concept rather than a broad performance test.
How the system captured the signal
The samples were deliberately different. One was a silver-stained adult zebrafish trunk section; the other combined granite and serpentinite in a core stack. That let the authors test chemically distinct fluorescence signals in a biological sample and structurally different materials in a geological one. The phase-contrast data supplied the structural reference for the fluorescence maps.
Full-field means the beam illuminated each sample in its entirety instead of moving a small excitation spot across it. The experiment used a parallel polychromatic beam and a HEXITEC hyperspectral detector set at 90 degrees to the beam. The detector's pixels were 250 micrometres across, and a 200-micrometre central pinhole sat in a 5-millimetre-thick, 20-millimetre-diameter aperture disc.
Calibration before reconstruction
Getting the geometry right was central to the reconstruction. The team first made an approximate pre-calibration, then refined the acquisition parameters with a self-consistent, data-driven fit. The initial calibration estimated detector rotation from the horizontal and vertical differences between the centres of two calibration ellipses. The refinement swept through candidate values, kept the one with the lowest loss, and applied a local parabolic fit using the two adjacent tested values.
The detector spectra were calibrated separately for each pixel with a linear model based on americium-241 lines. The data were represented in 769 energy bands from 4 to 100 keV, spaced by 0.125 keV. Six leading MNF components were retained, and 20-channel spectral windows were integrated to build the reconstructed volumes.
Fluorescence was collected at 121 angular steps around 360 degrees. Each zebrafish projection was integrated for 150 seconds and each granite-serpentinite projection for 170 seconds, while the accompanying phase-contrast scan used 5,000 projections. The reported fluorescence resolution was about 150 micrometres, with enough signal-to-noise to resolve elemental distributions, although higher-energy signals dominated.
What the two specimens showed
In the zebrafish, the detector resolved silver K-alpha and K-beta lines at 22.16 and 24.94 keV. The three-dimensional reconstruction showed a spatially varying silver-stain distribution through the soft tissue. That makes the result a map of where the stain appeared, not a measurement of silver concentration: the study did not report a quantitative concentration or detection limit.
The geological data produced a different kind of result. Using 20-channel energy windows, the analysis reconstructed iron-sensitive and barium-sensitive fluorescence contributions. Their spatial separation distinguished granite from serpentinite regions. The paper notes that the MNF components were not pure elemental maps, and the strongest reported signals were at higher energies.
Together, the two modalities gave the specimen both structural morphology and chemical contrast. In the core stack, the combined PCT-F3CT data included barium- and iron-sensitive features and detected weaker fluorescence in serpentinite. The pairing added discrimination in regions that would be difficult to identify from morphology alone.
A useful test case, not a benchmark
The limits are as important as the images. This was a two-specimen methods demonstration, with no independent benchmark or quantitative accuracy estimate. The analysis also reported no silver concentration, detection limit, or registration-error metric. It leaves unresolved how throughput, spatial resolution, sensitivity, dose and registration accuracy compare with focused pencil-beam XRF tomography.
The setup itself imposes trade-offs. The single-pinhole geometry exchanges throughput for resolution, and the current detector generation has comparatively slow temporal response. The authors describe F3CT as a complementary approach that favors full-field acquisition and compatibility with phase-contrast imaging. The experiment did not demonstrate real-time, in-situ or operando imaging.
Paper data and sources
Original title: Full-field fluorescence computed tomography (F3CT) using a calibrated virtual cone-beam pinhole geometry
Authors: Thomas Zillhardt, Yunhui Chen, Alexander Rack et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-28
DOI: Not available
Original paper · Full text