An X-ray micro-computed tomography, or micro-CT, workflow has given archived prostate tissue a three-dimensional view of its glandular architecture without staining or additional processing. In the proof-of-concept study, intact paraffin cores were scanned and then processed for histology, the preparation of thin tissue sections, allowing the same specimens to be compared as 3D volumes and conventional slides.
The authors call the work a proof-of-concept study. It involved five prostates, with one 2-mm-diameter paraffin core from each. The sampled regions included Gleason patterns 3, 4 and 5, plus two benign regions, so the material covered several tissue types but remained a small set.
A long scan, tuned for clarity
The cores were scanned intact with a 10× objective and 2× detector binning. The resulting images used 2.47 μm isotropic voxels, meaning the volume was sampled at the same scale in all three directions. Each scan covered a full 360 degrees, with 1,600 projections and one frame per projection.
The researchers tuned the tube voltage first, using contrast-to-noise ratio, a measure of how clearly tissue stands out from image noise. The pilot core scored 8.76 at 50 kV, compared with 8.41 at 40 kV, 8.21 at 60 kV and 6.97 at 80 kV. They selected 50 kV for the later scans.
Exposure time then became a practical trade-off. At 1,600 projections, 25 seconds per projection preserved discernible glandular morphology and took approximately 11 hours. The same setup would have taken about 22 hours at 50 seconds and 33 hours at 75 seconds, so 25 seconds was selected.
Noise reduction improved the short scan's similarity to a 75-second reference volume. In the pilot data, the unfiltered 25-second scan had SSIM 0.829, NRMSE 0.112 and PSNR 24.96 dB. With non-local-means filtering, the scores were 0.936, 0.109 and 25.21 dB, respectively. All tested methods improved the numerical comparison, while differences among the strongest methods were small.
Anisotropic diffusion was ultimately applied to all datasets after the numerical scores were considered alongside qualitative image review. The combined assessment led the researchers to use it for the rest of the study.
What the images showed
Across all five cores, matched micro-CT sections and histology showed qualitative agreement in tissue contours, gland-rich and stromal regions, epithelial contours, lumina, concretions and fibromuscular stroma. Here, qualitative agreement means the structures lined up visually; it was not a quantified diagnostic-performance measure.
The match was not perfect: localized discrepancies were consistent with processing artifacts. That is why the authors describe the correspondence as qualitative rather than as a precise diagnostic measurement.
In one illustrative exercise, the team manually segmented a region from sample 1, which represented predominant Gleason pattern 3 tissue. The rendering showed glandular epithelium and luminal spaces continuing through the depth of the tissue. The exercise demonstrated a possible 3D view, but it was a single-region technical demonstration without quantitative segmentation accuracy.
Shape change remains exploratory
The researchers also compared dimensions before sectioning with dimensions measured in matched histology. At the whole-core level, dimensions rose by 5.23% on average, with a standard deviation of 3.01 percentage points, perpendicular to the slicing direction, and fell by 1.73%, with a standard deviation of 6.57 percentage points, parallel to it. The reported directional difference was -6.95, with a standard deviation of 5.49. The three reported p values were 0.0625, 0.6250 and 0.1250.
None of those tests reached conventional statistical significance. Each of the five cores was treated as the analysis unit, so the deformation estimates were exploratory rather than definitive.
At the glandular-lumen level, the average change was 5.54% perpendicular to slicing, with a standard deviation of 4.98 percentage points, and -2.72% parallel to slicing, with a standard deviation of 6.67 percentage points. The directional difference was -8.26 percentage points, with a standard deviation of 6.44 percentage points. The reported p values were 0.0625, 0.4375 and 0.1250. These measurements were more variable than the whole-core measurements, and they too fell short of conventional significance.
Measurements were nevertheless repeatable when the same observer repeated them. Intraclass correlation coefficients, or ICCs, ranged from 0.807 to 0.985 for whole-core measurements and from 0.860 to 0.906 for glandular-lumen measurements. ICC is a consistency measure, so these ranges indicate high intra-observer repeatability in this dataset.
The evidence stays at feasibility level
Taken together, the work supports a minimally manipulated, non-destructive way to add three-dimensional structural information to archived paraffin tissue. Its strongest evidence is qualitative agreement between virtual sections and matched slides, alongside an illustrative rendering of glandular and luminal structures through tissue depth. The small sample and manual, single-region segmentation keep the result at the proof-of-concept stage.
Paper data and sources
Original title: An X‐ray micro‐CT protocol enables non‐destructive 3D visualization of glandular architecture in archived paraffin‐embedded prostate cancer tissue
Authors: Laguna‐Castro S, Mäkinen E, Tolonen T et al.
Journal/Repository: The journal of pathology. Clinical research
Status: Peer-reviewed
First online: 2026-10-13
DOI: Not available
Original paper · Full text