EUVIMEDEUVIMED entdecken
Crossref

Coherence, alignment and fibre-profile geometry in a hierarchical texture-descriptor tensor: application to H&E sections of skeletal muscle in wild boar (Sus scrofa) and domestic pigs

Ştefan Ţălu

Porcine Research · 30.06.2027 · 17 (1) · 1

DOI: 10.67784/0098 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Porcine Research, online erschienen am 31.10.2026. Verfasst von Ştefan Ţălu. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 40 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Muscle fibre size, shape and the organisation of intramuscular connective tissue are central to pork quality, but histological sections are usually summarised by fibre counts and diameters, which require segmentation and ignore direction. This study extends a hierarchical texture-descriptor tensor (channel × scale × orientation × region of interest, ROI) with three results that decompose its anisotropy index A: (1) an exact factorisation of A into energy-weighted local coherence and alignment of local axes, which trade off monotonically with the integration scale and cross at √A (for 0 < A < 1); (2) an exact decomposition of A over tissue compartments of arbitrary area, from which a two-compartment cancellation law and a dilution law follow; and (3) a closed-form boundary-profile law giving the anisotropy produced by elliptical fibre profiles of aspect ratio a in terms of complete elliptic integrals. The tensor (10 channels, 4 scales, 4 orientations; 1 and 4 ROIs) was applied to eight published, calibrated haematoxylin-eosin images of longissimus dorsi and biceps femoris: four of wild boar (three transverse, one longitudinal) and two magnifications each of Xiang and Large White pigs. The boundary-profile law was reproduced on synthetic ellipses within 0.003 and on synthetic mosaics within 0.004. All eight sections were anisotropic relative to isotropic surrogates (P ≤ 0.005 at s = 2 px), with A from 0.019 (Large White, detail) to 0.697 (wild boar, longitudinal; texture axis 90.8°, along the fibres). The interfibre band carried 0.83–0.93 of the gradient energy on 0.24–0.50 of the area; where wide perimysial septa were present they dominated A, and in the Large White detail field septal and endomysial contributions with nearly perpendicular axes cancelled, reducing A by 90 %. At the fascicle scale (s = 22 µm) the Large White overview was strongly anisotropic and the Xiang overview was not (A = 0.480 versus 0.071, non-overlapping intervals), and the characteristic scale of Xiang muscle agreed across magnifications (13.5 and 13.6 µm), whereas that of Large White did not (15.9 and 5.2 µm). Median minor-axis lengths of segmented fibres were 42.4 µm (Xiang, detail) and 61.6 µm (Large White, overview), close to the diameters reported by the source study, but the other two fields gave 53.5 and 53.3 µm, so agreement depended on the field. The results show that the tensor separates fibre-level and fascicle-level organisation of porcine muscle; the tensor itself requires no segmentation, whereas the compartment and fibre-profile analyses do; breed or quality differences are not claimed from single published images.

Zitieren

APA

Ţălu, Ş. (2027). Coherence, alignment and fibre-profile geometry in a hierarchical texture-descriptor tensor: application to H&E sections of skeletal muscle in wild boar (Sus scrofa) and domestic pigs. Porcine Research, 17(1), 1. https://doi.org/10.67784/0098

Vancouver

Ţălu Ş. Coherence, alignment and fibre-profile geometry in a hierarchical texture-descriptor tensor: application to H&E sections of skeletal muscle in wild boar (Sus scrofa) and domestic pigs. Porcine Research. 2027;17(1):1. doi:10.67784/0098

BibTeX
@article{lu2027,
  author = {Ţălu, Ştefan},
  title = {Coherence, alignment and fibre-profile geometry in a hierarchical texture-descriptor tensor: application to H&E sections of skeletal muscle in wild boar (Sus scrofa) and domestic pigs},
  journal = {Porcine Research},
  year = {2027},
  volume = {17},
  number = {1},
  pages = {1},
  doi = {10.67784/0098}
}
Kostenlos starten →