Poeciliid Research / Article

Quantitative multiscale analysis of spatial texture organization in liver histol...

Research Article
Quantitative multiscale analysis of spatial texture organization in liver histology of Poecilia reticulata
Ştefan Ţălu
The Directorate of Research, Development and Innovation Management (DMCDI), The Technical University of Cluj-Napoca, Cluj-Napoca, Romania. Corresponding author: Ş. Ţălu, stefan.talu@auto.utcluj.ro
Published2026
JournalPoeciliid Research
Volume / Issue16(1)/2026
Pagespp. 75-105
AccessOpen Access

Abstract

Quantitative characterization of liver histology requires analytical approaches capable of describing spatial heterogeneity and directional organization beyond conventional visual assessment and global intensity statistics. This study presents an exploratory multiscale texture analysis of four published liver histology panels of Poecilia reticulata reported in a study investigating co-exposure to iron oxide nanoparticles and glyphosate-based herbicides. The original investigation documented hepatic histopathological alterations using complementary histological, histochemical, and ultrastructural methods. Building on this biological context, the present study examines image-derived texture properties without assuming that individual descriptors directly measure specific lesions or toxicological mechanisms. Ten complementary texture descriptors were evaluated at four Gaussian scales and four orientations and organized into a multidimensional descriptor tensor (160 entries per panel for a single global region of interest, and 640 entries per panel for a four-region partition). Global measurements and regional analysis were used to investigate scale-dependent responses, directional anisotropy, spatial heterogeneity, and inter-descriptor relationships. In the analysed panels, the control panel (A) and the vasocongested panel (B) showed measurable directional organization (sampled anisotropy index A₄ = 0.19 and 0.18 at s = 2 px), with texture axes near 93° and 177°, whereas the steatotic panel (C) was close to isotropic (A₄ = 0.04). The unstained vacuole-like area fraction increased from 2 % in A to 17 %, 31 % and 22 % in B, C and D. Regional analysis showed that spatial variability depended on the descriptor: GLCM contrast and energy varied by up to 12 % and 15 % between regions, whereas entropy, GLCM correlation and homogeneity were stable (coefficient of variation < 3 %). The gradient channels D1, D2 and D5 were practically redundant (Spearman ρ ≥ 0.99). The study is intended to establish an exploratory quantitative description of the selected histological images rather than a validated classification of liver condition. Because the analysis is based on selected published image panels rather than independent biological replicates, any observed differences must be interpreted at the image level. The approach provides a structured basis for investigating multiscale and multidirectional texture organization in guppy liver histology, while highlighting the need for standardized image acquisition and independent biological validation before broader biological or diagnostic conclusions can be drawn.

Keywords

Poecilia reticulata directional anisotropy fish liver histology histological image analysis multiscale texture analysis spatial heterogeneity texture descriptors quantitative image analysis
Download PDF ← Back to Issue