Gill Alterations in Adult Zebrafish Exposed to Avobenzone: Morphometric Profiling and Machine-Learning Analysis

Gill histopathology is widely used in fish toxicology, but lesion scores and targeted measurements capture only part of the structural information in routine sections. We tested whether a morphometric profile could quantify avobenzone-related changes in adult zebrafish gills. The 40× dataset comprised gill sections from 35 adult zebrafish across five conditions represented in the image dataset: a 4-day solvent-control group and fish exposed to nominal avobenzone concentrations of 3 or 9 $\mathsf{\mu }$g/L for 4 or 21 days. From 750 fields, quality control and automated segmentation yielded 503 accepted masks; 360 fields (72 per condition) were used in the balanced analysis. Thirty-nine descriptors quantified tissue geometry, interlamellar spaces, branch-like organisation, and thickness heterogeneity. Validation was repeated 30 times with all fields from each fish kept in the same partition. The time-matched 4-day control-versus-exposed comparison reached 89.4 ± 4.1% balanced accuracy. The 3-versus-9 $\mathsf{\mu }$g/L comparison at 4 days reached 88.7 ± 4.6%, and 4 versus 21 days within the 9 $\mathsf{\mu }$g/L group reached 88.5 ± 5.0%. Performance for direct five-condition classification was lower (68.4 ± 5.1%). Because 21-day control images were not included, exposure-related inference was restricted to the time-matched 4-day comparison. Morphometric profiling provides a quantitative complement to conventional histopathology while preserving individual fish identity as the grouping unit for image analysis.