According to the publication by Sveen et al. (2025), Atlantic salmon (Salmo salar) are one of the most valuable finfish in aquaculture, but salmon lice and their removal treatments hamper their sustainable production.
Salmon lice are the biggest biological bottleneck in marine farming. Farmers spend over one billion dollars yearly worldwide just on treatments, not counting lost growth. In Norway alone, the price tag is at around 6.6 billion NOK annually. Lice-driven mortalities, drug resistance, and welfare rules are squeezing margins across a €30 billion global industry.
Due to difficulties in establishing infestation in the Pacific salmonid species, a total of three consecutive challenge trials were conducted on the same group of fish. The first two trials, performed on unsedated fish challenged with L. salmonis (salmon louse), were unsuccessful, as they did not result in any lice attaching to pink salmon and chum salmon. The third trial, conducted on sedated fish, was successful – and resulted in tissue samples being collected at six different time points post-infestation.
The collected skin samples were examined in all four salmonid species, focusing on quantitative differences in epithelial area, mucous cell number, and mucous cell area. Molecular and histological analyses of the lice attachment site in coho salmon and Atlantic salmon were also conducted to better characterize the local responses at the lice attachment site.
After scanning, all sample images were updated to the Aiforia® Platform for the AI analysis. The team used their own skin-AI model, developed in 2019 using Aiforia® Create: its annotation tools were used on thousands of fish-skin slides to develop a ready-to-run model. The model could be used on four salmon species without extra training, letting the team compare coho, chum, pink, and Atlantic fish side-by-side.
After the AI analysis, a trained histologist manually assessed the color overlay presented by the model to determine which features were assigned to each class.
Aiforia’s automated counts revealed a three to four times higher mucous-cell density and early neutrophil surge in coho/chum skin—objective numbers that underpinned the paper’s key conclusion about lice resistance.
“Aiforia has become an essential part of our histology work. What began in 2019 with a model for segmenting fish skin and detecting mucous cells has evolved into a tool that now supports our analysis. It complements manual scoring, immunohistochemistry, and other traditional methods as highlighted in our recent paper,” Lene Sveen explains.
“Further, to make the AI models function reliably, we had to professionalize our entire workflow, from sample collection and sectioning to staining, ensuring reproducibility across trials. This structured approach has enabled us to generate high-quality data consistently,” she continues.
“Working with Aiforia has improved our entire workflow, and it has changed how we use histology in research.” – Lene Sveen, Researcher - Fish Health at Nofima
The results demonstrate that the Pacific salmon species investigated can reject most copepodid lice before they reach the chalimus stage, while Atlantic salmon do not exhibit this rejection capability, allowing lice counts to persist.
Other key findings were:
Read the related publications: