“Machine Learning-Based Quantification of Cardiomyopathy Syndrome (CMS) Inflammation in Atlantic Salmon Hearts” (7201)

Machine Learning-Based Quantification of Cardiomyopathy Syndrome (CMS) Inflammation in Atlantic Salmon Hearts

INTRODUCTION

Cardiomyopathy syndrome (CMS) is a viral heart disease in farmed Atlantic salmon, caused by the piscine myocarditis virus (PMCV), leading to high morbidity and mortality. The disease is characterized by inflammation in the atrium and the spongious layer of the ventricle, resulting in poor fish welfare and substantial economic losses for aquaculture producers.

To increase resistance to CMS, selective breeding programs are conducted alongside vaccine development. Reliable methods for assessing disease severity are crucial for evaluating these interventions, and histopathology is considered the gold standard. Traditionally, CMS pathology has been graded histologically using a visual scoring model (0–4 scale), but this method is limited in precision, reproducibility, and efficiency. To address these limitations, a machine learning-based analysis, has been introduced for more precise and reproducible grading.

METHOD

A commercially available software utilizing algorithm-based machine learning was used to analyze digital histological sections of Atlantic salmon hearts. An application (APP), designed and trained by histopathologists, quantitatively distinguishes inflammation from healthy myocardial tissue in the atrium and spongious ventricle layer. The analysis thus quantifies the area of inflammation relative to healthy tissue within the sample.

RESULTS

Applying machine learning to histological heart samples with CMS inflammation provided a high resolution, quantitative histopathological endpoint (“PatoPrecision score”) on a continuous scale from 0 to 100, resulting in more precise scores than a traditional semi-quantitative score. Further the method ensures consistent and fully reproducible results, eliminating observer bias. However, the APP is sensitive to tissue section quality and artefacts, which must be considered.

CONCLUSIONS

The PatoPrecision score for CMS inflammation has been applied in multiple infection trials, collectively analyzing over 10,000 heart samples. The resulting (percentage inflammation) data have supported ongoing selective breeding efforts to enhance resistance to CMS by providing geneticists a novel tool for assessment of a phenotypic trait for selection, with excellent precision and consistency.

1. YRJOENHEIKKI, ANNELI, PATOGEN, Presenter
2. BAILY, JOHANNA, PATOGEN, Author
3. GAMLEM, NORALF, PATOGEN, Author
4. THOEN, EVEN, PATOGEN, Author