DigiHeart project

Machine learning to predict and prevent aquaculture mortality losses

The DigiHeart project aims to develop machine learning models to uncover new knowledge of the hitherto unknown progression of cardiac diseases in salmonids. Identifying the underlying acute mortality factors and linking these to environmental and operational parameters will improve decision making and implementation of preventive measures to reduce mortality levels.

Each year up to 20% of farmed salmonids die before slaughter. The death of large fish is of particular concern, considering the investment required for growing fish to this size. A high mortality rate also raises questions about the ethics and sustainability of the aquaculture industry.   Typically, mortality occurs during stressful incidents and interventions such as parasite treatment, deteriorating water quality, and transport to slaughter. Preliminary research has identified impaired cardiac health, often linked to deviating heart morphology, as a subsidiary element in viral diseases, such as Cardiomyopathy Syndrome (CMS), also known as “heart rupture”. The causes and effects of impaired cardiac health are poorly understood and the DigiHeart project aims to elucidate these in order to improve decision making in all production stages.    

Developing tools to improve diagnostic accuracy   

Working with partners in other Nordic countries, DNV will take advantage of the knowledge regarding different production practices used and the prevalence of heart disease in the three countries to analyze and understand the causes behind this issue. Today, operators lack models and tools to fully understand the causal relations between operations and cardiac health and predict the outcome of stressful interventions.

DigiHeart demonstrates how DNV manages to bring together experts from various domains to create new insights and value for our customers.

  • Lisa Terese de Jager ,
  • North Europe Seafood Director ,
  • DNV AS

The project aims to develop machine learning systems to improve diagnostic accuracy to enable the forecasting of disease progression and mortality based on environmental and operational factors. This approach will increase the efficiency of the diagnostic process and uncover new knowledge of previously unknown disease development, enabling more data driven planning and decision making.  The next steps of the project will include the delivery of computer vision models for heart morphology analysis, a probabilistic causal model relating operational and environmental parameters to mortality and production, and a decision support tool concept in line with the DNV’s Assurance of Digital Assets (ADA) framework in collaboration with the DNV Digital Assurance Research Center.  


The identification of operational practices and environmental factors that underlie the development of deviating cardiac morphology and health is an important step to mitigate production losses in aquaculture. The overall actionable knowledge derived from this project will allow for long term strategic planning so that fish farmers can develop more robust fish to tolerate stressful events such as transport and mechanical delousing. In short, succeeding with this cutting-edge project will optimize today’s fish farming practices, leading to lower mortality rates, reduced environmental footprint and improved fish welfare.