Genetic Analysis of Swimming Performance in Rainbow Trout (Oncorhynchus Mykiss) Using Image Traits Derived from Deep Learning

The physical and physiological condition of fish directly influences their swimming performance, which is crucial for their health and survival. This study explored how physical characteristics affect swimming performance in rainbow trout. 3D images were used to capture the holistic morphology of fish and assess its impact on critical swimming speed (Ucrit), measured via individual swim tests. A convolutional neural network (CNN) was utilized to predict Ucrit from these images. Using Gradient-weighted Class Activation Maps (GradCAM), image regions that contributed to Ucrit predictions were visualized. These regions were further refined into areas that are biologically relevant to Ucrit, leading to the definition of four swim traits: head volume, caudal fin volume, epaxial muscle volume, and shape. The interpretation of the CNN model via GradCAM was contextualized for fish physiology and coupled with genetic analysis as a quantitative evaluation to enrich the understanding of these swim traits contributing to swimming performance. Our findings indicated that Ucrit is moderately heritable. Genetically, heavier fish demonstrated poorer swimming performance; among fish of the same weight, those with larger and broader epaxial muscles, larger heads, and smaller caudal fins performed worse. Although a genetic improvement on Ucrit is feasible, caution is advised due to the potential genetic response to the reduction of the body volume and epaxial muscle volume. In this study, the workflow from data collection to model construction, visualization, interpretation, definition, and evaluation serves as a guideline for utilizing image-based deep learning models to formulate data-driven hypotheses, explain model predictions, and explore fundamental morphological characteristics to understand other complex traits in aquacultural studies.