Artificial Intelligence for Computer-Assisted Diagnosis of Hyperplasia in Atlantic Salmon Gill Histology Images

Measuring hyperplasia in Atlantic salmon gills can provide valuable insights into fish health. In this study, we propose an innovative technique for classifying histology images to identify regions of hyperplasia. Our pipeline utilises novel signal processing techniques in conjunction with prototypical deep learning methods to analyse image texture. We hypothesise and demonstrate that our method effectively captures distinct features of gill histopathology whole-slide images, thereby enhancing the classification task. Compared to conventional deep learning methods, our hybrid approach exhibits exceptional performance in speed and accuracy. When further developed, the concept can support conventional histopathological assessment by providing a computer-assisted hyperplasia score as an objective quantitative histopathological endpoint. The workflow is translatable to other gill conditions and histopathology images beyond gills.