Utilizing machine learning to automate analysis of white blood cell profiles in largemouth bass (Micropterus salmoides) and smallmouth bass (Micropterus dolomieu) [abstract]

Authors: LEET, JESSICA - Us Geological Survey (USGS); BRETZ, JOSEPH - Us Geological Survey (USGS); BRADSHAW, LILLIAN - Us Geological Survey (USGS); CLAUNCH, RACHEL - Us Geological Survey (USGS); Iwanowicz, Luke; EDWARDS, THEA - Us Geological Survey (USGS)

Submitted to: SETAC Conference
Publication Type: Abstract Only
Publication Acceptance Date: 8/30/2023
Publication Date: 11/15/2023
Citation: Leet, J.K., Bretz, J., Bradshaw, L., Claunch, R.A., Iwanowicz, L.R., Edwards, T.M. 2023. Utilizing machine learning to automate analysis of white blood cell profiles in largemouth bass (Micropterus salmoides) and smallmouth bass (Micropterus dolomieu) [abstract]. SETAC Conference. 11 12-16.

Interpretive Summary:

Technical Abstract: Fish are sensitive to adverse conditions and changes in their environment. Stressors for fish can include disease, contaminants, high temperatures, and other environmental stressors related to climate change and anthropogenic activity. Various parameters can be used to assess changes in fish health resulting from stress, and these parameters include changes in the relative abundance profiles of different white blood cell (WBC) types. However, the process of analyzing blood smears to determine the WBC profile is time-consuming and requires expertise to distinguish the different types of WBCs. We are developing a tool that automates WBC identification and counting to increase through-put and the ability of non-expert users to gain information about general fish health. A machine learning model has been trained to identify and count WBCs in largemouth bass (LMB) blood smear photos. This tool is being validated using blood smears from a study investigating the physiological and molecular mechanisms of antiviral immune response in smallmouth bass (SMB) compared to LMB. Polyinosinic:polycytidylic acid (poly I:C) was administered in vivo to juvenile SMB and LMB to mimic a viral infection and induce an antiviral immune response. Blood smears from ten fish per species intraperitoneally injected with 10 µl g-1 fish of poly I:C or sterile phosphate-buffered saline (PBS) were manually analyzed and run through the model. The results from this study indicated a shift in leukocyte profile in the poly IC challenged groups in both species to a higher neutrophil : lymphocyte ratio compared to the PBS control groups, indicating an immune response was induced. The comparison also suggested a trend toward greater magnitude of the immune response in LMB relative to SMB. To our knowledge this is the first study to compare the antiviral immune response of smallmouth and largemouth bass, as well as the first tool of this nature to automate differential blood cell analysis in bass. The start-to-end workflow developed during the establishment and validation of this tool can be implemented for future automation of additional fish health parameters to further reduce bottlenecks in fish health assessment and increase access to fish health assessment to non-expert users