High-frequency video analysis extends beyond the capabilities of valvometry in acute behavioral disturbance detection in bivalves

Bivalve filtering behavior has been extensively used as a sensitive in vivo indicator of water environment changes. However, greater use of this technique is hindered by the usually complex, invasive, and laborious technical approaches (glued-on sensors) needed to obtain high-resolution data on the movement of bivalve shells. Here, we introduce and test a high-frequency (5 frames per second) video analysis method for studying the potential for acute behavioral disturbance detection in bivalves. The method was tested by monitoring the behavioral response of the freshwater mussel (Anodonta anatina) to a reference toxicant, and the results were compared to data obtained through traditional valvometric evaluation using magnetic Hall sensors. Both methods showed high levels of sensitivity (Video: 0.97, 95% CI = 0.87-1; Valvometry = 0.97, 95% CI = 0.17-0.97) and specificity (Video: 0.97, 95% CI = 0.82-1; Valvometry: 0.92, 95% CI = 0.82-1) with no sig-nificant differences between the methods. Additionally, both methods performed equally well according to most binary classification metrics, such as accuracy (Video = 97.5, Valvometry = 95), positive predictive value (Video = 97.6, Valvometry = 93), negative predictive value (Video = 97.4, Valvometry = 97.3) and the area under the ROC curve (Video = 0.99, Valvometry = 0.96). A comparison between reaction times in response to stimuli of two reference toxicant concentrations (250 mg/L and 500 mg/L of nitrate-nitrogen) showed that reaction time measured from video data was significantly shorter (mean difference in reaction times = 1.56 +/- 0.89 s, paired t-test4: p = 0.01) in the 250 mg/L group due to the siphons closing first when exposed to the toxicant. We believe that approaches like the one presented here will allow future studies based on video data collection and analysis with a higher resolution than previously possible, complementing traditional gaping frequency measures and increasing our acute behavioral disturbance monitoring capabilities in bivalves.