Deep learning for the segmentation of preserved photoreceptors on en face optical coherence tomography in two inherited retinal diseases.

In this study, Atlantic salmon were: (i) implanted with heart rate (fH) data storage tags (DSTs), pharmacologically stimulated to maximum fH, and warmed at 10°C h-1 (i.e. tested using a 'rapid screening protocol'); (ii) fitted with Doppler® flow probes, recovered in respirometers and given a critical thermal maximum (CTmax) test at 2°C h-1; and (iii) implanted with fH DSTs, recovered in a tank with conspecifics for 4 weeks, and had their CTmax determined at 2°C h-1. Fish in respirometers and those free-swimming were also exposed to a stepwise decrease in water oxygen level (100% to 30% air saturation) to determine the oxygen level at which bradycardia occurred. Resting fH was much lower in free-swimming fish than in those in respirometers (∼49 versus 69 beats min-1) and this was reflected in their scope for fH (∼104 versus 71 beats min-1) and CTmax (27.7 versus 25.9°C). Further, the Arrhenius breakpoint temperature and temperature at peak fH for free-swimming fish were considerably greater than for those tested in the respirometers and given a rapid screening protocol (18.4, 18.1 and 14.6°C; and 26.5, 23.2 and 20.2°C, respectively). Finally, the oxygen level at which bradycardia occurred was significantly higher in free-swimming salmon than in those in respirometers (∼62% versus 53% air saturation). These results: highlight the limitations of some lab-based methods of determining fH parameters and thermal tolerance in fishes; and suggest that scope for fH may be a more reliable and predictive measure of a fish's upper thermal tolerance than their peak fH.