Figure 2

Figure 2

Comparison of eye tracking using traditional software and the DeepLabCut model. (A) Representative eye movement traces of larval zebrafish recorded during the optokinetic reflex (OKR) assay and analysed using three different pipelines: Stytra (top), OKRtrack in MATLAB (middle) and DeepLabCut (bottom). Stytra traces show missing segments (highlighted by red arrows), likely due to its reliance on image contrast thresholds and elliptical shape fitting. In contrast, both OKRtrack and DLC produce smoother and more complete traces of eye position over time, with consistent tracking of the slow-phase and saccadic components of the OKR. (B) Bland–Altman agreement plot comparing OKRtrack and DLC models. Each point represents the difference in slow-phase eye velocity between the two methods plotted against the mean value. The solid blue line denotes the mean difference (−0.043), while the outer dashed lines represent the 95% limits of agreement (−0.25 to 0.16), indicating good concordance between the methods. (C) Scatter plot comparing absolute slow-phase eye velocity values calculated by DeepLabCut (DLC) and OKRtrack for each fish. The diagonal line represents the line of identity (y = x), indicating perfect agreement. The results support the interchangeable use of DLC and OKRtrack for eye movement quantification in zebrafish OKR assays.