Three-dimensional vision-driven assessment of lambda-cyhalothrin-induced alterations in zebrafish swimming behavior

Swimming behaviour is considered a sensitive indicator of the potential impacts of pesticides and other environmental contaminants on fish. Leveraging a dual-view imaging and refraction-corrected 3D reconstruction workflow, and building upon improved object detection and 3D multi-object tracking, this study developed a set of techniques for identifying and analysing group-level fish swimming behaviour in three-dimensional space. From a 3D perspective, we quantified behavioural features and assessed the potential effects of lambda-cyhalothrin (LCH) on zebrafish swimming behaviour. With the improved YOLOv10n and 3D tracking methods, the proposed 3D vision-driven fish-school behavioural analysis pipeline achieved dual-view mAP@0.5 values of 99.3% and 98.3%; the maximum multi-object tracking accuracy (MOTA) increased to 93.9%, and the maximum identity consistency score (IDF1) increased to 96.8%. Behavioural analyses showed that after 9 h of exposure to 0.2 μg/L LCH, zebrafish swimming posture and orienting ability were significantly affected, with increased postural reorientation amplitude per unit time, abnormal movement paths, and more frequent non-linear swimming. Regarding the volumetric spatial distribution of fish groups, the LCH microcapsule suspension (CS) and the emulsifiable concentrate (EC) exhibited distinct effects. The LCH CS induced pronounced vertical shuttling between layers and expanded spatial coverage, whereas the LCH EC caused a contraction of the distribution range, with swimming activity largely concentrated near the tank bottom. These results provide new data support for understanding the potential risks of LCH to fish. The established 3D analysis methodology for schooling fish swimming behaviour offers new perspectives and methodological support for ecological risk assessment of pesticides and other environmental contaminants.