Hoof lesion detection in lactating Holsteins: Part I. External validation of autonomous camera-derived locomotion...

Hoof lesion detection remains a challenge in lameness management. Recent studies have proposed locomotion score-based thresholds using an autonomous camera system (AUTO), which generates locomotion scores (LS) ranging from 0 (sound) to 100 (severely lame). However, these thresholds have been developed and evaluated within the same study populations, and their performance on independent herds remains unknown, raising concerns about generalizability. The primary objective of this study was to evaluate the classification performance of AUTO-derived locomotion metrics for identifying cows with hoof lesions. Specifically, we assessed the performance of a weekly average LS threshold ≥50 (50_BIN) and 4 monthly AUTO-derived locomotion metrics: average LS (mAVG), minimum LS (mMIN), maximum LS (mMAX), and percentage of LS scores above 50 (mPLS). Performance metrics, including positive predictive value (PPV) and negative predictive value (NPV), were calculated for each threshold. Holstein cows (n = 511) of 8 to 100 DIM, with no active lesions and no more than 1 prior noninfectious lesion, were enrolled from 2 sites. Cows were subsequently classified as typically treated, which included cases of digital dermatitis, sole ulcer, toe lesion, white line disease, or other hoof lesions requiring treatment; or as typically not treated, which included cows without hoof lesions or without lesions typically requiring treatment. For positive classifications at the overall parity level, 50_BIN achieved the highest PPV (n = 451; PPV = 25.3%, 95% CI: 16.1-35.1), whereas mPLS achieved the highest NPV (n = 389; NPV = 92.4%, 95% CI: 89.0-95.6). These results highlight that threshold-based approaches failed to reliably identify cows with lesions, mainly due to low PPV. Based on these findings, we do not support the on-farm adoption of these AUTO-derived threshold methods for the identification and differentiation of cows with hoof lesions. The low PPV indicates that adopting these thresholds could increase unnecessary hoof trimming, associated labor demands, and economic costs. Future research should prioritize improving PPV to accurately identify cows with lesions and avoid potentially unnecessary hoof trimming visits.