Multimodal-Sensor Fusion and Machine-Learning for 24 h Early Warning of Sub-clinical Ketosis in Dairy Cattle: A Systematic Review and Meta-analysis

Sub-clinical ketosis (SCK) affects >30 % of fresh cows and predisposes them to displaced abomasum, mastitis and reduced fertility. Commercial monitors still rely on single-modality alerts (e.g., neck accelerometers or milk FTIR). Since 2023 the field has moved toward multimodal fusion models that combine vision, acoustics, IMU, FTIR and ambient THI streams, often with attention layers or cross-modal transformers (ResearchGate, PMC). No systematic review has yet quantified how much diagnostic accuracy is gained by fusing ≥2 modalities, nor has anyone assessed adherence to emerging AI-reporting check-lists (PRISMA-DTA, STARD-AI).