Holographic 3D Spatial Localization for Bioassay Based on Depth-Label Encoding and End-to-End Attention Network

Current lens-free holographic imaging systems face limitations in achieving large depth of field and volumetric sample detection capacity. Herein, we propose a holographic bioassay platform employing depth-label encoding and end-to-end attention-enhanced 3D spatial localization (3D_EEA biosensor). It enhances the depth of field and volumetric detection capacity, significantly improving sensitivity and accuracy for detection. We integrate a depth-label encoding strategy into an attention-based neural network architecture, transforming the challenge of 3D particle localization into a dual-task framework combining bounding box detection and depth classification. This strategy overcomes the limitations of traditional generative holographic reconstruction algorithms, including slow inference speed and high cost of mechanical 3D scanning, thereby enabling accurate real-time 3D localization. This holographic 3D spatial localization facilitates extended depth-of-field applications and increases countable microsphere density, enabling chloramphenicol detection across a wide dynamic range of 5 pg/mL to 100 ng/mL and a 10-fold enhancement in sensitivity compared with conventional 2D hologram-based bioassays. Interference tests and real-sample validations demonstrate the excellent detection performance and practical potential of this platform. This work provides a scalable solution for trace-level antibiotic detection in complex matrices, with applications in food safety and environmental monitoring.