Answers are grounded in this paper and cite their sources.
Experimental evaluation demonstrates a 33% scanning time reduction in clinical pathological slides, suggesting a high-throughput solution for digital pathology scanners.
High-throughput whole slide imaging is fundamentally bottlenecked by the strict focal alignment required for high-resolution objectives, causing conventional predictive methods such as the static global plane interpolation approach to fail under unpredictable tissue thickness variations and slide warping. To address this fundamental limitation, a single-path synergistic active autofocusing system is proposed to establish a hardware-synchronized closed-loop control paradigm. By multiplexing a near-infrared auxiliary laser into the primary imaging channel, this system creates a real-time optical proxy sensor without spatial interference or mechanical deceleration overhead. Driven by this hardware, a novel computational framework seamlessly fuses a weighted radius defocus estimator to quantify absolute defocus magnitude, and a Gaussian topology direction discriminator to rigorously break spatial polarity ambiguity. Extensive experiments under continuous flying-spot scanning conditions using complex clinical pathological slides demonstrate a rapid closed-loop response latency of 37 milliseconds and a continuous tracking rate of 72 frames per second. Compared to the industrial static global plane interpolation (SGPI) standard, this synergistic control mechanism reduces overall scanning time by over 33 percent. Furthermore, it achieves an approximately 25-fold throughput acceleration relative to the state-of-the-art image-based autofocusing (IBAF) method, while maintaining comparable near-zero defocus rates. Ablation studies also strongly validate the necessity of the GTDD module in maintaining dynamic tracking accuracy, collectively establishing a highly efficient optomechatronic solution for next-generation digital pathology scanners.
No takes yet. Share an insight, caveat, or question.
Liao et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: