DeepAf: One-Shot Spatiospectral Auto-Focus Model for Digital Pathology

TL;DR

DeepAf addresses the autofocus bottleneck in digital pathology by delivering a single-shot spatiospectral focus predictor that uses a dual-encoder network to infer the optimal focal distance from one image. The approach integrates a compact robotic microscope with a spatiospectral model, enabling real-time high-quality slide imaging without focal stacking. On open data and a brain-tissue case study, it achieves a mean focus error of $0.18$ in focus distance with robust cross-lab generalization ($0.72\%$ false direction, $90\%$ within the DoF) and demonstrates practical diagnostic value with a brain cancer classifier achieving an AUC of $0.90$ at $4\times$ magnification. Overall, the work presents a hardware-software stack that democratizes high-precision digital pathology for resource-constrained settings while preserving diagnostic performance.

Abstract

While Whole Slide Imaging (WSI) scanners remain the gold standard for digitizing pathology samples, their high cost limits accessibility in many healthcare settings. Other low-cost solutions also face critical limitations: automated microscopes struggle with consistent focus across varying tissue morphology, traditional auto-focus methods require time-consuming focal stacks, and existing deep-learning approaches either need multiple input images or lack generalization capability across tissue types and staining protocols. We introduce a novel automated microscopic system powered by DeepAf, a novel auto-focus framework that uniquely combines spatial and spectral features through a hybrid architecture for single-shot focus prediction. The proposed network automatically regresses the distance to the optimal focal point using the extracted spatiospectral features and adjusts the control parameters for optimal image outcomes. Our system transforms conventional microscopes into efficient slide scanners, reducing focusing time by 80% compared to stack-based methods while achieving focus accuracy of 0.18 μm on the same-lab samples, matching the performance of dual-image methods (0.19 μm) with half the input requirements. DeepAf demonstrates robust cross-lab generalization with only 0.72% false focus predictions and 90% of predictions within the depth of field. Through an extensive clinical study of 536 brain tissue samples, our system achieves 0.90 AUC in cancer classification at 4x magnification, a significant achievement at lower magnification than typical 20x WSI scans. This results in a comprehensive hardware-software design enabling accessible, real-time digital pathology in resource-constrained settings while maintaining diagnostic accuracy.

Figures (3)

• Figure 1: Left: Implemented prototype and the schematic of the microscopic setup. Step motors $M_1$, $M_2$, $M_3$ control slide movement using processor $PI$ along $x,y,z$ respectively (a), with $M_3$ controlling the focus position (b). At each step, a low-resolution image $I_{low}$ is captured by the camera ($C$), and the non-empty images are fed to the DeepAf network to adjust the focus (c). Finally, the high resolution images $I_{high}$ are obtained with the correct focus (d).
• Figure 2: Focal Distance from Optimal Focal Point. Left: data from the same protocol, Right: data from the different protocol.
• Figure 3: Classification Performance. Receiver operating curve for the binary brain cancer classifier with an AUC score of 0.9.