Quantitative Assessment of Focus Quality in Whole-Slide Imaging of Thyroid Liquid-Based Cytology Using Laplacian Variance

Reliable focus is essential for digital cytopathology, yet three-dimensional cell clusters in liquid-based cytology (LBC) slides pose persistent challenges for single-plane whole-slide imaging. We sought to quantitatively compare focus performance among three whole-slide scanners and to evaluate the use of Laplacian variance as an objective focus quality metric. A total of 102,573 image tiles (512 × 512 pixels) were generated from 36 whole-slide images after scanning twelve thyroid LBC slides at 40× magnification using three scanners. Single-plane mode with automatic focusing was applied. Image sharpness for each tile was computed in Python using the variance of the Laplacian, quantified as a focus score and classified into five ordinal categories, ultimately dichotomized as in-focus (variance ≥ 10) versus out-of-focus. Statistical comparisons between scanners were performed. Laplacian variance scores showed strong concordance with visual assessments of focus quality made by pathologists. One scanner consistently achieved the highest median focus scores and the most uniform spatial distribution for SurePath slides, while the other two demonstrated lower median values and greater variability. ThinPrep slides exhibited high focus quality across all scanners (> 95% in-focus), with smaller but still measurable inter-scanner differences. Overall, more than 70% of tiles were in-focus across all scanners, suggesting preserved diagnostic usability. Laplacian variance is a practical and reproducible tool for assessing focus quality in digital cytology. Focus performance varied among scanners, particularly for slides with three-dimensional cell clusters, whereas differences were less pronounced for monolayer ThinPrep preparations. Quantitative focus assessment may support scanner selection, quality assurance, and computational cytopathology workflows.