Z-Stack Scanning can Improve AI Detection of Mitosis: A Case Study of Meningiomas

TL;DR

Mitosis detection in meningiomas on whole slide images is challenged by the small, subtle nature of the feature and variability across scanners. The authors compare single-layer versus z-stack WSIs across three scanners using a two-stage DL pipeline (segmentation models: PSPNet, SegFormer, DeepLabV3+; CNN ensemble: EfficientNet variants) with per-plane processing and random-forest fusion, calibrated per scanner. Z-stack scanning yields an average sensitivity gain of $+17.14%$ (from $0.601$ to $0.704$) with precision largely unchanged ($0.753$ to $0.757$); the largest improvement reaches $+39.24%$ for a specific scanner/pipeline combination. This demonstrates that z-stack scanning can enhance AI-assisted mitosis detection in meningiomas across scanners and DL configurations, while increasing file size by about $3.8 imes$ and prompting optimization of plane count and compression for deployment.

Abstract

Z-stack scanning is an emerging whole slide imaging technology that captures multiple focal planes alongside the z-axis of a glass slide. Because z-stacking can offer enhanced depth information compared to the single-layer whole slide imaging, this technology can be particularly useful in analyzing small-scaled histopathological patterns. However, its actual clinical impact remains debated with mixed results. To clarify this, we investigate the effect of z-stack scanning on artificial intelligence (AI) mitosis detection of meningiomas. With the same set of 22 Hematoxylin and Eosin meningioma glass slides scanned by three different digital pathology scanners, we tested the performance of three AI pipelines on both single-layer and z-stacked whole slide images (WSIs). Results showed that in all scanner-AI combinations, z-stacked WSIs significantly increased AI's sensitivity (+17.14%) on the mitosis detection with only a marginal impact on precision. Our findings provide quantitative evidence that highlights z-stack scanning as a promising technique for AI mitosis detection, paving the way for more reliable AI-assisted pathology workflows, which can ultimately benefit patient management.

Figures (4)

• Figure 1: Deep learning mitosis detection pipeline for (a) single-layer WSIs and (b) z-stacked WSIs.
• Figure 2: Examples of mitoses annotated on WSIs from the Pannoramic 250 scanner (+0.0$\mu m$ focus plane), bar=5$\mu m$.
• Figure 3: Examples of mitoses under the single-layer and z-stack scanning with three scanners, bar=5$\mu m$.
• Figure 4: Examples of mitoses missed by the deep learning pipeline with DeepLabV3+ segmentation model on single-layer WSIs but were captured under z-stacked WSIs, bar=5$\mu m$.