Saturday 15 March 2025
Researchers have made a significant breakthrough in the field of digital pathology, demonstrating that whole slide imaging technology can improve the accuracy of artificial intelligence (AI) systems in detecting mitotic figures on glass slides.
The study used z-stack scanning, a technique that captures multiple focal planes alongside the z-axis of a glass slide, to enhance the resolution and depth information of the images. The researchers then compared the performance of three AI pipelines on both single-layer and z-stacked whole slide images (WSIs) from three different scanners.
The results showed that in all combinations of scanners and AI pipelines, the z-stacked WSIs significantly improved the sensitivity of mitotic figure detection, with an average increase of 17.14%. The precision of the AI systems also remained high, with only marginal improvements observed across the different conditions.
One of the key findings was that the z-stack scanning technology provided enhanced focus control by covering a broader range of depth information, resulting in higher-quality images. For instance, the study demonstrated that while single-layer WSIs did not exhibit significant out-of-focus issues, z-stacked WSIs captured more nuanced chromosomal features, potentially enhancing AI performance.
The researchers used three different deep learning pipelines, each with its own strengths and weaknesses, to analyze the images. The PSPNet segmentation model achieved a high sensitivity on all scanners, while the Segformer model showed improved precision on two of the scanners. The DeepLabV3+ model demonstrated strong performance across all conditions.
The study’s findings have significant implications for the field of digital pathology, where AI systems are increasingly being used to aid human pathologists in diagnosing diseases. By improving the accuracy and reliability of AI-based mitotic figure detection, z-stack scanning technology has the potential to enhance workflow integration and reduce errors in diagnosis.
The researchers suggest that future studies explore optimal settings for balancing file size and imaging quality, including the number of planes, interplane distance, and compression settings, while taking scanner hardware into consideration. This could lead to more efficient and effective use of z-stack scanning technology in digital pathology.
Overall, this study demonstrates the potential of z-stack scanning technology to enhance AI-based mitotic figure detection on glass slides. By providing higher-quality images, z-stack scanning can improve the accuracy and reliability of AI systems, ultimately benefiting patient diagnosis and management.
Cite this article: “Enhancing AI Accuracy in Digital Pathology with Z-Stack Scanning Technology”, The Science Archive, 2025.
Digital Pathology, Whole Slide Imaging, Artificial Intelligence, Mitotic Figures, Z-Stack Scanning, Image Resolution, Depth Information, Focus Control, Deep Learning Pipelines, Segmentation Model
