Scanners for Digital Pathology: Turning Tissue into Pixels
Tabular Overview of scanners for Digital Pathology
Download Overview of scanner for Digital Pathology
Digital slide scanners are at the heart of digital pathology, forming a steady beat of slide scans and a constant data flow. While very mature and proven models are available, this technology is still evolving, and the number of new scanning devices with novel features is steadily growing further.
In the past decade, digital slide scanners got significantly faster in digitizing slides. Scanning a hematoxylin eosin (HE)-stained tissue slide can be as quick as 30 seconds, fully automatic without manual interaction, at a high resolution (e. g., 0.5 to 0.25 µm/pixel). This contrasts with older devices that require manual selection of the tissue area, optional correction of focus points, and with a slow scanning time over five or ten minutes. Various models with advanced features are today available, e. g., with high slide capacity, particular robustness to preparation artefacts, with integrated slide archival, with scalability to grow as the demand increases, with a very affordable price, with support of online live AI analysis, models that are remote controllable and support live-streaming, or that are portable like a smartphone.
Big advances have further been achieved for automation, integration, and compatibility. Artificial Intelligence (AI) automatically detects the tissue area on a slide and controls image quality during and after the scan. Continuous loading of the scanners without interrupting the scanning process and automatic processing of the loaded racks and slides without manual interaction allows for seamless and easy handling. Workflow compatibility on hardware level, e. g. with the slide racks of the stainer machines, and on software level, e. g. for data exchange with the Laboratory Information System (LIS) or with the digital slide viewing software, promise plug’n-play-like integration of new scanners into the lab.
Current Challenges with Digital Slide Scanners ...
In a daily routine workflow, however, the digitization process still faces a few challenges. Scanner vendors naturally introduced their own image management software, varying in slide storage organization, unique image identifiers, whole slide image data formats and visual appearance of the images. This vendor and model heterogeneity complicates a mixture of scanners in one clinical lab, as they all come with their own software, interfaces, maintenance workflows, and cross-device compatibilities, increasing the workflow complexity. A single system is easier to administrate and favorable.
However, from a research perspective and for specific requirements and features of individual scanner models (e. g. fluorescence scanning, polarization scanning, cytology, or others), a mixture of scanners would be favorable. The heterogeneity in visual slide representations of different scanners can make AI models more robust and generalizable, if learned during training. Also, downstream software applications, such as clinical and non-clinical AI-models, increase the heterogeneity in the lab further, might be compatible with only a subset of scanner devices and introduce additional interfaces and viewing software for human interaction.
... and Their Solutions
This gap between simplicity for an efficient lab workflow and complexity due to hardware and software heterogeneity is currently being closed. E. g., an image format towards a digital imaging and communications in medicine (DICOM) standard is being developed and discussed, e. g. by the European bigpicture consortium1. More and more scanners generate DICOM images that can be shared across different systems and viewers. Further, middleware software such as digital pathology solutions fill the gap between LIS, scanners and AI. Such solutions are not necessarily needed but drastically detangle the specific requirements of individual system combinations and configurations providing more freedom and flexibility when interacting in a broader laboratory landscape with multiple vendors.