An integrative web-based software tool for multi-dimensional pathology whole-slide image analytics

Abstract

Objective.In the era of precision medicine, human tumor atlas-oriented studies have been significantly facilitated by high-resolution, multi-modal tissue based microscopic pathology image analytics. To better support such tissue-based investigations, we have developed Digital Pathology Laboratory (DPLab), a publicly available web-based platform, to assist biomedical research groups, non-technical end users, and clinicians for pathology whole-slide image visualization, annotation, analysis, and sharing via web browsers.Approach.A major advancement of this work is the easy-to-follow methods to reconstruct three-dimension (3D) tissue image volumes by registering two-dimension (2D) whole-slide pathology images of serial tissue sections stained by hematoxylin and eosin (H&E), and immunohistochemistry (IHC). The integration of these serial slides stained by different methods provides cellular phenotype and pathophysiologic states in the context of a 3D tissue micro-environment. DPLab is hosted on a publicly accessible server and connected to a backend computational cluster for intensive image analysis computations, with results visualized, downloaded, and shared via a web interface.Main results.Equipped with an analysis toolbox of numerous image processing algorithms, DPLab supports continued integration of community-contributed algorithms and presents an effective solution to improve the accessibility and dissemination of image analysis algorithms by research communities.Significance.DPLab represents the first step in making next generation tissue investigation tools widely available to the research community, enabling and facilitating discovery of clinically relevant disease mechanisms in a digital 3D tissue space.

Keywords: digital pathology; high performance computing; image analysis; image registration; web computing; whole slide images.

Figure 1.

Serial pathology image registration results. Two result sets with serial WSIs of (Top) Glioblastoma and (Bottom) Liver tissue sections in H&E and Masson’s trichrome stain are demonstrated. (A) High-resolution serial WSI stacks before registration, (B) extracted low-resolution serial WSI stacks before registration, (C) resulting low-resolution WSI stacks after the low image resolution registration, and (D) high-resolution serial image stacks after the high image resolution registration, are visualized in a 3D virtual tissue space.

Figure 2.

3D Visualization and slicing interface. Users can visualize imaging data in 3D volumes directly from a browser. The 3D visualization interface allows for a variety of mouse gestures and adjustable settings to pan, zoom and slice/cut into image volumes. Options for opacity adjustment, slice spacing/thickness, as well as 3D guidelines and markers are also available for use. (A) Image volume at a thickness of 3 units, with gridlines and boundaries visible; (B) Image volume at a thickness of 1 unit, with gridlines and boundaries visible; (C) Image volume with an adjustable 3D marker placed around a feature of the image stack; (D) Image volume with diagonal clipping, revealing features displayed by cross-sectional views within the volume.

Figure 3.

3D Visualization support for multiple image modalities. (A) Registered serial H&E WSIs; (B) Registered serial IHC WSIs; (C) Registered serial WSIs stained by H&E and IHC biomarker alternatively; (D) Digital Imaging and Communications in Medicine (DICOM) radiology serial image volume of a patient with a brain neoplasia; and (E) DICOM CT serial image volume of a patient, with sinus cavities visualized.

Figure 4.

Typical 2D image analysis result visualizations from deployed 2D algorithms. Algorithms return green contours and masks in colors to represent detection and segmentation results. (A) Nuclei segmentation; (B) IHC positive pixel counting; (C) Fibrosis quantification; (D) Cell seed detection; (E) Steatosis segmentation; and (F) Nuclei segmentation with multiplexed Hyperion biomarker data.

Figure 5.

Digital Pathology server architecture and design. The splitting architecture of the DPLab server separates the web server (blue) from the computational server/cluster (red). A unified job queue and database hosted on the web server is used to manage a clustered pool of worker threads on the computational server for analyses in parallel. The architectural design of separating the computational server from the application server preserves the web application performance and provides consistent user experience, even under high computational load. It also allows for flexible scaling of the computational cluster by managing computational worker threads asynchronously, allowing new machines to be easily added to or removed from the cluster.

Figure 6.

Scalable worker and database architecture. A single unified Redis job queue is used to hold and manage jobs in a first-in-first-out priority queue on the primary web application server. A scalable pool of any number of remote worker machines with access permissions to listen to the job queue on a specific network port can retrieve jobs from the application server Redis queue and perform them asynchronously in parallel. Computed data is next written to a separately hosted cluster of database servers. Database connections from computational threads are round-robined across database servers, and databases are kept in-synchronized with Postgres built-in master-slave replication functionality. The number of machines within the database cluster and computational worker pool is scalable and flexible to real-time demand.