gms | German Medical Science

Introduction: Digital pathology provides novel ways for pathologists to view, store, transmit, manage, process, and analyze tissue samples. As microscopic images are more and more being used in a digital format, their well-organized integration into laboratory information management systems (LIMS) becomes significantly more important for primary diagnosis, research, and education. Currently, there are no standard guidelines regarding the handling of whole slide images unlike radiology images, for which DICOM (Digital Image Communication in Medicine) became the de-facto standard. The very large virtual microscope images are hitting some limitations of DICOM, such as the maximum size of uncompressed image pixel data (4GB) caused by the underlying 32-bit architecture [1].

As a subproject of the Competence Network Multiple Sclerosis (KKNMS) [2], the Multiple Sclerosis Brain Bank (MS-BB) serves as a use case for the application of Digital Pathology system (DPS) in clinical research. Within the scope of the project, high-resolution scans of histopathologic samples are to be provided to researchers before shipping specific brain tissue to facilitate a digital view into the sample. The objective of the work is to find out significant requirements for the DPS in the environment of the research network, in order to develop an interoperable microscope image handling system, efficiently integrated into the existing IT infrastructure of the KKNMS.

Materials and methods: In order to determine the requirements, the following five steps were performed. At first, the workflow of digital pathology was examined, where the characteristics of the MS lesion, staining procedures, clinical and biomaterial data were assessed. Second, the scanning process was analyzed and several histopathologic samples supplied by the Department of Neuropathology were scanned. During this, technical details like file size, format, and structure, scanning time, and batch processing functionality were evaluated. Third, the documentation process for biomaterial with the local LIMS was assessed, where tissue samples of patients are registered. Forth, an investigation of image viewers and handling systems for digital pathology was carried out on the market. In the final step, the requirements were drawn from the materials collected in the previous four steps according to the procedure of theoretical assumptions of the requirements engineering [3].

Results: During the analysis of the workflows, three significant stages were identified: preparation, digitalization, and provision. The preparation process comprises the standard operating procedures for the autopsy of the MS patients, tissue collection, the histological staining procedures and the characterization of the MS lesions with regard to the localization and lesion activity.

During the assessment of the digitalization stage, we used the Olympus dotSlide microscope and significant parameters such as file size and scanning time were regarded. Digital pathology works with huge data volumes and it requires support of big data storage solutions. Average file size of a 15mm x 20mm tissue specimen varies between 2.8GB for 40x and 1GB for 20x magnification. Additionally, the scanning duration plays an important role, especially when using the higher magnification. Corresponding scanning durations are 1h 25min for 40x and 25min for 20x scan. Historically, scanning speed has been improved with every new generation of scanners, but data volume, especially if fluorescence tissues or z-stack images are scanned, has also gone up. The file format of our images is .vsi, but contemporary digital scanners support conversion of own format to standard file formats such as TIFF or JPEG2000 with loss of accompanying metadata.

To provide an overview of tissue samples for requested research questions, the clinical parameters of patients in conjunction with the digital images are important for researchers of KKNMS. The conclusion of this is that DPS needs to be integrated into a comprehensive IT infrastructure including phenotype and biomaterial data. Tissue samples, registered in the LIMS, are assigned to the specific brain area and have a unique sample ID generated by the LIMS. DPS should be separated from the LIMS and includes an image server, database, and web viewer for microscopic images. Digital slides in conjunction with metadata are manually uploaded to the DPS after the completion of the scanning process. A URL and thumbnail of each digital slide are registered for tissue sample into the LIMS in order to enable launching of a DPS viewer directly from within the LIMS [4]. Most of the LIMS can work with web-services that are the most common mechanism for interfacing with external systems (e.g. SOAP, REST).

Finally, the current web viewers for microscopic images were investigated that identified supplementary features for the DPS. 15 different systems were examined, basing on different platform and supporting different file formats. Based on depicted investigations, the following features have been determined as required for a DPS: tools for uploading images; storing of accompanying metadata in conjunction with the images; web viewer with navigation tools for image viewing; support of high resolution images and different file formats; storage infrastructure for large digital files. In addition, desirable functionalities of system are: tools for adding annotations; support of multiple images viewing; questionnaire of digital images for researchers; take snapshot of region of interest. Last but not least, security of data and user account is one of the important features of system.

Discussion: The work presented here sums up the functional, technical, system, and environmental requirements for integration of digital pathology into the already existing IT infrastructure of a matured research network. Based on described features catalog, next stage of research is to determine, implement, and integrate a viable solution of MS-BB IT infrastructure for archiving, retrieving, and distribution of the microscopic images into the LIMS. The limitations of this project need to be considered. First, currently the study presented here has not evaluated the usage of digital scanners of different vendors, but it is planned to develop vendor neutron DPS within the scope of KKNMS. Second, underlying workflows resemble the MS-BB project; Different projects may have distinct workflows and thus different needs, but the goal was to implement a working prototype that can later be extended to other projects. It is expected that web-based DPS will probably become useful product to facilitate collaboration among the researchers of KKNMS.

Acknowledgements: This work was supported by the Competence Network Multiples Sclerosis (01GI1304B), funded by the German Federal Ministry of Education and Research.