The tissue micro-array data exchange specification: a web based experience browsing imported data

Abstract

Background: The AIDS and Cancer Specimen Resource (ACSR) is an HIV/AIDS tissue bank consortium sponsored by the National Cancer Institute (NCI) Division of Cancer Treatment and Diagnosis (DCTD). The ACSR offers to approved researchers HIV infected biologic samples and uninfected control tissues including tissue cores in micro-arrays (TMA) accompanied by de-identified clinical data. Researchers interested in the type and quality of TMA tissue cores and the associated clinical data need an efficient method for viewing available TMA materials. Because each of the tissue samples within a TMA has separate data including a core tissue digital image and clinical data, an organized, standard approach to producing, navigating and publishing such data is necessary. The Association for Pathology Informatics (API) extensible mark-up language (XML) TMA data exchange specification (TMA DES) proposed in April 2003 provides a common format for TMA data. Exporting TMA data into the proposed format offers an opportunity to implement the API TMA DES. Using our public BrowseTMA tool, we created a web site that organizes and cross references TMA lists, digital "virtual slide" images, TMA DES export data, linked legends and clinical details for researchers. Microsoft Excel and Microsoft Word are used to convert tabular clinical data and produce an XML file in the TMA DES format. The BrowseTMA tool contains Extensible Stylesheet Language Transformation (XSLT) scripts that convert XML data into Hyper-Text Mark-up Language (HTML) web pages with hyperlinks automatically added to allow rapid navigation.

Results: Block lists, virtual slide images, legends, clinical details and exports have been placed on the ACSR web site for 14 blocks with 1623 cores of 2.0, 1.0 and 0.6 mm sizes. Our virtual microscope can be used to view and annotate these TMA images. Researchers can readily navigate from TMA block lists to TMA legends and to clinical details for a selected tissue core. Exports for 11 blocks with 3812 cores from three other institutions were processed with the BrowseTMA tool. Fifty common data elements (CDE) from the TMA DES were used and 42 more created for site-specific data. Researchers can download TMA clinical data in the TMA DES format.

Conclusion: Virtual TMAs with clinical data can be viewed on the Internet by interested researchers using the BrowseTMA tool. We have organized our approach to producing, sorting, navigating and publishing TMA information to facilitate such review. We have converted Excel TMA data into TMA DES XML, and imported it and TMA DES XML from another institution into BrowseTMA to produce web pages that allow us to browse through the merged data. We proposed enhancements to the TMA DES as a result of this experience. We implemented improvements to the API TMA DES as a result of using exported data from several institutions. A document type definition was written for the API TMA DES (that optionally includes proposed enhancements). Independent validators can be used to check exports against the DTD (with or without the proposed enhancements). Linking tissue core images to readily navigable clinical data greatly improves the value of the TMA.

Figure 1

Producing a TMA block and slide. A slide from a source block is marked to indicate the area of interest. The marked slide is used to locate the place to punch out a core. The core is extracted from the source block and then inserted into a wax TMA block with cores from other source blocks. TMA slides are then cut from the TMA block.

Figure 2

Sample image viewed with Virtual Microscope.

Figure 3

Implementation. A DTD governs the allowed data elements in each type (TMA block list, TMA LDE, TMA DES export & style) of XML file. The TMA block list uses BrowseTMA to process each listed block file and style. BrowseTMA produces a single HTML TMA block list file, a TMA block HTML file for each specified block and an HTML TMA style file for each style used.

Figure 4

Producing a block list. The TMA file list is written in a simple XML language. The language is defined by tmaList.dtd, a document type definition that allows 13 common data elements (ex. tma-file-name, style-id). Referencing this DTD causes validation of the entries indicating which TMA DES export files and blocks to use with which styles. For each TMA file in the TMA file list, BrowseTMA processes the specified TMA blocks and displays a row of information about each. In processing a TMA block, BrowseTMA uses that block's data from the XML export file to display the export file name, block identifier, purpose and description and the TMA file list entry to display the style file name and links to the style and possibly to an image.

Figure 5

Producing a legend. The style file specifies the layout of the array (or matrix) and the legend cell contents and colors using a simple XML language defined in the addStyles conditional section of tmades.dtd. These 26 definitions (ex. core-location-id, row-label, cell-color) allow specification of the TMA layout. They are supplemented by common and locally defined data elements that are nested under the core data element (ex. core_TISSUE, core_histo-repository_donor-block) which may be used to determine the legend cell color and contents and the details columns. Referencing this DTD causes validation of the entries that specify labelling, row and column grouping of cell identifiers, legend cell contents and colors. Cell identifiers (ex. A1, R1C1, or 1.1) are specified in rows with row and column labels interspersed inside the matrix element. The legend-cell-content element contains a list of clinical data items that should be placed in each cell in the HTML legend table. The legend-cell-colors element contains a clinical data item and a list of cell-color elements. The data from the XML export file corresponding to the specified core data item (core_TISSUE is used here) is compared against the substring and exact-match entries inside the cell-color elements and the corresponding color is used for the legend cell in the HTML legend table when a match is found.

Figure 6

Producing the clinical details table. The style file specifies an ordered list of columns and their headings using a simple XML language defined in the addStyles conditional section of tmades.dtd, definitions that allow any data element nested inside the core CDE in the XML export file (ex. core_histo-repository) including locally defined data elements to be used. Activating addStyles causes validation of the entries that specify each details table column. Entries nested inside one or more data elements inside core are allowed. Each core data entry in the XML export file provides one row of cells in the resulting HTML details table.