DESCRIPTION

LIST OF KNOWN PROPERTIES

Properties generated by OpenSlide

openslide.background-color

The background color of the slide, given as an RGB hex triplet. This property is not always present.

openslide.bounds-height

The height of the rectangle bounding the non-empty region of the slide. This property is not always present.

openslide.bounds-width

The width of the rectangle bounding the non-empty region of the slide. This property is not always present.

openslide.bounds-x

The X coordinate of the rectangle bounding the non-empty region of the slide. This property is not always present.

openslide.bounds-y

The Y coordinate of the rectangle bounding the non-empty region of the slide. This property is not always present.

openslide.comment

A free-form text comment.

openslide.mpp-x

Microns per pixel in the X dimension of level 0. May not be present or accurate.

openslide.mpp-y

Microns per pixel in the Y dimension of level 0. May not be present or accurate.

openslide.objective-power

Magnification power of the objective. Often inaccurate; sometimes missing.

openslide.quickhash-1

A non-cryptographic hash of a subset of the slide data. It can be used to uniquely identify a particular virtual slide, but cannot be used to detect file corruption or modification.

openslide.vendor

The name of the vendor backend.

Properties for TIFF-based formats

tiff.Artist

The contents of the TIFF Artist tag.

tiff.Copyright

The contents of the TIFF Copyright tag.

tiff.DateTime

The contents of the TIFF DateTime tag.

tiff.DocumentName

The contents of the TIFF DocumentName tag.

tiff.HostComputer

The contents of the TIFF HostComputer tag.

tiff.ImageDescription

The contents of the TIFF ImageDescripton tag.

tiff.Make

The contents of the TIFF Make tag.

tiff.Model

The contents of the TIFF Model tag.

tiff.ResolutionUnit

The contents of the TIFF ResolutionUnit tag.

tiff.Software

The contents of the TIFF Software tag.

tiff.XPosition

The contents of the TIFF XPosition tag.

tiff.XResolution

The contents of the TIFF XResolution tag.

tiff.YPosition

The contents of the TIFF YPosition tag.

tiff.YResolution

The contents of the TIFF YResolution tag.

Vendor-specific properties

A list of vendor-specific properties can be found on the pages for each vendor format, linked from m[blue]Supported Virtual Slide Formatsm[][1].

TRESTLE FORMAT

Format

single-file pyramidal tiled TIFF, with non-standard metadata and overlaps; additional files contain more metadata and detailed overlap info

File extensions

.tif

OpenSlide vendor backend

trestle

Detection

Trestle slides are stored in single-file TIFF format. OpenSlide will detect a file as Trestle if:

1. The file is TIFF.

2. The TIFF Software tag starts with MedScan.

3. The ImageDescription tag is present.

4. All images are tiled.

Relevant TIFF tags

Extra data stored in ImageDescription

The ImageDescription tag contains semicolon-delimited key-value pairs. A key-value pair is equals-delimited. We use the OverlapsXY and Background Color keys from the ImageDescription, and ignore the rest. All of these values are stored as properties starting with "trestle.".

TIFF Image Directory Organization

The first image in the TIFF file is the full-resolution image. The subsequent images are assumed to be decreasingly sized reduced-resolution images.

Overlaps

The OverlapsXY pseudo-field encodes a list of tile overlap values as ASCII.

Example: "64 64 32 32 16 16" (note the initial space).

These values represent the standard overlaps between adjacent tiles in X and Y, in pixels. This example encodes 3 levels worth of overlaps. Further overlaps are assumed to have the value 0.

Individual tile overlaps may differ from the standard overlaps. These individual overlaps are recorded in .tif-Nb files adjacent to the .tif file, where N is the level number. OpenSlide does not read these files, though they have been partially decoded; see m[blue]issue 21m[][2] for details.

Associated Images

macro

the image with a filename extension of ".Full" (optional)

Known Properties

All data encoded in the ImageDescription TIFF field is represented as properties prefixed with "trestle.".

openslide.mpp-x

copy of tiff.XResolution (note that this is a totally non-standard use of this TIFF tag)

openslide.mpp-y

copy of tiff.YResolution (note that this is a totally non-standard use of this TIFF tag)

openslide.objective-power

normalized trestle.Objective Power

Test Data

m[blue]http://openslide.cs.cmu.edu/download/openslide-testdata/Trestle/m[]

HAMAMATSU FORMAT

Format

multi-file JPEG/NGR with proprietary metadata and index file formats, and single-file TIFF-like format with proprietary metadata

File extensions

.vms, .vmu, .ndpi

OpenSlide vendor backend

hamamatsu

Detection

OpenSlide will detect a file as Hamamatsu if:

1. The file given is a INI-style text file.

2. It has a [Virtual Microscope Specimen] (VMS) or [Uncompressed Virtual Microscope Specimen] (VMU) group.

3. If VMS, there are at least 1 row and 1 column of JPEG images (NoJpegColumns and NoJpegRows).

or if:

1. The file has a TIFF directory structure.

2. The Software tag starts with NDP.scan.

Overview

The Hamamatsu format has three variants. VMS and VMU consist of an index file, 2 or more image files, and (in the case of VMS) an "optimisation" file. NDPI consists of a single TIFF-like file with some custom TIFF tags. VMS and NDPI contain JPEG images; VMU contains NGR images (a custom uncompressed format).

Multiple focal planes are ignored, only focal plane 0 is read.

JPEG does not allow for files larger than 65535 pixels on a side. In VMS, multiple JPEG files are used to encode large images. To avoid having many files, VMS uses close to maximum size (65K by 65K) JPEG files. NDPI, instead, stuffs large levels into a single JPEG and sets the overflowed width/height fields to 0.

Unfortunately, JPEG provides very poor support for random-access decoding of parts of a file. To get around this, JPEG restart markers are placed at regular intervals, and these offsets are specified in the optimisation file (in VMS) or in a TIFF tag (in NDPI). With restart markers identified, OpenSlide can treat JPEG as a tiled format, where the height is the height of an MCU row, and the width is the number of MCUs per row divided by the restart marker interval times the width of an MCU. (This often leads to oddly-shaped and inefficient tiles of 4096x8, for example.)

Unfortunately, the VMS optimisation file does not give the location of every restart marker, only the ones found at the beginning of an MCU row. It also seems that the file ends early, and does not give the location of the restart marker at the last MCU row of the last image file.

Thus, the optimisation file can only be taken as a hint, and cannot be trusted. The entire set of JPEG files must be scanned for restart markers in order to facilitate random access. OpenSlide does this lazily as needed, and also in a background thread that runs only when OpenSlide is otherwise idle.

The VMS map file is a lower-resolution version of the other images, and can be used to make a 2-level JPEG pyramid. JPEG also allows for lower-resolution decoding, so further pyramid levels are synthesized from each JPEG file.

VMS File

The .vms file is the main index file for the VMS format. It is a Windows INI-style key-value pair file, with sections. Only keys in the Virtual Microscope Specimen group are read by OpenSlide.

Here are known keys from the file:

VMU File

The .vmu file is the main index file for the VMU format. Only keys in the Uncompressed Virtual Microscope Specimen group are read by OpenSlide.

Here are known keys from the file:

NDPI File

NDPI uses a TIFF-like structure, but libtiff cannot read the headers of an NDPI file. This is because NDPI specifies the RowsPerStrip as the height of the file, and after doing out the multiplication, this typically overflows libtiff and it refuses to open the file. Also, the TIFF tags are not stored in sorted order.

NDPI stores an image pyramid in TIFF directory entries. In some files, the lower-resolution pyramid levels contain no restart markers. The macro image, and sometimes an active-region map, seems to come last.

JPEG files in NDPI are not necessarily valid. If ImageWidth or ImageHeight exceeds the JPEG limit of 65535, then the width or height as stored in the JPEG file is 0. libjpeg will refuse to read the header of such a file, so the JPEG data stream must be altered when fed into libjpeg.

Here are the observed TIFF tags:

Optimisation File (only for VMS)

The optimisation file contains a list of 32- (or 64- or 320- ?) bit little endian values, giving the file offset into an MCU row, each offset starts at a 40-byte alignment, and the last row (of the entire file, not each image) seems to be missing. The offsets are all packed into 1 file, even with multiple images. The order of images is left-to-right, top-to-bottom.

Map File (only for VMS/VMU)

The VMS map file is a standard JPEG file. Its restart markers (if any) are not included in the optimisation file. The VMU map file is in NGR format. This file can be used to provide a lower-resolution view of the slide.

Image Files (only for VMS/VMU)

These files are given by the VMS/VMU ImageFile keys. They are assumed to have a height which is a multiple of the MCU height. They are assumed to have a width which is a multiple of MCUs per row divided by the restart interval.

For VMS, these files are in JPEG, for VMU they are in NGR format.

NGR Format

The NGR file contains uncompressed 16-bit RGB data, with a small header. The files we have encountered start with GN, two more bytes, and then width, height, and column width in little endian 32-bit format. The column width must divide evenly into the width. Column width is important, since NGR files are generated in columns, where the first column comes first in the file, followed by subsequent files. Columns are painted left-to-right.

At offset 24 is another 32-bit integer which gives the offset in the file to the start of the image data. The image data we have encountered is in 16-bit little endian format.

Associated Images

macro

the image file given by the MacroImage value in the VMS/VMU file, or SourceLens of -1 in NDPI

Known Properties

All key-value data stored in the VMS/VMU file, and known tags from the NDPI file, are encoded as properties prefixed with "hamamatsu.".

openslide.mpp-x

for NDPI, calculated as 10000/tiff.XResolution, if tiff.ResolutionUnit is centimeter

openslide.mpp-y

for NDPI, calculated as 10000/tiff.YResolution, if tiff.ResolutionUnit is centimeter

openslide.objective-power

normalized hamamatsu.SourceLens

Test Data

NDPI format

m[blue]http://openslide.cs.cmu.edu/download/openslide-testdata/Hamamatsu/m[]

VMS format

m[blue]http://openslide.cs.cmu.edu/download/openslide-testdata/Hamamatsu-vms/m[]

APERIO FORMAT

Format

single-file pyramidal tiled TIFF, with non-standard metadata and compression

File extensions

.svs, .tif

OpenSlide vendor backend

aperio

Vendor Documentation

m[blue]http://www.aperio.com/documents/api/Aperio_Digital_Slides_and_Third-party_data_interchange.pdfm[]

Detection

Aperio slides are stored in single-file TIFF format. OpenSlide will detect a file as Aperio if:

1. The file is TIFF.

2. The initial image is tiled.

3. The ImageDescription tag starts with Aperio.

Relevant TIFF tags

Extra data stored in ImageDescription

For tiled images, the ImageDescription tag contains some dimensional downsample information as well as what look like offsets. Additionally, vertical line-delimited key-value pairs are stored, in at least the full-resolution image. A key-value pair is equals-delimited. These key-values are stored as properties starting with "aperio.". Currently, OpenSlide does not use any of the information present in these key-value fields.

For stripped images, the ImageDescription tag may contain a name, followed by a carriage return. This is used for naming the associated images. The second image in the file does not have a name, though it is an associated image.

TIFF Image Directory Organization

m[blue]http://www.aperio.com/documents/api/Aperio_Digital_Slides_and_Third-party_data_interchange.pdfm[] page 14:

The first image in an SVS file is always the baseline image (full resolution). This image is always tiled, usually with a tile size of 240x240 pixels. The second image is always a thumbnail, typically with dimensions of about 1024x768 pixels. Unlike the other slide images, the thumbnail image is always stripped. Following the thumbnail there may be one or more intermediate "pyramid" images. These are always compressed with the same type of compression as the baseline image, and have a tiled organization with the same tile size.

Optionally at the end of an SVS file there may be a slide label image, which is a low resolution picture taken of the slide's label, and/or a macro camera image, which is a low resolution picture taken of the entire slide. The label and macro images are always stripped.

JPEG 2000 (compression types 33003 or 33005)

Some Aperio files use compression type 33003 or 33005. Images using this compression need to be decoded as a JPEG 2000 codestream. For 33003: YCbCr format, possibly with a chroma subsampling of 4:2:2. For 33005: RGB format. Note that the TIFF file may not encode the colorspace or subsampling parameters in the PhotometricInterpretation field, nor the YCbCrSubsampling field, even though the TIFF standard seems to require this. The correct subsampling can be found in the JPEG 2000 codestream.

Associated Images

thumbnail

the second image in the file

label

optional, the name "label" is given in ImageDescription

macro

optional, the name "macro" is given in ImageDescription

Known Properties

All key-value data encoded in the ImageDescription TIFF field is represented as properties prefixed with "aperio.".

openslide.mpp-x

normalized aperio.MPP

openslide.mpp-y

normalized aperio.MPP

openslide.objective-power

normalized aperio.AppMag

Test Data

m[blue]http://openslide.cs.cmu.edu/download/openslide-testdata/Aperio/m[]

MIRAX FORMAT

Format

multi-file with very complicated proprietary metadata and indexes

File extensions

.mrxs

OpenSlide vendor backend

mirax

Detection

OpenSlide will detect a file as MIRAX if:

1. The file is not a TIFF.

2. The filename ends with .mrxs.

3. A directory exists in the same location as the file, with the same name as the file minus the extension.

4. A file named Slidedat.ini exists in the directory.

Overview

MIRAX can store slides in JPEG, PNG, or BMP formats. Because JPEG does not allow for large images, and JPEG and PNG provide very poor support for random-access decoding of part of an image, multiple images are needed to encode a slide. To avoid having many individual files, MIRAX packs these images into a small number of data files. The index file provides offsets into the data files for each required piece of data.

The camera on MIRAX scanners takes overlapping photos and records the position of each one. Each photo is then split into multiple images which do not overlap. Overlaps only occur between images that come from different photos.

To generate level n + 1, each image from level n is downsampled by 2 and then concatenated into a new image, 4 old images per new image (2 x 2). This process is repeated for each level, irrespective of image overlaps. Therefore, at sufficiently high levels, a single image can contain one or more embedded overlaps of non-integral width.

Index File

The index file starts with a five-character ASCII version string, followed by the SLIDE_ID from the slidedat file. The rest of the file consists of 32-bit little-endian integers (unaligned), which can be data values or pointers to byte offsets within the index file.

The first two integers point to offset tables for the hierarchical and nonhierarchical roots, respectively. These tables contain one record for each VAL in the HIERARCHICAL slidedat section. For example, the record for NONHIER_1_VAL_2 would be stored at nonhier_root + 4 * (NONHIER_0_COUNT + 2).

Each record is a pointer to a linked list of data pages. The first two values in a data page are the number of data items in the page and a pointer to the next page. The first page always has 0 data items, and the last page has a 0 next pointer.

There is one hierarchical record for each zoom level. The record contains data items consisting of an image index, offset and length within a file, and a file number. The file number can be converted to a data file name via the DATAFILE slidedat section. The image index is equal to image_y * GENERAL.IMAGENUMBER_X + image_x. Image coordinates which are not multiples of the zoom level's downsample factor are omitted.

Nonhierarchical records refer to associated images and additional metadata. Nonhierarchical data items consist of three zero values followed by an offset, length, and file number as in hierarchical records.

Data Files

A data file begins with a header containing a five-character ASCII version string, the SLIDE_ID from the slidedat file, the file number encoded into three ASCII characters, and 256 bytes of padding. The remainder of the file contains packed data referenced by the index file.

Slide Position File

The slide position file is referenced by the VIMSLIDE_POSITION_BUFFER.default nonhierarchical section. It contains one entry for each camera position (not each image position) in row-major order. Each entry is nine bytes: a flag byte, the X pixel coordinate of the photo (4 bytes, little-endian, may be negative), and the Y coordinate (4 bytes, little-endian, may be negative). In slides with CURRENT_SLIDE_VERSION ≥ 1.9, the flag byte is 1 if the slide file contains images for this camera position, 0 otherwise. In older slides, the flag byte is always 0.

In slides with CURRENT_SLIDE_VERSION ≥ 2.2, the slide position file is compressed with DEFLATE and referenced by the StitchingIntensityLayer.StitchingIntensityLevel nonhierarchical section.

Associated Images

thumbnail

the image named "ScanDataLayer_SlidePreview" in Slidedat.ini (optional)

label

the image named "ScanDataLayer_SlideBarcode" in Slidedat.ini (optional)

macro

the image named "ScanDataLayer_SlideThumbnail" in Slidedat.ini (optional)

Known Properties

All key-value data stored in the Slidedat.ini file are encoded as properties prefixed with "mirax.".

openslide.mpp-x

normalized MICROMETER_PER_PIXEL_X from the Slidedat section corresponding to level 0 (typically mirax.LAYER_0_LEVEL_0_SECTION.MICROMETER_PER_PIXEL_X)

openslide.mpp-y

normalized MICROMETER_PER_PIXEL_Y from the Slidedat section corresponding to level 0 (typically mirax.LAYER_0_LEVEL_0_SECTION.MICROMETER_PER_PIXEL_Y)

openslide.objective-power

normalized mirax.GENERAL.OBJECTIVE_MAGNIFICATION

See Also

m[blue]Introduction to MIRAX/MRXSm[][3]. Note that our terminology has changed since that document was written; where it says "tile", substitute "image", and where it says "subtile", substitute "tile".

Test Data

m[blue]http://openslide.cs.cmu.edu/download/openslide-testdata/Mirax/m[]

SAKURA FORMAT

Format

SQLite database containing pyramid tiles and metadata

File extensions

.svslide

OpenSlide vendor backend

sakura

Detection

OpenSlide will detect a file as Sakura if:

1. The file is a SQLite database.

2. The DataManagerSQLiteConfigXPO table contains exactly one row, and its TableName field refers to a unique table.

3. The unique table contains a row with id = "++MagicBytes" and data = "SVGigaPixelImage".

File Organization

Sakura slides are SQLite 3 database files written by the eXpress Persistent Objects ORM. Tables contain slide metadata, associated images, and JPEG tiles. Tiles are addressed as (downsample, level-0 X coordinate, level-0 Y coordinate, color channel), with separate grayscale JPEGs for each color channel. Despite the generality of the address format, tiles appear to be organized in a regular grid, with power-of-two level downsamples and without overlapping tiles. The structure of the file allows scans to be sparse, but it is not clear if this is actually done.

SQL Tables

Some irrelevant tables and columns have been omitted from the summary below. DataManagerSQLiteConfigXPO.PP Useful only to get a reference to the unique table. OpenSlide requires this table to contain exactly one row.

SVSlideDataXPO.PP High-level metadata about a slide. OpenSlide assumes this table will contain exactly one row.

SVHRScanDataXPO.PP A single high-resolution scan of a slide from SVSlideDataXPO. OpenSlide assumes this table will contain exactly one row.

SVScannedImageDataXPO.PP Contains associated images other than the thumbnail.

tile.PP This table is most naturally used to map tile coordinates to tile IDs, but is not suitable for individual lookups because it has no useful indexes.

Unique table

This is the table named by DataManagerSQLiteConfigXPO.TableName. It contains named blobs including the JPEG tile data.

This table stores a variety of blob types. IDs for image tiles appear to be mechanically generated from the tile coordinates, but OpenSlide does not construct them directly. Instead, it uses the tile table to obtain IDs for each tile.

Associated Images

label

SVScannedImageDataXPO.Image corresponding to SVSlideDataXPO.m_labelScan

macro

SVScannedImageDataXPO.Image corresponding to SVSlideDataXPO.m_overviewScan

thumbnail

SVHRScanDataXPO.ThumbnailImage

Known Properties

sakura.Creator

SVSlideDataXPO.Creator

sakura.Date

SVSlideDataXPO.Date

sakura.Description

SVSlideDataXPO.Description

sakura.DiagnosisCode

SVSlideDataXPO.DiagnosisCode

sakura.FocussingMethod

SVHRScanDataXPO.FocussingMethod

sakura.Keywords

SVSlideDataXPO.Keywords

sakura.NominalLensMagnification

SVHRScanDataXPO.NominalLensMagnification

sakura.ResolutionMmPerPix

SVHRScanDataXPO.ResolutionMmPerPix

sakura.ScanId

SVHRScanDataXPO.ScanId

sakura.SlideId

SVSlideDataXPO.SlideId

openslide.mpp-x

calculated as 1000 * sakura.ResolutionMmPerPix

openslide.mpp-y

calculated as 1000 * sakura.ResolutionMmPerPix

openslide.objective-power

normalized sakura.NominalLensMagnification

Test Data

No public data available. Contact the m[blue]mailing listm[][4] if you have some.

GENERIC TILED TIFF FORMAT

Format

single-file pyramidal tiled TIFF

File extensions

.tif

OpenSlide vendor backend

generic-tiff

Detection

OpenSlide will detect a file as generic TIFF if:

1. No other detections succeed.

2. The file is TIFF.

3. The initial image is tiled.

TIFF Image Directory Organization

The first image in the TIFF file is the full-resolution image. Any other tiled images in the file with the "reduced resolution" bit set are assumed to be reduced-resolution versions of the original.

Associated Images

None.

Known Properties

Many TIFF tags are encoded as properties starting with "tiff.".

AUTHORS

The Carnegie Mellon School of Computer Science.

This manual page was written by Mathieu Malaterre <[email protected]> for the Debian GNU/Linux system (but may be used by others).

NOTES

1.

Supported Virtual Slide Formats

http://openslide.org/formats/

2.

issue 21

http://openslide.orghttps://github.com/openslide/openslide/issues/21#issuecomment-23615583

3.

Introduction to MIRAX/MRXS

http://lists.andrew.cmu.edu/pipermail/openslide-users/2012-July/000373.html

4.

mailing list

http://lists.andrew.cmu.edu/mailman/listinfo/openslide-users/