The Zegami Machine Learning Suite is centered around images. Our image exploration platform has always supported the most common image formats including .jpg, .png, .gif and .bmp. However, our customers often work with a variety of images stored in other formats. We seek to make importing visual data as straightforward as possible. Therefore, while it’s usually possible to convert these formats to .png before uploading, we’ve extended our platform over time to provide built-in support for a number of different formats.
This built-in support brings multiple benefits:
- Raw data files can be uploaded with no need to convert to a standard format
- Automatic extraction through the Zegami platform of any non-image metadata embedded within files, and adding to the collection data
- Source files can be downloaded from Zegami for further investigation after identifying images of particular interest
What image formats are supported by Zegami?
DICOM (Digital Imaging and Communications in Medicine)
DICOM is a widespread standard for storing and exchanging medical imaging data. Rather than being a single, homogenous format, DICOM covers a range of different encodings and topologies, including 3D and time sequences.
Zegami is able to consume static 2D DICOM images with ease since they map naturally to the 2D viewing interface. However, many images come with 3D data (MRI, CT), and some capture sequences of images. Images with these extra dimensions – temporal or spatial – can be automatically broken apart into the constituent frames or slices. These slices can be added to a single main body of images, or alternatively treated as separate sources.
DICOM files often include metadata fields which would be of use for filtering and graphing within the Zegami platform. Our processing engine is able to extract these columns and add them to the row data that accompanies each image.
TIFF (or TIF)
A TIFF is a flexible container that can contain pixel data in several formats, with various compression techniques.
Most TIFF files are accepted on the Zegami platform, and some common features that are important to many scientific applications are also supported:
- 16-bit colour depth. In a 16-bit images, each pixel is represented by a 16-bit value rather than the typical 8 bits. This format enables recording precise values across a wide spectrum of brightness values. In Zegami we convert these images to 8 bits, losing the extra precision, but the conversion can be configured to use a specific range window.
- Layered images. TIFFs can contain a stack of images each at the same resolution but carrying different information. Typically, this might represent images captured for different optical wavelengths, or may represent slices of a third spatial dimension. In this instance the layers can be automatically broken apart into separate image sources.
Whole Slide Images (WSI)
Whole Slide Images are a family of related formats which encapsulate a set of high-resolution microscopic images covering the entirety of a microscope slide. Formats include .svs, .scn and .mrxs, and many are also packaged as TIFs. These images typically cover an area with an extremely high number of pixels (e.g 200,000 x 100,000) and are technically composites of multiple smaller images.
Instead of storing the entire image as a single enormous map of pixel values, WSI formats store multiple different scaled resolutions, each as a set of tiles, to facilitate viewing the image via zooming and panning. This arrangement is often referred to as a pyramid: At the tip is an image encompassing the entire slide, but at a relatively low resolution; in effect a thumbnail. The next level down is double the resolution but is separated into quadrants so made up of 4 distinct images or ‘slices’. The next, four times the thumbnail resolution, with 16 slices. This carries on until we have a set of many tiles at the original capture resolution.
WSI images come in a variety of vendor-specific formats, but Zegami can ingest many of these to be processed for viewing in our visualisation platform.
There are many formats in existence beyond those described her. Although we don’t currently support every format out there, our flexible processing engine means it’s relatively easy for us to add direct support for new formats. If you’d like to find out about how Zegami can handle your image data, don’t hesitate to get in touch.