11.2 The DICOM classes

The DicomElement class is a simple container for DICOM attribute information. It has three main properties:

• •

  an integer tag

• •

  a value representation (VR)

• •

  a value

These properties can be obtained using the corresponding get function: getTag(), getVR(), getValue(). The tag refers to the concatenated group/element tag. For example, the transfer syntax UID which corresponds to group 0x0002 and element 0x0010 has a numeric tag of 0x00020010. The VR is represented by an enumerated type, DicomElement.VR. The ‘value’ is the raw value extracted from the DICOM file. In most cases, this will be a String. For raw numeric values (i.e. stored in the DICOM file in binary form) such as the unsigned short (US), the ‘value’ property is exactly the numeric value.

For VRs such as the integer string (IS) or decimal string (DS), the string will still need to be parsed in order to extract the appropriate sequence of numeric values. There are static utility functions for handling this within DicomElement. For a ‘best-guess’ of the desired parsed value based on the VR, one can use the method getParsedValue(). Often, however, the desired value is also context-dependent, so the user should know a priori what type of value(s) to expect. Parsing can also be done automatically by querying for values directly through the DicomHeader object.

When a DICOM file is parsed, all meta-data (attributes apart from the actual pixel data) is assembled into a DicomHeader object. This essentially acts as a map that can be queried for attributes using one of the following methods:

The first method returns the full element as described in the previous section. The remaining methods are used for convenience when the desired value type is known for the given tag. These methods automatically parse or convert the DicomElement’s value to the desired form.

If the tag does not exist in the header, then the getIntValue(...) and getDecimalValue(...) will return the supplied defaultValue. All other methods will return null.

The DicomPixelBuffer contains the decoded image information of an image slice. There are three possible pixel types currently supported:

• •

  byte grayscale values (PixelType.BYTE)

• •

  short grayscale values (PixelType.SHORT)

• •

  byte RGB values, with layout RGBRGB...RGB (PixelType.BYTE_RGB)

The pixel buffer stores all pixels in one of these types. The pixels can be queried for directly using getPixel(idx) to get a single pixel, or getBuffer() to get the entire pixel buffer. Alternatively, a DicomPixelInterpolator object can be passed in to convert between pixel types via one of the following methods:

These methods populate an output array or buffer with converted pixel values, which can later be passed to a renderer. For further details on these methods, refer to the Javadoc documentation.

A single DICOM file contains both header information, and one or more image ‘frames’ (slices). In ArtiSynth, we separate each frame and attach them to the corresponding header information in a DicomSlice. Thus, each slice contains a single DicomHeader and DicomPixelBuffer. These can be obtained using the methods: getHeader() and getPixelBuffer().

For convenience, the DicomSlice also has all the same methods for extracting and converting between pixel types as the DicomPixelBuffer.

An complete DICOM acquisition typically consists of multiple slices forming a 3D image stack, and potentially contains multiple 3D stacks to form a dynamic 3D+time image. The collection of DicomSlices are thus assembled into a DicomImage, which keeps track of the spatial and temporal positions.

The DicomImage is the main object to query for pixels in 3D(+time). To access pixels, it has the following methods:

The inputs {x, y, z} refer to voxel indices, and time refers to the time instance index, starting at zero. The four voxel dimensions of the image can be queried with: getNumCols() getNumRows(), getNumSlices(), and getNumTimes().

The DicomImage also contains spatial transform information for converting between voxel indices and patient-centered spatial locations. The affine transform can be acquired with the method getPixelTransform(). This allows the image to be placed in the appropriate 3D location, to correspond with any derived data such as segmentations. The spatial transformation is automatically extracted from the DICOM header information embedded in the files.