Use Cases

Leadership: Nicolas A Karakatsanis, Weill Cornell Medical College, New York, NY, USA

The “Use Cases” subgroup is responsible for developing using the Yardl generated code in different languages (C++, Python, etc.) software tools to read, process, and write raw data files of the proposed standardized format.

In the ETSIhacker’s GitHub public repository we share a set of software tools developed during our hackathon events demonstrating a set of practical use-cases for the data standard including:

• conversion of simulated or real list-mode data in the standard format (e.g. from Monte-Carlo simulations or vendor scanners)
• basic list-mode data operations (e.g. subsampling, gating etc.),
• interfaces linking the data standards with existing open-source image reconstruction software (e.g. CASToR, PyTomography, STIR etc.), and
• visualization of the scanner geometry description embedded in the data standard

Examples of such usage tools range from extraction of basic statistics and features of the raw data, to histogramming and precorrection processes and creation of interfaces to feed the raw data into basic image reconstruction algorithms implemented in open source software packages, such as CASToR, NiftyPET, and STIR.

More specific “use case” software tool examples are indicated below:

• extraction of info from the standardized list data, such as numbers or histograms of different types of coincidence events (prompts, randoms) and noise-equivalent counts corresponding to user-defined energy windows, acquisition times, Time-of-Flight bins, detector pairs of combinations thereof.
• extraction of info from the standardized histogram (a.k.a. sinogram) data, such as order of sinogram segments/TOF bins, applied “compression” (span/mash factors), maximum ring difference and plane indexing per segment
• replacement of detector pair info with a detector pair bin address of every event in the list data or vice-versa
• binning of the different types of listed events to histograms/sinograms according to a user-defined time frame sequence, or energy window or TOF bin range or combinations thereof
• splitting, concatenating or gating of list data streams (optionally using tags to drive those operations)
• preprocessing of list-mode or sinogram/histogram data to become compatible to open-source FBP, ML-EM and OS-EM image reconstruction with existing open-source libraries (CASToR, NiftyPET, STIR etc.)
• performance of basic math operations on histograms to support pre-corrections of raw data before reconstruction such as scalar/element-wise addition, subtraction, multiplication, division and exponential operations (e.g. for building attenuation correction histograms)
• compressing or mashing histograms/sinograms or applying the single-slice rebinning (SSRB) operation to aggregate all raw data to direct planes
• resorting of sinogram segments or TOF bins in different orders and automatically determining that order on existing sinograms
• expansion from a small set of normalization components to a full normalization correction sinogram
• conversion of simulated/synthetic emission output list-mode data (e.g. in ROOT format) or histograms/sinograms produced with analytic forward projectors (e.g. “STIR”, “SMART”, “PET Sim & Recon” etc.) or Monte Carlo simulations (e.g. GATE, SimSET, SIMIND etc.) to the standardized PET raw data format
• development of conversion tools, in collaboration with scanner manufacturers, to attain full compatibility between the proposed standardized data format and vendors proprietary data formats. Such tools may not necessarily be open source if the vendors wish to protect intellectual property associated with their proprietary formats, but they would need to be freely available to anyone wishing to convert to/from vendors’ proprietary formats from/to the standardized data format.
• construction of scanner geometric representations (.stl/.obj data formats) from embedded/header info in the raw data and vice versa
• 3D viewing of PET event histograms (per plane or per segment)

In addition, our emission tomography raw data standards including the model data elements and the associated software development kits (SDKs) can be found at ETSI’s GitHub public repository.