Image reconstruction by multivariate kernel density estimation

The application, written in R language, is devoted to image reconstruction for J-PET, using multivariate kernel density estimation (KDE).

Prerequisites

Tested on Ubuntu 16.04 LTE with R version 3.4.4 installed. The script operates with a number of shell commands (for example, readlink -f <incomplete_path_to_file>), but. basically, all of them are standard.

Additional R packages are required:

• jsonlite
• data.table
• ks
• reshape2 (for ASCII format only)
• RColorBrewer (optional - for plotting images)

Input data and parameters

The application proceeds with GOJA output format (16-column ASCII) as the input. The parameters of the reconstruction, as well as the geometry of the scanner, are stored in JSON format: it is self-describing and easy to undersand. Please refer to the example/ directory for the details.

Script architecture

The script is designed to work as quick as possible, hence, for data processing, R environments are utilised instead of classes:

• SetMappingEnvironment.R
• SetAnhPtsEnvironment.R
• SetKDEEnvironment.R

In comparison, the performance is not crucial for the read of input parameters, so Reference classes are used for controllers (wrapped into a list by InitParamsFromJSON.R):

• JSONReader.R
• KDEController.R

More information on the performance of R classes and environments could be found here.

Usage

There are three main executables:

• MapHitsToCentresOfStrips.R - remaps all XY-coordinates of hits to the centres of strips and saves the result to file, according to the format chosen (ASCII or .RData). Only first 8 columns are saved (pairs of coordinates and times of hits)
• DetectAnnihilationPoints.R - remaps XY-coordinates of hits to the centres of strips and estimates exact positions of annihilation points, using time-of-flight (estimated from times of hits). The result - three columns for (X,Y,Z) - is stored according to the format chosen (ASCII or .RData).
• EngageKDEReconstruction.R - executes KDE on the data for annihilation points. Depends strongly on input JSON file and requires DetectAnnihilationPoints.R to run first.

A single shared argument, a .json file with parameters is passed to all executables (see the example/ directory):

$ Rscript [--vanilla] MapHitsToCentresOfStrips.R <parameters.json>
$ Rscript [--vanilla] DetectAnnihilationPoints.R <parameters.json>
$ Rscript [--vanilla] EngageKDEReconstruction.R <parameters.json>

The option --vanilla prevents Rscript from reading R history, profile, or environment files, as well as reloading data or objects from previous sessions.