Corresponding Paper:
"2.5D Thermometry Mapsfor MRI-guided Tumor Ablation" by Alpers and Reimert et al.

Data Sets:
The data suite consists of three different data types:
1) PerfusionPhantom_1-6 correspond to the test data simulating a possible heat sink effect by including a PVC tube inside of the phantom.
2) Phantom_1-5 correspond to the test data without additional tubes or other structures. These phantoms show a homogeneous necrosis growing over the time of ablation.
3) TempPhantom_1-2 are equal to the data mentioned under 2) but these two phantoms also included two temperature sensors each to verify the temperature accuracy exemplarily.

Data Structure:
Every data set consists of the following four subfolder:
1) "RawData" includes the magnitude and phase images for the differen rotation angles 0°, 22.5°, 45°, 67.5°, 90°, 112.5°, 135° and 167.5°.
2) "HeatMap" includes the computed heat maps for every rotation angle and every time point according to the proton resonance frequency shift methods with image "0.IMA" always beeing the reference image.
3) "Reconstruction" includes the reconstruction result given by our algorithm for the end of the ablation taking into acount the latest time points from every rotation angle. The algorithm is available for download at https://github.com/jalpers/2.5DThermometryReconstruction
4) "GroundTruth" includes the manually annotated necrosis based on the corresponding "PostTreatment" data set

Results Excel:
The excel sheet "Results.xlsx" shows the result from our evaluation using every dataset provided. Computation of the Dice Scroe was performed according to the paper of "Popovic, A., De la Fuente, M., Engelhardt, M., & Radermacher, K. (2007). Statistical validation metric for accuracy assessment in medical image segmentation. International Journal of Computer Assisted Radiology and Surgery, 2(3), 169-181." 

Data Observations:
1) Phantom_1 is missing the magnitude images due to a technical problem during image acquisition
2) PerfusionPhantom_1 and PerfusionPhantom_2 show strong MR inhomogeneities resulting in bad reconstructions of the volumetric heat map.
3) All other data sets show a usual signal-to-noise ratio.