Kuopio Tomography Challenge 2023 open electrical impedance tomographic dataset (KTC 2023)

Notice: v3 update 17.4.2023. Ground truth images were corrected slightly. The older ground truth images had a geometric distortion caused by cropping the photographs to the outer boundary of the imaging chamber, instead of the inner boundary on which the measurement electrodes were located. In addition, version 3 now contains both the limited EIT data used in the challenge evaluation, as well as the full EIT data with no measurements left out.

It is recommended to download version v3 and replace the v2 version with it if using the data for research purposes.

This dataset is primarily designed for the Kuopio Tomography Challenge 2023 (KTC 2023), but it can be used for generic algorithm research and development in 2D Electric Impedance Tomography (EIT) image reconstruction. The purpose of KTC 2023 is to develop algorithms for electric impedance tomography image reconstruction with limited data.

The dataset contains electric impedance tomography data collected from phantoms made up of a cylindrical water chamber with inclusions of varying shapes and sizes present. The data consists of electrical measurements collected using electrodes placed at the chamber boundary, i.e. the values of electric current injected between two electrodes at a time, and the resulting voltages between adjacent electrode pairs. The data has been stored in the MATLAB .mat format.

The training dataset contains data measured from five training phantoms. These are designed to facilitate algorithm development and benchmarking for the challenge itself. Four of the training phantoms contained inclusions, which were either resistive or conductive in comparison to the water in the imaging chamber. The fifth training phantom was the imaging chamber filled with only water. We encourage subsampling these datasets to create limited datasets and comparing the reconstruction results to the ground truth obtainable from the full electrode data.

The actual challenge data is arranged into seven different difficulty levels, labeled 1-7, with each level containing three different phantoms, labeled A-C. As the difficulty level increases, the number of inclusions increases and their shapes become increasingly complex. Furthermore, the measurement data is made more limited by removing measurements collected by some of the electrodes. At each difficulty level above level 1, the data collected from two of the boundary electrodes is removed.

To illustrate a solution to the KTC2023, we have included a simple example reconstruction algorithm, and a Finite Element Method based forward solver used by the reconstruction algorithm. These are provided as both Matlab and Python codes. The reconstructions were computed using linearized difference imaging, and the resulting (EIT) images were then segmented using Otsu's method.

The competitors do not need to follow the above procedure, and are encouraged to explore various image reconstruction and segmentation techniques.

The full data for all the test phantoms will be released after the Kuopio Tomography Challenge 2023 has ended. This data has now been added to this dataset as EvaluationData.zip.

All measurements were conducted at the Process Tomography Laboratory at the University of Eastern Finland.

注意事项：本数据集v3版本更新于2023年4月17日。本次对真值图像（ground truth）进行了小幅修正。旧版真值图像存在几何畸变问题，其成因是将拍摄图像裁剪至成像腔的外边界，而非测量电极所在的内边界。此外，v3版本同时包含了挑战赛评估所用的受限电阻抗断层成像（Electric Impedance Tomography, EIT）数据，以及未遗漏任何测量项的完整EIT数据。

若将本数据集用于科研工作，建议下载v3版本并替换原有v2版本。

本数据集主要为2023年库奥皮奥断层成像挑战赛（Kuopio Tomography Challenge 2023, KTC 2023）设计，同时也可用于二维电阻抗断层成像（EIT）图像重建的通用算法研发。KTC 2023的核心目标是研发面向受限数据场景的EIT图像重建算法。

本数据集包含从仿体采集的EIT数据，该类仿体由圆柱形水腔构成，内部包含形状、尺寸各异的内含物。数据通过置于腔室边界的电极采集得到，具体包括：单次在两个电极间注入的电流值，以及相邻电极对间产生的电压值。所有数据以MATLAB .mat格式存储。

训练数据集包含5个训练仿体的测量数据，旨在为挑战赛的算法开发与性能基准测试提供支持。其中4个训练仿体带有内含物，相较于腔室内的纯水，其导电率分别呈阻性或导性；第5个训练仿体仅填充纯水。我们建议对该数据集进行下采样以构建受限数据集，并将重建结果与完整电极数据可获取的真值图像进行对比。

挑战赛正式数据分为7个难度等级（标记为1至7），每个等级包含3个不同的仿体（标记为A至C）。随着难度等级提升，仿体内含物的数量逐渐增加，形状也愈发复杂；同时，通过移除部分电极的采集数据进一步受限化数据集。在等级1以上的所有难度中，均会移除2个边界电极的采集数据。

为展示KTC2023的一种解决方案，本数据集附带了一款简易的图像重建示例算法，以及该算法所用的基于有限元法（Finite Element Method, FEM）的前向求解器，同时提供了MATLAB与Python两种语言的实现代码。重建过程采用线性化差分成像方法，最终得到的EIT图像通过大津法（Otsu's method）进行分割。

参赛选手无需遵循上述流程，欢迎探索各类图像重建与分割技术。

所有测试仿体的完整数据将于2023年库奥皮奥断层成像挑战赛结束后公开，目前该部分数据已以EvaluationData.zip的形式加入本数据集。

所有测量工作均在东芬兰大学过程断层成像实验室完成。