A guide to performance evaluation algorithm (real measurement)

• Guide
• data analysis
• drift analysis
• image analysis

The examples and demos show how the algorithms for the evaluation of EIT performance based on pre-measured data. Simulated demo is also provided for image analysis in 2D and 3D (Detectability and GREIT parameters) in next section. To evaluate any EIT system, the user can define what type of evaluation they want to do after placing the measured data to a proper directory or modifying the directory locations in the code. Simulated tutorial for image analysis is provided in next section.

Download the algorithms in a zip file and then unzip it.

Main folders: EIT_system_evaluation consists of following 4 subfolders: code, data, documents and figures. Data repository: We created folders to store the data under the data folder which include accuracy, drift, detectability (1 object) and distinguishability (2 objects) subfolders.

Guide to evaluation algorithm:

1. data_analysis.m
2. drift_analysis.m
3. image_analysis.m

Note:drift analyses takes long time to to be processed (~20-30 min depending on the PC specification) so a separate sub-function is provided out of data analysis algorithm. The algorithm is provided for data and drift analysis below with pre-measured data. The measures for the accuracy and symmetric electrode measurement calculated based on mean values of multiple measurements for each position.

data_analysis.m includes

• SNR
• Accuracy
• Reciprocity Accuracy
• File name:data_analysis.m
• Usage:[snr, accuracy,sym_err]= data_analysis(Reshape);  For example,[snr, accuracy,sym_err]= data_analysis(0)

Reshape is a status/input parameter used to check if reshape of signal from U-shape to plateau signal shape is needed or not - 0 not needed, 1 needed. To perform data analysis, the measurement sequence may need to be rearranged to be consistent with the presentation of the results. Measured and simulated voltages need to have plateau signal shape form for ease of analysis as figure (A) (to get plateau signal shape). If the voltage sequence is based on U-shape signal (the Sheffield measurement sequence or the sequence of measurement followed by the injecting electrodes) as figure (B) or  , the algorithm will reshape the signal using U2plateau.m.

 

Figure (A): Plateau shaped signal,(B) U-shape signal



Figure: (A) SNR values, (B) Accuracy, (C) symmetric measurement error

Test of drift

• File name: drift_analysis.m
• Usage:[Df]=drift_analysis

Users have to check several parameters if it matches what they measured to the default values that are defined in the code like number of frames, number of data points for each frame, and measurement sequences. Input: nframes is the number of frames for drift measurement; nmeas is number of measurements for each frame (mostly 208, in some cases 416). The output value is the values of Allan variance.

Image analysis

• File name : image_analysis.m
• Usage: [DET, greit_para] = image_analysis, this function computes and plots following measures as output:
  • Detectability
  • Distinguishability
  • Amplitude response (AR)
  • Position error (PE)
  • Resolution (RES)
  • Shape deformation (SD)
  • Ringing (RNG)

To have a measurement sequence with a plateau signal shape please look at Figure 10 of  paper source: Gagnon H, Cousineau M, Adler A, Hartinger A 2010 A resistive mesh phantom for assessing the performance of EIT systems {\em IEEE T Biomed. Eng.} 57 (9):2257--2266, 2010.

• U2plateau.m converts Sheffield voltage measurement sequence (input) to plateau shape (output) described above.
• Usage: [Vp]= U2plateau(V)
• V should be based on U-shape (Sheffield voltage measurement sequence or the measurement sequence followed by the injecting electrodes), produce plateau shaped wave form (Vp)

Last Modified: $Date: 2017-02-28 13:12:08 -0500 (Tue, 28 Feb 2017) $ by $Author: aadler $