Accuracy assessment of method for 3D reconstruction of phantoms based on MRI images

As a part of treatment planning of solid, deep-seated tumors with electrochemotherapy an algorithm that reconstructs a 3D model of a tumor has been developed. The model is calculated from MRI images. In order to verify the accuracy of the reconstructed model and assess the quality of the method it was necessary to use an object whose dimensions can be physically measured. To conduct this examination a commercial cylindrical phantom with constant diameter of 162 mm made of plastic material was chosen. The phantom was filled with special MRI sensitive liquid. Scanning of the phantom with MRI device gave a collection of 134 MRI images which were used as the entry data set for algorithm for 3D model reconstruction. For each reconstructed slice number of pixels in that slice was calculated. Based on pixel size and known phantom diameter, areas of each slice in 3D model were calculated, expressed in mm², along with actual referential area of the phantom. Results of the calculation were then statistically analyzed giving the maximum absolute error of 3.52%. The calculated error is in fact a spatial geometric distortion composed from image segmentation error and reconstruction error from segmented data.