This paper presents an alternative technique of hole-filling method in pixel nearest neighbour for 3D volume reconstruction, in order to reconstruct the voxels that are not captured by the medical imaging scanner such as ultrasound. However, many existing works claim that the pixel nearest neighbour may blur the reconstruction results especially the boundary of the anatomy. Thus, a hole-filling method based on the modified butterfly interpolation scheme is proposed. The experiments are conducted on the ultrasound and CT scan data sets. The artificial holes are simulated on these data sets in order to test the capability of various hole-filling methods, such as mean, median, Olympic, oriented sticks operation, and the proposed method. The evaluation is made based on the root mean square error (RMSE) and mean absolute error (MAE) by the comparison between original data and the reconstructed data. The visualization of the results is also conducted. The result shows that the proposed method is robust and efficient in reconstructing the missing voxel value in the ultrasound and CT scan data sets.
The full article can be downloaded at https://ieeexplore.ieee.org/abstract/document/8980511
