Three-dimensional reconstruction and 3D printing of kidney from computed tomography

This paper presents a novel system to reconstruct 3D kidney structure from CT images. Before reconstruction, the kidney region should be well segmented from each CT image. This paper presents a deep learning method to segment each kidney region roughly from the CT image as initial starting points to guide a contour tracking to refine its final boundaries. However, due to the higher radiation risk from CT, a patient cannot be scanned densely so that the resolution of CT images in the Z-axis is not good enough for 3D reconstruction; that is, the distance between layers is larger than 5mm. To tackle this problem, a novel interpolation method is proposed to enhance the reconstruction results not only from the cross-section view but also the longitudinal-section view. However, the two views are not well aligned. Then, before interpolation, an alignment scheme is proposed to register the two views well. After alignment, the fine-grained 3D structure of kidney can be well reconstructed from this set of CT images with a lower-resolution in the Z axis.