A Neural Network Architecture to Identify the Bone Tissue for Solid Freeform Fabrication
Computed tomography produces sets of tomograms for medical interpretation. Typical interpretation consists of imaging and simple observation on a 2D display screen, so that feature extraction and tissue differentiation is based primarily on human expertise. Solid freeform fabrication offers the promise of fabrication of prostheses based on actual patient anatomy. Use of CT data for this purpose requires automated interpretation. This paper presents a system architecture based on neural networks for the segmentation and classification of tissues of interest in tomograms. This approach produces a quantitative recovery of the available information by applying a feed-forward neural net trained with the back-propagation algorithm. The neural network architecture selected was tested on fabricated CT image matrices of the lower extremity.