Methods of the measurement of three dimensional (3-D) shapes in medical and biomedical fields are discussed. 3-D reconstruction methods of the left ventricle from x-ray, ultrasound and magnetic resonance imaging are especially focused. Some results of 3-D shapes of the left ventricle reconstructed and 3-D functional images are shown.
In this paper, we establish a new mathematical basis of CT image reconstruction based on the spectral decomposition of Radon transform. Operator representations of unified image reconstruction algorithms from continuously viewed projection data are presented, and the fundamental problems are studied by constructing the spectral decomposition of these operators. The structure of reconstruction error due to measurement noise and the image reconstruction capability from limited view angle projection data are discussed in the reconstruction algorithm using Helgason-Ludwig consistency condition. Throughout the investigation in this paper, the theoretical possibility and limitation of CT imaging technology can be obtained.
The mathematical analysis of CT image reconstruction based on the spectral decomposition of Radon transform, which was developed in previous paper, is extended to the image reconstruction problem from discretely viewed projection. In the discretely viewed situation, the image reconstruction is carried out under the minimum norm least square criterion, because the mathematically exact reconstruction is impossible. The structure of reconstruction error due to the discretely viewed and the limited viewed projection is discussed. The relationship between reconstruction error and number of projection can be obtained. Consequently, the theoretical possibility and limitation in practical CT imaging technology are made clear.
A simple quantitative analysis system was developed using personal computer-basis and a cineangio-projector system for coronary artery, at injection of two medical agents. The accuracy and precision of the contour detection procedure were validiated on the basis of phantom models filled with a contrast medium.
This paper describes the researches on parallel processors for image and signal processings in the United States and the United Kingdom. The processors introduced here have new and high-performance architectures : a real-time image processor using commercial LSI array chips (GAPP) and a general purpose image processor (PASM) at Purdue University, the Wavefront Array Processor at University of Southern California, a systolic array processor (Warp) at Carnegie Mellon University, and the Associative String Processor (ASP) at Brunel University.
We propose a high-level language specification (DFLIP) for the image processing with data flow machines. And we outline the compiler which effectively translates the von Neumann program written by the proposed language into the data flow graph. This compiler consists of syntax analyzer, vactor analizer, semantics analyzer, and code generator. The vactor analizer abstracts the parallelism from the program and makes the data flow graph. The experimental results with the data flow processor μ PD7281 show the effectiveness of the proposed compiler.