Generation of Three-Dimensional Images Using Ultrasonic Pulse-Echo Data for Fault Characterization

Abstract

The nondestructive testing using ultrasonic pulse-echo data is an effective method especially for metal structure. Typically, the ultrasonic pulse-echo data is processed and the results are shown in A-scan, B-scan or C-scan data formats. In conventional testing using ultrasonic pulse-echo data, the inspector is able to identify the location of faults as well as their rough dimensions upon viewing B-scan and C-scan data. In this paper, we propose an approach using 3-D visualization of ultrasonic pulse-echo data. The main idea of the proposed method is to use various 3-D visualization methods frequently used in medical image visualization systems, namely, surface rendering (SR), volume rendering (VR) and maximum intensity projection (MIP). Such 3-D visualization of ultrasonic pulse-echo data enables easier identification of the location and dimension of faults more accurately. Introduction The nondestructive testing aims to detect faults (or discontinuities) inside or outside the structure and to evaluate physical and mechanical characteristics of the structure without harming it [1]. Of various nondestructive testing methods, the ultrasonic pulse-echo method is an effective method especially for metal structures. The pulse-echo technique has been receiving increasing interests, giving rise to the development of various multi-transducer (or an array of transducers) techniques to improve spatial resolution in the reconstructed image [2,3,4]. Nondestructive testing of three-dimensional structures requires various tools in order for inspectors to efficiently visualize, detect, locate and size the faults that may or may not be present within the 3-D structures. Currently accepted testing procedure based on ultrasonic pulse-echo data infers various fault characterization from viewing B-scan and C-scan data simultaneously. The B-scan shows a cross-sectional side view of faults and C-scan shows a two-dimensional projection view of faults [5]. But, this method has a distinct disadvantage that the evaluation depends heavily upon inspector’s experience and expertise. This paper is aimed at generating clear 3-D pictures of the structures under evaluation, rather than a sequence of B-scan and C-scan images. Generation of 3-D images of human anatomy has recently gained popularity as patients are typically scanned volumetrically, sometimes producing a hundred or more slices [6,7]. These slices are typically rendered in 3-D using one of the three techniques: surface rendering (SR), volume rendering (VR) and maximum intensity projection (MIP) [8]. The B-scan data is essentially equivalent to slice images of X-ray CT data, for instance, and is amenable for such 3-D rendering. We have previously reported some initial results on the MIP technique to visualize the pulse-echo data [9,10] and in this paper we focus on the efficacy of all three rendering techniques. In this paper, the three rendering techniques will be analyzed as applicable for processing pulse-echo data and pre-processing steps that precede 3-D visualization process will be introduced: deconvolution of the pulse-echo data through Wiener filtering and interpolation kernel selection 2 Title of Publication (to be inserted by the publisher) minimizing rendering artifacts. It will be demonstrated that the 3-D visualization of pulse-echo data through various rendering techniques is effective for viewing and detecting the faults in 3-D structures. Data Acquisition and Processing Figure 1 shows the entire procedures of pulse-echo data acquisition and processing. The transducer is placed on top of the specimen and collects the pulse-echo data moving along the raster scan line. At each scan line, we acquire one B-scan data of the specimen and the entire B-scan data of the specimen is essentially a sequence of slice images of the specimen. These B-scan data pass through the Wiener filter and the linear interpolator before the 3-D visualization step for the generation of 3-D images. Wiener filtering. Because the ultrasonic beam from the transducer is diffracted and cone-shaped, the acquired B-scan data consists of horizontal blurring. Such horizontal blurring of the B-scan data will lead to blurred 3-D images. Therefore, the blurring of B-scan data must be eliminated before 3-D visualization step. Since the blurring can be assumed as a two-dimensional convolution in the y x − plane between the 3-D data and the point spread function (PSF), the acquired B-scan data ) , , ( z y x g can be represented as [9,10]: ) , , ( ) , , ( ) , , ( ) , , ( z y x n z y x h z y x f z y x g + ∗ = (1) where ) , , ( z y x f , ) , , ( z y x h and ) , , ( z y x n denote the desired 3-D data of the specimen, the PSF of the ultrasonic beam and the inherent measurement noise, respectively. The deblurring of B-scan data is realized through Wiener filtering. Wiener filtering produces ) , , ( ˆ z y x f that estimates original 3-D shape of specimen ) , , ( z y x f from ) , , ( z y x g . ) , , ( ˆ z y x f minimizes the mean-squared error between ) , , ( z y x f and ) , , ( ˆ z y x f [11]. The actual filtering process is performed iteratively in the frequency domain and can be expressed as the following [10]: Fig. 1 The acquistion and processing of the ultrasonic pulse-echo data. ) , , ( ) , , ( ˆ / ) , , ( ) , , ( ) , , ( ˆ 2 ) ( 2 ) ( ) 1 ( z G z F K z H z H A z F y x k y x y x y x k k y x ω ω ω ω ω ω ω ω ω ω + = ∗ + (2)