Computerised prostate boundary estimation of ultrasound images using radial bas-relief method

Abstract

A new method is presented for automatic prostate boundary detection in ultrasound images taken transurethrally or transrectally. This is one of the stages in the implementation of a robotic procedure for prostate surgery performed by a robot known as the robot for urology (UROBOT). Unlike most edge detection methods, which detect object edges by means of either a spatial filter (such as Sobel, Laplacian or something of that nature) or a texture descriptor (local signal-to-noise ratio, joint probability density function etc.), this new approach employs a technique called radial bas-relief (RBR) to outline the prostate boundary area automatically. The results show that the RBR method works well in the detection of the prostate boundary in ultrasound images. It can also be useful for boundary detection problems in medical images where the object boundary is hard to detect using traditional edge detection algorithms, such as ultrasound of the uterus and kidney.

Texture Guided Active Appearance Model Propagation for Prostate Segmentation

“…However, to have an overall qualitative estimate of the functioning of our method we observe that the mean segmentation time of 0.72 ± 0.05 seconds for an image is comparable to [2](less than a second), [15](2.1 second), [11](5 seconds), [3](5 seconds) and inferior only to [20] that achieves segmentation time of 0.3sec in C++ and ITK framework. To have an estimate of overlap accuracy DSC value of 0.95 ± 0.01 is comparable to different measure of overlap accuracy value obtained by [12](Area difference 8.48%), [8](Area difference 4.79 ± 0.68%), [10](Average similarity 89%), [18](Area overlap error 3.98±0.97%), [3](Area overlap 93 ± 0.9%), [16](Area overlap 93%), [21](Area overlap 91%) and [4](Area accuracy 94.05%).…”

A survey of prostate modeling for image analysis

“…These statistics are computed from the histogram of the gray levels of the image and depend only on the individual pixels and not on their neighborhood. In the prostate, first-order statistics have been used in the following models: the intensity profile model [47], gradient models [48,49], models using the gray level threshold of the regions extracted from a neural network [50], a radial basis relief model [51], an instantaneous variation coefficient (ICOV) model [52], a model using the local standard deviation in a multiresolution framework [53], posterior probability models [9,54], mixture probability distribution models [42,43,55,56], or models that are combined in many other ways [13,15,57,58]. Feng et al [49] proposed a weighted combination of gradient and probability distribution functions.…”

Tessellated Surface Reconstruction from 2D Contours

“…In our case, outlining the object boundary (e.g. prostate) of ultrasound images is done by employing a set of segmentation processes consisting of image pre-processing, edge enhancing using Radial Bas-Relief [2] followed by a model-based boundary extraction [3]. To form a 3D structure, the contours that use polygonal approximation [4] from adjacent cross sections are connected.…”