Computerized ultrasound characterization of breast tumors

Abstract

In this paper, we present a new approach to the segmentation and analysis of solid breast nodules in ultrasonography. We have applied computer vision techniques to segment the nodules and analyze a series of diagnostic criteria, which can help discriminate malignant and benignant tumors. The segmentation is carried out by means of a combination of a region growing algorithm and the active contour technique. On the other hand, the analysis of the diagnostic criteria involves several methods, such as the extraction of the minimum distance ellipse through gradient descent, pseudocorner location or structure tensor. The methods that we propose have provided quite satisfactory results and show the usefulness of image processing techniques in the diagnosis by means of medical imaging. The aim of this work is not the substitution of the specialist, but the generation of a series of parameters which reduce the need of carrying out the biopsy. D

Active contours in medical image processing. Theory and applications

“…The examples are multi-direction snakes: skin cancer images [14], topology-adaptive snakes: MR brain images and CT scans [23], gravitational force snakes: a variety of medical and non-medical images [15], narrow-band snakes: MRI and CT scan images of lungs [41], distance snake, GVF snake, balloon snake , "area and length" snakes, geodesic snakes, constrained snakes and level set method: MRI, CT and US images of brain, liver, kidney [32], region-competition snakes (originally [22]): CT scan slices of arteries [44], sectored snakes : abdominal CT scans [5], parametric snakes: US of breast masses [45], 3D-snakes: US breast cancer images [45], [46], GVF snakes with edge map preprocessing: US of the kidney tumors [9], GVF snakes combined with the region growing and the median filter :US breast tumors [10], sketch-snakes: chest X-Ray images [46], combination of snakes and active shape models: US of the human heart [47], the so-called early-vision and the discretesnake model: a variety of the US images [48], multi-resolution snake: echographic and echobrachial images [49], GGVF snakes combined with a continuous force field analysis: breast tumor US images [50] .…”

“…A comparison between the segmented image with and without the preprocessing shows that this module produces a great improvement in the accuracy of the boundary detection. A combination of filtering, edge map and initialization by a human operator [10] employs an iterative truncated median filter to reduce the speckle noise. In [57] the speckle noise is suppressed by anisotropic diffusion filter [58] and a stick filter [59].…”

“…Since the seminal work of Kass and colleagues, techniques based on active contours have been applied to many object extraction tasks with a different degree of success. In particular, snakes have been used to locate the objects in various applications of medical image processing such as segmentation of abnormalities in the images of the human heart, liver, brain, breast, etc [2]- [10].…”

Phase portrait analysis for automatic initialization of multiple snakes for segmentation of the ultrasound images of breast cancer

“…The examples are multi-directional snakes: skin cancer images [57], topology-adaptive snakes: MR brain images and CT scans [76], gravitational force snakes: a variety of medical and non-medical images [77], narrow-band snakes: MRI and CT scan images of lungs [78], distance snake [79], GVF snake, balloon snake [80], ''area and length'' snakes [72], geodesic snakes [71], constrained snakes [73] and level set method: MRI, CT and US images of brain, liver and kidney [74], region-competition snakes (originally [81]): CT scan slices of arteries [82], sectored snakes [83]: abdominal CT scans [84], parametric snakes: US of breast masses [85], 3D snakes: US breast cancer images [22,85], GVF snakes with an edge map preprocessing: US of the kidney tumors [86], GVF snakes combined with the region growing and the median filter: US breast tumors [87], sketch snakes [22]: chest X-ray images [22], combination of snakes and active shape models: US of the human heart [88], the early vision and the discrete snakes: the US images [89], multi-resolution snake: echographic and echobrachial images [90], GGVF snakes combined with a continuous force field analysis: breast tumors in the US images [91] and geodesic snakes and coupled geometric snakes: female pelvic organs in the MRI images [92][93][94].…”

Phase Portrait Analysis for Multiresolution Generalized Gradient Vector Flow

“…The snakes have been used to locate the object boundaries in various applications of medical image processing with a different degree of success. In particular, they have been successfully applied to segmentation of abnormalities in the images of the human heart, liver, brain, breast, etc [2]- [12].…”