Efficient border extraction of image feature

A system for mammography comprising:

• an XRAY source;
• a means of obtaining a digitalized X-Ray image of part of a female torso as a binary array;
• a computer configured to segmentize the digitalized image, wherein the image is a medical image, and identify pixels corresponding to border of a breast by a method comprising the steps of:
• (b) scanning a first horizontal row across the picture in a first direction from outermost pixel of picture towards the breast and calculating grayscale gradient across adjacent pixels until a gradient is detected as being significant;
• (c) confirming that the variation in adjacent pixels indicates a first true boundary pixel of the feature by ascertaining that grayscale gradient across subsequent pixels in series of several pixels along the line still significant;
• (d) iteratively repeating the steps (v) to (vii) of:
• (v) scanning next row in vicinity of said boundary pixel, and
• (vi) determining the boundary pixel of the adjacent row by assessing significant pixel intensity gradient along the row until a failure to find a predetermined significant gradient occurs in which case;
• (vii) attempt scanning vertically by repeating steps (v) and (vi) until the edge of picture is reached having segmentized the breast to one side of the first line, and
• (e) repeating step (d) on other side of the first row.

A method of extracting a boundary surrounding a feature from a monochrome digital image, wherein the image is a medical image, comprising a two dimensional array of pixels, the method comprising the steps of:

• (ii) scanning along a line of pixels in a first direction across the image from the outside edge of image towards the feature until a significant grayscale gradient occurs across a number of adjacent pixels;
• (iii) ascertaining that the gradient indicates a true boundary pixel by evaluating the gradient across subsequent pixels;
• (iv) identifying boundary pixels in an adjacent line by scanning grayscale intensity across pixels bordering the first boundary pixel;
• (v) iteratively scanning next adjacent line to identify adjacent boundary pixels until technique fails;
• (vi) rotating scanning direction to a direction orthogonal to the first direction;
• (vii) repeating steps (ii) to (vi) until the first boundary pixel is reached after tracing a closed loop enclosing the feature or an edge of the image is reached.

A method for preprocessing a raw digital image using a computer, wherein the image is a medical image, for subsequent segmentation of a feature therein comprising the steps of:

• (i) removing any frame;
• (ii) setting brightest pixel in picture to a maximum intensity value of a simplified scale;
• (iii) setting dullest pixel to a background intensity value of the simplified scale;
• (iv) converting intensities of intermediate intensity pixels to intermediate values of the simplified scale;
• (v) sampling the digital picture with an integer coefficient to produce a sampled image having a reduced resolution;
• (vi) computing a raw image histogram from the sampled image;
• (vii) where digital image is framed, identifying and removing frame from the sampled image to produce a clean image;
• (viii) selecting most intense pixel and least intense pixel from said clean image histogram;
• (ix) setting grey-scales of least intense and most intense pixels to standardized minimum and maximum arbitrary values;
• (x) creating a look-up-table (LUT) for transforming grayscales of pixels with intermediate intensities by calculating a logarithmic scale between said maximum and minimum intensities;
• and
• (xi) applying said LUT to the clean image to produce a normalized image.

A method of confirming that the variation in grey level intensity of a suspected boundary pixel of a feature in an axial line across digital image is a true boundary pixel of the feature comprising the steps of:

• i. determining a reference gradient range;
• ii. determining a minimum and a maximum line segment length;
• iii. scanning along the line towards feature;
• iv. setting line segment to its minimum length
• v. calculating a local average of grayscale values about each pixel by averaging grayscale intensities of a line segment prior to and adjacent the pixel;
• vi. calculating the local grayscale gradient at the boundary pixel;
• such that
• vii. if the local gradient is not within the reference gradient range then the suspected boundary pixel is moved to the end of the line-segment and steps iv, v and vi are repeated;
• viii. if the local gradient is within the reference gradient range and the line segment length is below the maximum segment length, then the segment length is increased and steps (v) and (vi) are repeated, and
• ix. if the local gradient is within the reference gradient range and the line segment length is not below the maximum segment length then the suspected boundary pixel is validated as being a true boundary pixel.