Nodule Detection

Abstract

A method of detecting a nodule in a three-dimensional scan image comprises calculating a three-dimensional sphericity index for each point in the scan image ( 310 - 330 ), applying a high sphericity threshold to the sphericity index ( 340 ) to obtain a candidate nodule region, and then performing region-growing ( 350 ) from the candidate region using a relaxed sphericity threshold to determine an extended region including less spherical parts connected to the candidate region. Optionally, spherical filtering may be applied to the image by matching the spherical filter to the extended region.

1 . A method of identifying a nodule in a computed tomography scan image of a lung, comprising:
 (a) identifying a region of high sphericity within the image;   (b) extending the region to include connected points of lower sphericity; and   (c) outputting the extended object as an identified nodule.   
   
   
       2 . The method of  claim 1 , wherein step (a) includes calculating a sphericity index map of the image, and identifying connected points with a high sphericity index as belonging to said region. 
   
   
       3 . The method of  claim 2 , wherein the sphericity index indicates the variation in curvature of an iso-intensity surface in the region around each point. 
   
   
       4 . The method of  claim 3 , wherein the sphericity index is calculated from a partial derivative of intensity around each point. 
   
   
       5 . The method of  claim 4 , wherein the sphericity index is calculated from the first and second partial derivatives of intensity in three dimensions. 
   
   
       6 . The method of  claim 2 , wherein the sphericity index map is calculated on a smoothed image. 
   
   
       7 . The method of  claim 2 , wherein the step of identifying points having a high sphericity index comprises detecting whether each said point has a sphericity index above a predetermined high sphericity index threshold. 
   
   
       8 . The method of  claim 7 , wherein step (b) includes performing three-dimensional region growing from said region to add said connected points of lower sphericity. 
   
   
       9 . The method of  claim 8 , wherein said connected points are determined as having lower sphericity if their sphericity index is above a relaxed sphericity index threshold lower than said high sphericity index threshold. 
   
   
       10 . The method of  claim 1 , further including performing filtering around the extended region using a spherical filter. 
   
   
       11 . The method of  claim 10 , wherein the spherical filter comprises an inner spherical region of positive weight and an outer spherical region of negative weight. 
   
   
       12 . The method of  claim 11 , wherein the inner and outer spherical regions have approximately equal volumes. 
   
   
       13 . The method of  claim 11 , wherein the inner spherical region has a diameter approximately equal to the diameter of the extended region. 
   
   
       14 . The method of  claim 11 , wherein the filtering step comprises convolving the inner and outer spherical regions with the scan image. 
   
   
       15 . The method of  claim 14 , wherein the diameter of the inner spherical region is varied so as to determine a maximum strength of said convolution, and the spherical filtering corresponding to the maximum convolution strength is applied to the scan image to generate a spherically enhanced output image. 
   
   
       16 . The method of  claim 2 , further including performing filtering around the extended region using a spherical filter comprising an inner spherical region of positive weight and an outer spherical region of negative weight, wherein the filtering step comprises convolving the inner and outer spherical regions with the sphericity index map. 
   
   
       17 . The method of  claim 16 , wherein the diameter of the inner spherical region is varied so as to determine a maximum strength of said convolution, and the spherical filtering corresponding to the maximum convolution strength is applied to the sphericity index map to generate a spherically enhanced output image. 
   
   
       18 . Apparatus for identifying a nodule in a computed tomography scan image of a lung, comprising:
 (a) means for identifying a region of high sphericity within the image;   (b) means for extending the region to include connected points of lower sphericity; and   (c) means for outputting the extended object as an identified nodule.   
   
   
       19 . The apparatus of  claim 18 , wherein the means for identifying the region includes means for calculating a sphericity index map of the image, and for identifying connected points with a high sphericity index as belonging to said region. 
   
   
       20 . The apparatus of  claim 19 , wherein the sphericity index indicates the variation in curvature of an iso-intensity surface in the region around each point. 
   
   
       21 . The apparatus of  claim 20 , wherein the sphericity index is calculated from a partial derivative of intensity around each point. 
   
   
       22 . The apparatus of  claim 21 , wherein the sphericity index is calculated from the first and second partial derivatives of intensity in three dimensions. 
   
   
       23 . The apparatus of  claim 19 , wherein the sphericity index map is calculated on a smoothed image. 
   
   
       24 . The apparatus of  claim 19 , wherein the means for identifying points having a high sphericity index comprises means for detecting whether each said point has a sphericity index above a predetermined high sphericity index threshold. 
   
   
       25 . The apparatus of  claim 24 , wherein the means for extending the region is arranged to perform three-dimensional region growing from said object to add said connected points of lower sphericity. 
   
   
       26 . The apparatus of  claim 25 , wherein said connected points are determined as having lower sphericity if their sphericity index is above a relaxed sphericity index threshold lower than said high sphericity index threshold. 
   
   
       27 . A method of identifying a nodule in a computed tomography scan image of a lung, comprising:
 (a) calculating a sphericity index map of the image, wherein the sphericity index indicates the variation in curvature of an iso-intensity surface around each point of the image;   (b) identifying a connected region of high sphericity index within the image;   (c) performing three-dimensional region growing from said region to add connected points of lower sphericity; and   (d) outputting the extended object as an identified nodule.   
   
   
       28 . Apparatus for identifying a nodule in a computed tomography scan image of a lung, comprising:
 (a) means for calculating a sphericity index map of the image, wherein the sphericity index indicates the variation in curvature of an iso-intensity surface around each point of the image;   (b) means for identifying a connected region of high sphericity index within the image;   (c) means for performing three-dimensional region growing from said region to add connected points of lower sphericity; and   (d) means for outputting the extended object as an identified nodule.   
   
   
       29 . A computer program product readable by a machine and embodying program code arranged to perform the following method steps for identifying a nodule in a computed tomography scan image of a lung, when executed by the machine:
 (a) calculating a sphericity index map of the image, wherein the sphericity index indicates the variation in curvature of an iso-intensity surface around each point of the image;   (b) identifying a connected region of high sphericity index within the image;   (c) performing three-dimensional region growing from said region to add connected points of lower sphericity; and   (d) outputting the extended object as an identified nodule.   
   
   
       30 . An article comprising a medium for storing instructions to enable a processor-based system to:
 (a) identify a region of high sphericity based on a computed tomography scan image of a lung;   (b) extend the region to include connected points of lower sphericity; and   (c) output the extended object as an identified nodule.