(a) Original breast MR image; (b) region of interest is defined.
(a) Simulated curve for chest wall. (c) Result of chest removal in the breast image.
Determination of the turning point through the middle point and point of tangency.
(a) Four histograms of pixel distributions are created to determine the nipple position. (b) The region of the nipple is detected and highlighted in bright color.
Representatives of nine sets of breast MR images used for performance evaluation.
The different sampling scopes of contour points for curve fitting. The scopes of Dmin, Dmax, and (Dmin+Dmax)/2 are the smallest, middle, largest circles, respectively.
Nipple regions covered for computing Dice coefficient. The region delineated by the expert: the darker regions. The region detected by the proposed method: brighter regions + darker regions.
RMSE comparison of simulated breast contour using four different sampling methods for Case 3. (a) YBnd: left = 2.9369, right = 2.0299; (b) Dmax: left = 2.7435, right = 2.0688; (c) Dmin: left = 0.6724, right = 1.4765; (d) (Dmin+Dmax)/2: left = 1.2377, right = 0.6250.
RMSE comparison of simulated breast contour using four different sampling methods for Case 6. (a) YBnd: left = 1.7205, right = 1.6747; (b) Dmax: left = 1.2519, right = 1.5747; (c) Dmin: left = 3.0436, right = 2.0368; (d) (Dmin+Dmax)/2: left = 2.5429, right = 1.9511.
RMSE comparison of simulated breast contour using four different sampling methods for Case 9. (a) YBnd: right = 3.0001; (b) Dmax: right = 3.1413; (c) Dmin: right = 1.2067; (d) (Dmin+Dmax)/2: right = 2.228.
A comparison of different methods for estimating root mean square error (RMSE). Unit: pixel.
The three sets of null hypotheses to be tested.
Three ANOVA test results.
Dice coefficients of the nine cases measured by the proposed method.
Centroid difference of the nine cases measured in the compared methods.
Article metrics loading...
Full text loading...