1887
banner image
No data available.
Please log in to see this content.
You have no subscription access to this content.
No metrics data to plot.
The attempt to load metrics for this article has failed.
The attempt to plot a graph for these metrics has failed.
The simulation of 3D microcalcification clusters in 2D digital mammography and breast tomosynthesis
Rent:
Rent this article for
USD
10.1118/1.3662868
/content/aapm/journal/medphys/38/12/10.1118/1.3662868
http://aip.metastore.ingenta.com/content/aapm/journal/medphys/38/12/10.1118/1.3662868

Figures

Image of FIG. 1.
FIG. 1.

(a) A flow chart that explains the main steps to acquire 3D models of microcalcification clusters. (b) A flow chart describing the processing steps in the simulation framework.

Image of FIG. 2.
FIG. 2.

(a) A photograph of a biopsy specimen containing microcalcification clusters after extraction from a patient. (b) An example of a 3D model of microcalcification cluster from the database built during this study and considered as full cluster (voxel size was 0.02 mm). (c) An example of a 3D model of microcalcification cluster from the database and considered as subcluster (voxel size was 0.03 mm).

Image of FIG. 3.
FIG. 3.

(a) A flow chart explaining the steps involved in the segmentation procedure of the micro-CT reconstruction images to form a 3D model. (b) All the steps in (a) are illustrated by images as output of each step.

Image of FIG. 4.
FIG. 4.

The different outputs of the segmentation procedure applied on each reconstructed micro-CT image shown for visual check with the original input in gray scale and the borders of the output represented by the red contour. (a) The output of algorithm 1 in the segmentation procedure. (b) The output of algorithm 2 in the segmentation procedure. (c) The final output of the segmentation procedure after the combination of algorithms 1 and 2 through logical AND.

Image of FIG. 5.
FIG. 5.

Illustration of the out of plane artifacts of a real microcalcification cluster. These are 8 out of the 15 planes shown to the radiologists starting from the top left as 1, 3, 5, …, 15.

Image of FIG. 6.
FIG. 6.

ROC analysis of the 2D FFDM study with area under the ROC curve (AUC) per observer. The false positive fraction (FPF) is plotted against the true positive fraction (TPF). The AUC is calculated using the trapezoidal method.

Image of FIG. 7.
FIG. 7.

Examples of false positive cases in the 2D FFDM study, i.e., simulated microcalcification clusters thought to be real by at least three radiologists. (a) The simulated cluster is shown within the processed mammogram and highlighted with the arrow, in the left bottom corner a magnification view of the cluster. (b) Other false positive cases in magnification view.

Image of FIG. 8.
FIG. 8.

Examples of false negative cases in the 2D FFDM study, i.e., real microcalcification clusters thought to be simulated by at least three radiologists. (a) A processed mammogram with a real cluster highlighted with the arrow, on the left bottom corner a magnification of this cluster. (b) Other false negative clusters in magnification view.

Image of FIG. 9.
FIG. 9.

ROC analysis of the tomosynthesis study with area under the ROC curve (AUC) per observer. The false positive fraction (FPF) is plotted against the true positive fraction (TPF). The AUC is calculated using the trapezoidal method.

Image of FIG. 10.
FIG. 10.

Examples of false positive cases (i.e., simulated microcalcification clusters thought to be real cases by at least three radiologists) in the tomosynthesis study with the in-focus plane presented, while the microcalcifications were distributed over multiple planes within the cluster. The window level was changed for every image for better visualization. (a) The simulated cluster is shown within the breast with the arrow pointing to it, in the right bottom part the cluster in magnification view. (b) Other false positive cases in magnification view.

Image of FIG. 11.
FIG. 11.

Examples of false negative cases (i.e., real microcalcification clusters thought to be simulated by at least three radiologists) in the tomosynthesis study with only the in-focus plane presented, but most of the microcalcifications were distributed over multiple planes within the same cluster. The window level was modified for every image for better visualization. (a) The real cluster is shown within the breast with the arrow pointing to it, in the right bottom part the cluster in magnification view. (b) Other false negative cases in magnification view.

Image of FIG. 12.
FIG. 12.

Some examples of true negative cases, i.e., simulated microcalcification clusters where at least three radiologists agreed that they were simulated, in the 2D FFDM study. (a) The cluster had high contrast. (b) The cluster looked super-imposed and artificial when magnified.

Tables

Generic image for table
TABLE I.

The description of the patients and clusters used in the validation studies and the training studies in terms of real and simulated population along with the total number of cases and clusters used in the overall study. The patients in the “group with real clusters” are patient cases with real microcalcification clusters, while the patients within the “group with simulated clusters” are real patient cases where their background were used for simulating microcalcification clusters.

Generic image for table
TABLE II.

The description of the models used in the validation studies in terms of quantities, distribution, and malignancy/benign description.

Generic image for table
TABLE III.

Area under the ROC curve (AUC) along with the 95% CI for the 2D study per observer; last raw reports the average AUC over all readers and the corresponding 95% CI.

Generic image for table
TABLE IV.

Area under the ROC curve (AUC) along with the 95% CI for the tomosynthesis study per observer; last raw reports the average AUC over all readers and corresponding 95% CI.

Generic image for table
TABLE V.

The agreement, as measured by the kappa statistic, between the 2D FFDM and the tomosynthesis studies for the simulated population and the real population. Listed are the values of the kappa statistic, the 95% confidence interval, and the conventional interpretations of the kappa values for each observer and the average.

Loading

Article metrics loading...

/content/aapm/journal/medphys/38/12/10.1118/1.3662868
2011-11-28
2014-04-24
Loading

Full text loading...

This is a required field
Please enter a valid email address
752b84549af89a08dbdd7fdb8b9568b5 journal.articlezxybnytfddd
Scitation: The simulation of 3D microcalcification clusters in 2D digital mammography and breast tomosynthesis
http://aip.metastore.ingenta.com/content/aapm/journal/medphys/38/12/10.1118/1.3662868
10.1118/1.3662868
SEARCH_EXPAND_ITEM