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.
Computer-aided classification of breast masses using speckle features of automated breast ultrasound images
Rent:
Rent this article for
USD
10.1118/1.4754801
/content/aapm/journal/medphys/39/10/10.1118/1.4754801
http://aip.metastore.ingenta.com/content/aapm/journal/medphys/39/10/10.1118/1.4754801

Figures

Image of FIG. 1.
FIG. 1.

A sample of 3D automated breast ultrasound (ABUS) images. (a) A series of continuous 2D slices in axial view. (b) The reconstructed 3D volumetric image can be observed from axial, sagittal, and coronal view. The VOI is used to specify the tumor area. (c) The three views of the VOI. (d) The 3D view of the VOI.

Image of FIG. 2.
FIG. 2.

Illustration of speckle density in a 2D slice of VOI from axial view. (a) A 2D slice of VOI of a benign case. (b) The speckle pixels of (a) are shown by a white appearance in the slice. (c) A 2D slice of VOI of a malignant case. (d) The speckle pixels of (c) are shown by a white appearance in the slice. The speckle density in (b) (the benign case) is higher than that in (d) (the malignant case).

Image of FIG. 3.
FIG. 3.

The receiver operating characteristic (ROC) curves of the speckle features, the morphological features, and the combined feature set. The Az of combing speckle and morphological features was significantly better than that of using morphological features only (0.96 [95% CI: 0.919–0.980] vs 0.91 [95% CI: 0.856–0.949], p-value = 0.0154).

Image of FIG. 4.
FIG. 4.

The cases classified correctly by the speckle features and misclassified by the morphological features. (a) A 2D slice of VOI of a benign case. (b) The speckle pixels of (a) are shown by a white appearance in the slice. (c) The segmentation result in 3D view of the benign case. (d) A 2D slice of VOI of a malignant case. (e) The speckle pixels of (d) are shown by a white appearance in the slice. (f) The segmentation result in 3D view of the malignant case. For the benign case in (a), the likelihood of malignancy predicted by the speckle features (b) and the morphological features (c) was 0.01 and 0.99, respectively. For the malignant case in (d), the likelihood of malignancy predicted by the speckle features (e) and the morphological features (f) was 0.64 and 0.23, respectively.

Image of FIG. 5.
FIG. 5.

The cases classified correctly by the morphological features and misclassified by the speckle features. (a) A 2D slice of VOI of a benign case. (b) The speckle pixels of (a) are shown by a white appearance in the slice. (c) The segmentation result in 3D view of the benign case. (d) A 2D slice of VOI of a malignant case. (e) The speckle pixels of (d) are shown by a white appearance in the slice. (f) The segmentation result in 3D view of the malignant case. The likelihood of malignancy of the benign case in (a) was predicted as 0.87 and 0.05 according to the speckle features (b) and the morphological features (c), respectively. In (d), the likelihood of malignancy of the malignant case predicted based on the speckle features (e) and the morphological features (f) was 0.24 and 0.86, respectively.

Image of FIG. 6.
FIG. 6.

A malignant case classified correctly by the combined feature set and misclassified by both the speckle features and the morphological features. (a) A 2D slice of VOI. (b) The speckle pixels of (a) are shown by a white appearance in the slice. (c) The segmentation result in 3D view of the malignant case. The malignant case in (a) was classified correctly by the combined feature set (malignancy = 0.71) and misclassified by both the speckle features [(b), malignancy = 0.31)] and the morphological features [(c), malignancy = 0.44].

Tables

Generic image for table
TABLE I.

The significant speckle features for the classification of breast masses.

Generic image for table
TABLE II.

The performance of CAD with the speckle features, the morphological features, and the combined feature for the classification of breast masses.

Loading

Article metrics loading...

/content/aapm/journal/medphys/39/10/10.1118/1.4754801
2012-10-03
2014-04-20
Loading

Full text loading...

This is a required field
Please enter a valid email address
752b84549af89a08dbdd7fdb8b9568b5 journal.articlezxybnytfddd
Scitation: Computer-aided classification of breast masses using speckle features of automated breast ultrasound images
http://aip.metastore.ingenta.com/content/aapm/journal/medphys/39/10/10.1118/1.4754801
10.1118/1.4754801
SEARCH_EXPAND_ITEM