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.
Automated noninvasive classification of renal cancer on multiphase CT
Rent this article for


Image of FIG. 1.
FIG. 1.

Multiphase abdominal 4D CT data of a patient with VHL. 2D slices of 3D volumes are shown (a) before contrast and (b) at PVP, when tumors are better distinguished from the renal parenchyma.

Image of FIG. 2.
FIG. 2.

Examples of types of renal tumors as they appear at PVP. The intensity and homogeneity inside the lesion are indications used in diagnosis. These examples include small, large, cystic (lower intensity/enhancement), and solid (higher intensity/enhancement), homogeneous and heterogeneous tumors. A BHD tumor is shown in (a); in (b) and (f) there are homogeneous and heterogeneous VHL cancers; cysts can be found in (c) and (f); a HPRC lesion is shown in (d); and a small HLRCC cancer is presented in (e).

Image of FIG. 3.
FIG. 3.

Diagram of the algorithm flow. Multiphase (4D) CT data are registered between phases before the segmentation is performed on the PVP image. The lesion biomarkers are extracted using data from all phases of enhancement. Finally, classification is performed to determine the types of tumors.

Image of FIG. 4.
FIG. 4.

The correction of abdominal motion (preprocessing) shown on 2D slices of 3D volumes (a) the image before contrast; (b) at PVP; and (c) the noncontrast image after motion correction. Images are shown at the same body location, as seen at the vertebral body. However, the VHL tumor in the right kidney (shown in the highlights) was present only in the enhanced image (b) and not at the same location in image (a). The location of the tumor is corrected in (c) after nonlinear registration.

Image of FIG. 5.
FIG. 5.

The cascade of binary classifiers used for renal lesion classification. The number of correctly classified lesions (true positive ratio) that reaches each branch is presented using HCF.

Image of FIG. 6.
FIG. 6.

Segmentation of renal lesions. A VHL cancer was segmented in the PVP image (left), as shown by the semi-automated tumor contour, and the segmentation was propagated to the noncontrast data (right). Before contrast-enhancement the lesion was invisible to the eye, but through spatial registration and segmentation propagation, the cancer can be analyzed at all phases of enhancement.

Image of FIG. 7.
FIG. 7.

3D renderings of segmented lesions using the semi-automated technique for renal cancer quantification and classification. A total of 25 random lesions of five types are presented; there are five lesions per row, each from a different patient: From top to bottom, VHL, BHD, cysts, HPRC, and HLRCC cases.

Image of FIG. 8.
FIG. 8.

Examples of distribution of shape indexes between types of lesions: (a) two benign cysts; (b) a solid VHL cancer and a cystic HLRCC cancer; (c) two solid lesions: VHL and BHD; and (d) two cystic cancers: HPRC and HLRCC.

Image of FIG. 9.
FIG. 9.

Classification ROC curves by HCF vs. using only mean multiphase CT values. AUC data are given in Table III.


Generic image for table

Features used by the HCF descriptor.

Generic image for table

Renal lesion quantification variability.

Generic image for table

ROC analysis for the classification of renal lesions.


Article metrics loading...


Full text loading...

This is a required field
Please enter a valid email address
752b84549af89a08dbdd7fdb8b9568b5 journal.articlezxybnytfddd
Scitation: Automated noninvasive classification of renal cancer on multiphase CT