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.
A novel computer-aided lung nodule detection system for CT images
Rent:
Rent this article for
USD
10.1118/1.3633941
/content/aapm/journal/medphys/38/10/10.1118/1.3633941
http://aip.metastore.ingenta.com/content/aapm/journal/medphys/38/10/10.1118/1.3633941

Figures

Image of FIG. 1.
FIG. 1.

Top level block diagram of the proposed CAD system.

Image of FIG. 2.
FIG. 2.

Performance evaluation of the nodule detector in our CAD system for the training set, at different DNG threshold values (see Table I) applied to “isolated/juxtavascular; juxtapleural” nodules.

Image of FIG. 3.
FIG. 3.

(a) Example of a nodule with agreement level 4 (arrow); (b) the points where the DNG is maximal, detected at six different scales; (c) the nodule clusters detected with the nodule enhancement filter; (d) the nodule clusters selected with the DNG seed points in accordance with parameters specified in Sec. ???; (e) the DNG computed at s = 2.5 mm; (f) the divergence of the gradient (or Laplacian) computed at s = 2.5 mm.

Image of FIG. 4.
FIG. 4.

(a) Example of a nodule with agreement level 4 (arrow) in the picture; (b) the nodule clusters combined with the DNG seed points; (c) feature L vv is a good ridge or vessel detector; (d) feature L uu is good for distinguishing nodules; as expected, L uu and L vv both produce strong (in this case, negative) responses in the presence of nodules whereas L vv produces a strong response and L uu a weak response in the presence of blood vessels.

Image of FIG. 5.
FIG. 5.

FROC of FD-NEAT for nodules of the independent test set at four agreement levels.

Image of FIG. 6.
FIG. 6.

FROC of fixed-topology ANN for nodules of the independent test set at four agreement levels.

Image of FIG. 7.
FIG. 7.

FROC of SVM classifier for nodules of the independent test set at four agreement levels.

Image of FIG. 8.
FIG. 8.

FROC of SVM, fixed-topology ANN, FD-NEAT for nodules with agreement level 1.

Image of FIG. 9.
FIG. 9.

FROC of SVM, fixed-topology ANN, FD-NEAT for nodules with agreement level 2.

Image of FIG. 10.
FIG. 10.

FROC of SVM, fixed-topology ANN, FD-NEAT for nodules with agreement level 3.

Image of FIG. 11.
FIG. 11.

FROC of SVM, fixed-topology ANN, FD-NEAT for nodules with agreement level 4.

Image of FIG. 12.
FIG. 12.

Nodules from the LIDC database that are detected by our CAD system at (a) agreement level 1, (b) agreement level 2, (c) agreement level 3, and (d) agreement level 4.

Image of FIG. 13.
FIG. 13.

Examples of some of the nodules that are missed by the detection stage of our CAD system at (a) agreement level 1, (b) agreement level 2 and (c) agreement level 3 (only one nodule missed at agreement level 4).

Image of FIG. 14.
FIG. 14.

Examples of some of the FPs that are included by our CAD system.

Tables

Generic image for table
TABLE I.

Specific divergence threshold values applied to isolated/juxtavascular nodules (divvals) and to juxtapleural nodules (divjux) at the points on the graph displayed in Fig. 2.

Generic image for table
TABLE II.

Overview of all features that were computed for the classifier of our CAD system.

Generic image for table
TABLE III.

CAD system sensitivities of FD-NEAT, the fixed-topology ANN and the SVM classifiers at an FP rate of 4/scan for nodules at the four agreement levels.

Generic image for table
TABLE IV.

Performance comparison of our CAD system (using the fixed-topology ANN classifier) with other methods. The right-most columns display the sensitivities of our CAD system at agreement levels 2–4, operating at the average number of FPs/case of the benchmark system listed in column 3.

Generic image for table
TABLE V.

Performance comparison of our CAD system (using the fixed-topology ANN classifier) with other methods, for results reported at agreement level 1. The right-most column displays the sensitivity of our CAD system at agreement level 1, operating at the average number of FPs/case of the benchmark system listed in column 3.

Loading

Article metrics loading...

/content/aapm/journal/medphys/38/10/10.1118/1.3633941
2011-09-22
2014-04-18
Loading

Full text loading...

This is a required field
Please enter a valid email address
752b84549af89a08dbdd7fdb8b9568b5 journal.articlezxybnytfddd
Scitation: A novel computer-aided lung nodule detection system for CT images
http://aip.metastore.ingenta.com/content/aapm/journal/medphys/38/10/10.1118/1.3633941
10.1118/1.3633941
SEARCH_EXPAND_ITEM