Ensemble detection of colorectal lesions for CT colonography

Authors:
Janne J. Näppi;Daniele Regge;Hiroyuki Yoshida
Affiliations:
Department of Radiology, Massachusetts General Hospital and Harvard Medical School, Boston, MA;Institute for Cancer Research and Treatment, Turin, Italy;Department of Radiology, Massachusetts General Hospital and Harvard Medical School, Boston, MA
Venue:
MICCAI'11 Proceedings of the Third international conference on Abdominal Imaging: computational and Clinical Applications
Year:
2011

Citing 2
Cited 1

Random Forests

Machine Learning
Combining Pattern Classifiers: Methods and Algorithms

Combining Pattern Classifiers: Methods and Algorithms

Comparative performance of state-of-the-art classifiers in computer-aided detection for CT colonography

MICCAI'12 Proceedings of the 4th international conference on Abdominal Imaging: computational and clinical applications

Quantified Score

Hi-index	0.00

Visualization

Abstract

Even though different computer-aided detection (CAD) systems for computed tomographic colonography (CTC) have similar overall detection accuracies, they are known to detect different types of lesions and false positives. We implemented an ensemble CAD scheme for merging the detection results of different CAD systems in CTC. After normalizing of the lesion-likelihood data between different systems, a Bayesian classifier was used for determining the final detections. For evaluation, we collected 218 abnormal patients with 263 lesions ≥6 mm. The detection accuracies of three CAD systems were compared with that of their ensemble CAD scheme by use of independent training and testing. The preliminary results indicate that the ensemble CAD scheme can yield a higher overall detection accuracy than can individual CAD systems. In particular, the ensemble scheme was able to detect flat lesions at high sensitivity without compromising a high polyp detection accuracy.