Deep feature learning for pulmonary nodule classification in a lung CT
Authors:
- 1 February 2016
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 28, 1-3
Abstract
In this paper, we propose a novel method of identifying pulmonary nodules in a lung CT. Specifically, we devise a deep neural network by which we extract abstract information inherent in raw hand-crafted imaging features. We then combine the deep learned representations with the original raw imaging features into a long feature vector. By taking the combined feature vectors, we train a classifier, preceded by a feature selection via t-test. To validate the effectiveness of the proposed method, we performed experiments on our in-house dataset of 20 subjects; 3,598 pulmonary nodules (malignant: 178, benign: 3,420), which were manually segmented by a radiologist. In our experiments, we achieved the maximal accuracy of 95.5%, sensitivity of 94.4%, and AUC of 0.987, outperforming the competing method.Keywords
This publication has 9 references indexed in Scilit:
- Automatic classification of pulmonary peri-fissural nodules in computed tomography using an ensemble of 2D views and a convolutional neural network out-of-the-boxMedical Image Analysis, 2015
- Lung Nodule Classification Using Deep Features in CT Images2015
- Latent feature representation with stacked auto-encoder for AD/MCI diagnosisBrain Structure and Function, 2013
- Evaluation of Patients With Pulmonary Nodules: When Is It Lung Cancer?Chest, 2007
- Reducing the Dimensionality of Data with Neural NetworksScience, 2006
- Texture information in run-length matricesIEEE Transactions on Image Processing, 1998
- Gradient-based learning applied to document recognitionProceedings of the IEEE, 1998
- Learning representations by back-propagating errorsNature, 1986
- Textural Features for Image ClassificationIEEE Transactions on Systems, Man, and Cybernetics, 1973