Gaussian Highpass Filters-based Convolutional Neural Network for Pulmonary Nodules Detection in CT Images
ACM International Conference Proceeding Series
Association for Computing Machinery
New York, NY
First page number:
Last page number:
The segmentation of various types of nodules in CT images presents various challenges due to a large amount of information that needs to be processed. In this study, we proposed a Gaussian highpass filter-based convolutional neural network(CNN) for the fully-automated detection of pulmonary nodules in CT scans. In medical image analysis, the dataset sizes are usually too small to train the network. Therefore, for each training data, a set of 2-D patches from differently oriented planes are extracted. The extracted datasets are used as inputs for the proposed framework which comprises multiple streams of 2-D CNN, and the obtained outputs are combined to produce the final classification. We evaluate this strategy on a test set of 888 CT scans and compare it with other CNN or published methodologies using the same dataset. The results indicate that the proposed framework offers significant performance gains over other methods.
Gaussian highpass filters; Convolutional neural network; Small sample; CT; Pulmonary nodules
Analytical, Diagnostic and Therapeutic Techniques and Equipment
Gaussian Highpass Filters-based Convolutional Neural Network for Pulmonary Nodules Detection in CT Images.
ACM International Conference Proceeding Series, 2018
New York, NY: Association for Computing Machinery.