A Segmentation Method of Lung Tumor by using Adaptive Structural Deep Belief Network

Deep Learning has a hierarchical network architecture to represent the complicated feature of input patterns. In the previous research, the adaptive structure learning method of Deep Belief Network (Adaptive DBN) was developed, which can discover an optimal number of hidden neurons for given input data in a Restricted Boltzmann Machine (RBM) by neuron generation-annihilation algorithm, and can obtain appropriate number of hidden layers in DBN. In this paper, a new segmentation method using the Adaptive DBN was developed to extract lung tumor from 3D CT images. In the experiment, 156 cases collected by an open dataset in NSCLC Radiogenomics was used to evaluate our model. As a result, our model showed 0.884 dice coefficient for the test data, which was higher value than the 3D U-Net implemented on MONAI.