Detection of Lung Contour with Closed Principal Curve and Machine Learning

Abstract

Radiation therapy plays an essential role in the treatment of cancer. In radiation therapy, the ideal radiation doses are delivered to the observed tumor while not affecting neighboring normal tissues. In three-dimensional computed tomography (3D-CT) scans, the contours of tumors and organs-at-risk (OARs) are often manually delineated by radiologists. The task is complicated and time-consuming, and the manually delineated results will be variable from different radiologists. We propose a semi-supervised contour detection algorithm, which firstly uses a few points of region of interest (ROI) as an approximate initialization. Then the data sequences are achieved by the closed polygonal line (CPL) algorithm, where the data sequences consist of the ordered projection indexes and the corresponding initial points. Finally, the smooth lung contour can be obtained, when the data sequences are trained by the backpropagation neural network model (BNNM). We use the private clinical dataset and the public Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) dataset to measure the accuracy of the presented method, respectively. To the private dataset, experimental results on the initial points which are as low as 15% of the manually delineated points show that the Dice coefficient reaches up to 0.95 and the global error is as low as 1.47 × 10. The performance of the proposed algorithm is also better than the cubic spline interpolation (CSI) algorithm. While on the public LIDC-IDRI dataset, our method achieves superior segmentation performance with average Dice of 0.83.

Hybrid Automatic Lung Segmentation on Chest CT Scans

“…Considering the optimal performance of the CPL-BNN, we set the hidden neurons to be 10 [23]. At the same time, considering the co-influence of time and accuracy on the model, we set DBN to two hidden layers, one with 25 neurons, and the other with 30 neurons.…”

“…, S) and b 2,k (k = 1, 2) are the thresholds of the ith hidden neuron and the k-th output neuron, respectively. TABLE 1 shows the parameters used in this model, where # denotes the parameters validated in this paper, and * denotes the parameters discussed in our previous work [23].…”

“…In addition to DSC and , two volumebased metrics, we used ASSD and MSD as surface-based metrics for evaluation. In order to compare performance of different methods under similar conditions, we used the Closed Polygonal Line and Backpropagation Neural Network Model (CPL-BNNM) [23] and Hull-Closed Polygonal Line Method (Hull-CPLM) [44] methods. We trained the CPL-BNNM and Hull-CPLM with the same 50 cases as the PSCCL-CH-CPC discussed in Section III-B.…”

“…Recently, the principal curve model has attracted interests for segmenting abnormal organs from other neighboring normal structures, because of its strong abilities to effectively deal with noise input and obtain robust results [21]. You et al [22] have proposed a semi-supervised system based on the principal curve method for lung CT image segmentation, and Peng et al [23] have devised a principal curve model to detect lung contours; both of them detect tissue contours with high accuracy. Furthermore, machine learning models can be used as classifiers for early diagnosis [24].…”

A fully automatic segmentation algorithm for CT lung images based on random forest

“…Chen et al focused on pathological lung segmentation in clinical three‐dimensional (3D) low‐dose computed tomography images and proposed a sparse shape composition to reconstruct 3D surfaces of pathological lungs. Peng et al proposed a semi‐supervised lung contour detection algorithm. They took a set of points as initial input, extracted lung contours with a closed polyline algorithm and smoothed lung contours with a back‐propagation neural network model.…”