To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
Accurate Detection of Pulmonary Nodule and False Positive Reduction with Faster R-CNN and ResNet Models |
|
Author |
Taghreed Alzahrani, Suhare Solaiman |
| Citation |
Vol. 25 No. 2 pp. 1-10 |
|
Abstract |
The annual incidence of lung cancer has significantly increased. The detection of pulmonary nodule is critical for the diagnosis of lung cancer. In recent years, deep learning models have demonstrated encouraging results, surpassing traditional machine learning methods across different fields Researchers have used several deep learning methods to enhnace the effectiveness of computer-aided detection (CAD) systems that use computed tomography (CT) images for lung cancer diagnosis. Despite the capability of deep-learning models to detect pulmonary nodules, they still encounter significant challenges in terms of a high rate of false positive. In this study, we propose a model for pulmonary nodule detection and false positive reduction using CT images. This model combines a faster region-based convolutional neural network (Faster R-CNN) for accurate detection and a residual network (ResNet50) to reduce the false positive rate. Our model was evaluated using the public LUng Nodule Analysis (LUNA16) dataset, and the experimental results revealed that our model outperformed existing methods in the literature regarding competition performance metric (CPM) at 95.1%. An additional evaluation, the10-Fold Cross-Validation approach, was used to evaluate the proposed model, which demonstrated the reliability and capability of the generalization of our model, with an average accuracy of 91.1%. |
|
Keywords |
Keywords: Pulmonary nodule detection; false positive reduction; Faster R-CNN model; ResNet model; CT images; CAD |
|
URL |
