False Positive Reduction in Lung Computed Tomography Images using Convolutional Neural Networks

• DOI: 10.48550/arxiv.1811.01424
• Publication Date: 2018-01-01
• Authors: Görkem Polat; Uğur Halıcı; Yeşim Serinağaoğl...
• Keywords: Nodule (geology)

Recent studies have shown that lung cancer screening using annual low-dose computed tomography (CT) reduces mortality by 20% compared to traditional chest radiography. Therefore, CT has started be used widely all across the world. However, analyzing these images is a serious burden for radiologists. In this study, we propose novel and simple framework analyzes screenings convolutional neural networks (CNNs) false positives. Our shows even non-complex architectures are very powerful classify 3D nodule data when methods. We also use different fusions in order show their power effect on overall score. CNNs preferred over 2D because 3D, operations may result information loss. Mini-batch overcome class-imbalance. Proposed been validated according LUNA16 challenge evaluation got score of 0.786, which average sensitivity values at seven predefined positive (FP) points.