Deep Learning-based Artificial Intelligence Improves Accuracy of Error-prone Lung Nodules

Abstract

Introduction: Early detection of lung cancer is one way to improve outcomes. Improving the detection of nodules on chest CT scans is important. Previous artificial intelligence (AI) modules show rapid advantages, which improves the performance of detecting lung nodules in some datasets. However, they have a high false-positive (FP) rate. Its effectiveness in clinical practice has not yet been fully proven. We aimed to use AI assistance in CT scans to decrease FP. Materials and methods: CT images of 60 patients were obtained. Five senior doctors who were blinded to these cases participated in this study for the detection of lung nodules. Two doctors performed manual detection and labeling of lung nodules without AI assistance. Another three doctors used AI assistance to detect and label lung nodules before manual interpretation. The AI program is based on a deep learning framework. Results: In total, 266 nodules were identified. For doctors without AI assistance, the FP was 0.617-0.650/scan and the sensitivity was 59.2-67.0%. For doctors with AI assistance, the FP was 0.067 to 0.2/scan and the sensitivity was 59.2-77.3% This AI-assisted program significantly reduced FP. The error-prone characteristics of lung nodules were central locations, ground-glass appearances, and small sizes. The AI-assisted program improved the detection of error-prone nodules. Conclusions: Detection of lung nodules is important for lung cancer treatment. When facing a large number of CT scans, error-prone nodules are a great challenge for doctors. The AI-assisted program improved the performance of detecting lung nodules, especially for error-prone nodules.

Keywords: CT images; artificial intelligence; lung nodules.

Figure 1

Framework for pulmonary nodule detection. The 3D CNN is a pre-trained fully supervised detector that serves as the detector backbone to extract nodule proposals and features in weakly supervised settings. In addition to image-level labels to predict the pseudo labels for each proposal, this model additionally observed nodule numbers and slice index information from EMR to guide the learning process. Abbreviations: 3D CNN: 3-Dimentional convolutional neural network; NMS, non-maximum suppression; RoI pooling, region of interest pooling; FC layer, fully connected layer; ReLU, rectified linear unit; MIL, multiple instance learning; P.S., pseudo labels.

Figure 2

Overall nodular detection.

Figure 3

Left, central and right lung fields and nodular detection. Number of pulmonary nodular detection in left, center and right lung fields. False positive and sensitivity of nodular detection in left, center and right lung fields.

Figure 4

Upper, middle and lower lung fields and nodular detection. (A) Number of pulmonary noular detection in upper, milddle and lower lung fields. (B) False positive and sensitivity of nodular detection in upper, milddle and lower lung fields.

Figure 5

Nodular size and nodular detection. (A) Number of difference sizes of pulmonary nodules. (B) False positive and sensitivity of nodular detection in difference sizes of pulmonary nodules.

Figure 6

Nodular texture and nodular detection. Number of difference textures of pulmonary nodules. False positive and sensitivity of nodular detection in difference textures of pulmonary nodules.