Predictive value of a novel Asian lung cancer screening nomogram based on artificial intelligence and epidemiological characteristics. Issue 23 (28th October 2021)

• Record Type: Journal Article
• Authors: Liu, Dahai; Sun, Xiao; Liu, Ao; Li, Lun; Li, Shaoke; Li, Jinmiao; Liu, Xiaojun; Yang, Yu; Wu, Zhe; Leng, Xiaoliang; Wo, Yang; Huang, Zhangfeng; Su, Wenhao; Du, Wenxing; Yuan, Tianxiang; Jiao, Wenjie
• Is Part Of: Thoracic cancer. Volume 12:Issue 23(2021)
• Page Start: 3130
• Page End: 3140
• Publication Date: 2021-10-28
• Subjects: artificial intelligence -- Asians lung cancer screening -- convolutional neural networks -- epidemiological characteristics -- nomogram; Chest -- Cancer -- Periodicals; Chest -- Cancer -- Treatment -- Periodicals; Chest -- Surgery -- Periodicals; 616.99494005
• DOI: 10.1111/1759-7714.14140
• Languages: English
• ISSNs: 1759-7706

Abstract

Abstract: Background: To develop and validate a risk prediction nomogram based on a deep learning convolutional neural networks (CNN) model and epidemiological characteristics for lung cancer screening in patients with small pulmonary nodules (SPN). Methods: This study included three data sets. First, a CNN model was developed and tested on data set 1. Then, a hybrid prediction model was developed on data set 2 by multivariable binary logistic regression analysis. We combined the CNN model score and the selected epidemiological risk factors, and a risk prediction nomogram was presented. An independent multicenter cohort was used for model external validation. The performance of the nomogram was assessed with respect to its calibration and discrimination. Results: The final hybrid model included the CNN model score and the screened risk factors included age, gender, smoking status and family history of cancer. The nomogram showed good discrimination and calibration with an area under the curve (AUC) of 91.6% (95% CI: 89.4%–93.5%), compare with the CNN model, the improvement was significance. The performance of the nomogram still showed good discrimination and good calibration in the multicenter validation cohort, with an AUC of 88.3% (95% CI: 83.1%–92.3%). Conclusions: Our study showed that epidemiological characteristics should be considered in lung cancer screening, which can significantly improve the efficiency of the artificial intelligence (AI) model alone. We combined the CNN model score with Asian lung cancer epidemiological characteristics to develop a new nomogram to facilitate and accurately perform individualized lung cancer screening, especially for Asians. Abstract : A novel AI model was developed for LDCT detection specifically. Our study showed that epidemiological characteristics should be considered in lung cancer screening, which can significantly improve the efficiency of AI model alone. We combined the CNN model score with Asian lung cancer epidemiological characteristics to develop a new Nomogram to facilitate and accurately perform individualized lung cancer screening, especially for Asians.