原研多光谱智能分析仪诊断肺腺癌浸润程度的诊断性分析

Background and objective Lung cancer is one of the most common malignant tumors in the world. The accuracy of intraoperative frozen section (FS) in the diagnosis of lung adenocarcinoma infiltration cannot fully meet the clinical needs. The aim of this study is to explore the possibility of improving the diagnostic efficiency of FS in lung adenocarcinoma by using the original multi-spectral intelligent analyzer. Methods Patients with pulmonary nodules who underwent surgery in the Department of Thoracic Surgery, Beijing Friendship Hospital, Capital Medical University from January 2021 to December 2022 were included in the study. The multispectral information of pulmonary nodule tissues and surrounding normal tissues were collected. A neural network model was established and the accuracy of the neural network diagnostic model was verified clinically. Results A total of 223 samples were collected in this study, 156 samples of primary lung adenocarcinoma were finally included, and a total of 1,560 sets of multispectral data were collected. The area under the curve (AUC) of spectral diagnosis in the test set (10% of the first 116 cases) of the neural network model was 0.955 (95%CI: 0.909-1.000, P<0.05), and the diagnostic accuracy was 95.69%. In the clinical validation group (the last 40 cases), the accuracy of spectral diagnosis and FS diagnosis were both 67.50% (27/40), and the AUC of the combination of the two was 0.949 (95%CI: 0.878-1.000, P<0.05), and the accuracy was 95.00% (38/40). Conclusion The accuracy of the original multi-spectral intelligent analyzer in the diagnosis of lung invasive adenocarcinoma and non-invasive adenocarcinoma is equivalent to that of FS. The application of the original multi-spectral intelligent analyzer in the diagnosis of FS can improve the diagnostic accuracy and reduce the complexity of intraoperative lung cancer surgery plan.