Risk and success prediction analysis with artificial intelligence applications of lung cancer surgical applications and results

NSCLC is the most common histopathological type among lung cancers. 8.TNM staging is used in NSCLC staging. Surgical treatment is effective in early stage and resectable locally advanced NSCLC groups. Lung resection and systematic mediastinal lymph node dissection form the basis of NSCLC surgery. Postoperative pulmonary complications, intensive care admission, long-term hospital stay are important in terms of morbidity, mortality, nosocomial infections, cost, bed turnover rate and social aspects. Artificial intelligence applications are increasingly taking place in medical practice. Although its main areas of use are electronic recording and storage of data, learning and teaching practices, diagnostic aid algorithms are used in fields such as radiology, pathology, and dermatology, based on the principle of image processing. In our study, we aimed to determine the clinical data and postoperative processes of patients who underwent surgical resection for NSCLC using artificial intelligence modeling. In our study, preoperative data of 953 patients who underwent surgical resection for NSCLC were evaluated and postoperative clinical processes were predicted with high accuracy and success values through artificial intelligence algorithms. The F1 average value of the test data for postoperative pulmonary complication development using deep learning algorithms was 86.4%. The area under the curve in the ROC curve was 84%. With deep learning algorithms, the F1 average value of the test data for postoperative intensive care admission requirement was 82.3%. The area under the curve in the ROC curve was 83%. With deep learning algorithms, the F1 average value of the test data for postoperative hospital stay of more than 5 days was 67.1%. The area under the curve in the ROC curve was 83%. With deep learning algorithms, the F1 average value of the test data for postoperative hospital stay of less than 9 days was 70.3%. The area under the curve in the ROC curve was 82%. The R value of the test data for the number of days of postoperative hospital stay with deep learning algorithms was 68.2%. Using artificial intelligence algorithms, postoperative clinical predictions after NSCLC surgery can be made with high accuracy and success and will be used frequently in our clinical practice in the future.