Deep Learning Signature to Predict Postoperative Anxiety in Patients Receiving Lung

Qingqing Ji ^1* Guohua Zhou ² Xiangxiang Sun ³

• ¹ Shanghai University of Engineering Sciences, Shanghai, China
• ² Ningbo First Hospital, Ningbo, Zhejiang Province, China
• ³ Department of Thoracic Surgery, The First Affiliated Hospital of University of Science and Technology of China Anhui Provincial Hospital, Hefei, Anhui Province, China

This study aims on establishing and validate a deep learning signature based on magnetic resonance imaging (MRI) to predict postoperative anxiety in patients receiving lung cancer surgery. In the current study, 202 patients receiving lung cancer surgery were included.Preoperative MRI-T1WI images were collected to train the deep learning signature utilized the ResNet-152 algorithm. The relationships between clinical variables and postoperative anxiety were explored via Logistic regression and the predictive performances of the developed deep learning signature were evaluated via receiver operating characteristic analysis. Larger tumor size (odds ratio [OR], 2.044; 95% confidence interval [CI], 1.736-3.276; p=0.002) and occurrence of lymph node metastasis (OR, 2.078; 95% CI, 1.023-3.221; p=0.043) were revealed as independent predictors for postoperative anxiety. With the increase of deep learning scores, more patients experiencing postoperative anxiety were identified. Moreover, our deep learning signature yielded areas under the curve of 0.865 (95% CI, 0.800-0.930) and 0.822 (95% CI, 0.695-0.950) to predict postoperative anxiety. Therefore, our deep learning signature could help identify lung cancer patients with high risks of postoperative anxiety.