Prediction of Immune Checkpoint Inhibitors Treatment Response of Non-Small Cell Lung Cancer Patients from Serial Computed Tomography Scans Based on Global Self-Attention Mechanism

Abstract

Background: This study aimed to predict the response of NSCLC patients to immune checkpoint inhibitors (ICIs) by utilizing computed tomography (CT) images through deep learning methods.

Methods: The study retrospectively collected 624 serial CT images from 156 patients at Jiangsu Province Hospital, along with their clinical data.  The dataset was then divided into three cohorts: training (n=547), validation (n=64), and test (n=64).  Additionally, an external validation cohort included 37 CT images from 37 patients at Nanjing PuKou Peoples' Hospital, with comprehensive clinical data. Based on the modified Video Vision Transformer (ViViT) model with the global self-attention mechanism, patients who underwent ICIs treatment were analyzed to predict their response to immunotherapy. The performance of the ViViT model was evaluated using a confusion matrix and a receiver operating characteristic curve (ROC).

Results: The ViViT model exhibited predictive capabilities for ICIs response, with corresponding areas under the receiver operating characteristic curve (AUC) of 0.74 (95% CI: 0.69-0.78), 0.74 (95% CI: 0.61-0.86), 0.76 (95% CI: 0.62-0.88), and 0.69 (95% CI: 0.5-0.87) in the training, validation, test, and external validation sets, respectively. In the ViViT model, the overall accuracy was 0.74, 0.76, 0.75, and 0.68 in the training, validation, test, and external validation sets, respectively. The corresponding precision values were 0.80, 0.82, 0.81, and 0.63, recall values were 0.69, 0.75, 0.72, and 0.71, F1 values were 0.74, 0.78, 0.77, and 0.67, and specificity values were 0.80, 0.78, 0.78, and 0.65, respectively.

Conclusion: This study illustrates how a deep learning model provides a noninvasive means of predicting clinical outcomes for NSCLC patients undergoing immunotherapy. This model has the potential to significantly enhance personalized treatment strategies for lung cancer patients.