Investigation of optimization model for predicting immune checkpoint inhibitor treatment efficacy on contrast-enhanced computed tomography images of hepatocellular carcinoma using deep learning

Abstract

Background & Aims Although the use of immune checkpoint inhibitor (ICIs)-targeted agents for unresectable hepatocellular carcinoma (HCC) is promising, individual response variability exists. Therefore, we developed an artificial intelligence (AI)-based model to predict treatment efficacy using pre-ICI contrast-enhanced computed tomography (CT) imaging characteristics. Approach & Results We evaluated the efficacy of atezolizumab and bevacizumab or lenvatinib in 43 patients at the Nagasaki University Hospital from 2020–2022) using the modified Response Evaluation Criteria in Solid Tumors. A total of 197 Progressive Disease (PD), 271 Partial Response (PR), and 342 Stable Disease (SD) contrast CT images of HCC were used for training. We used ResNet18 as the Convolutional Neural Network (CNN) model and YOLOv7 as the You Only LookOnce (YOLO) model with precision-recall curves and class activation maps (CAMs) for diagnostic performance evaluation and model interpretation, respectively. The 3D t-distributed Stochastic Neighbor Embedding was used for image feature analysis. The CNN model had a PD prediction sensitivityof 84 %. However, the CAMs sites did not align with the tumors, suggesting the possible overlearning of non-tumor CT features. In contrast, the YOLO model demonstrated an area under the curve of 0.995 on the precision–recall curve for PD, providing accurate and clinically versatile predictions by identifying decisive points. Conclusions Preparing substantial training data for tumor drug effect prediction models is challenging compared to general tumor diagnosis models; hence, large-scale validation using an efficient YOLO model is warranted.