Anterior chamber angle analysis, especially the appositional angle indicating the early reversibility stage of angle-closure glaucoma, is important for the diagnosis and treatment of the disease. Most existing studies only analyzed the open and closed angle status based on static anterior segment optical coherence tomography (AS-OCT) images, while ignoring the clinically important appositional angle. In this paper, we propose a Dual-path Video Information Aggregation Network (DVIA-Net) based on AS-OCT dynamic video to achieve accurate classification of open, appositional, and synechial angles. Specifically, first, based on the clinical correlation of the above three angle states, we designed a dual-path video information architecture for the feature extraction and aggregation of the iris motion and the open-closed angle stationary status. Secondly, we use a dynamic region recognition (DRR) module to emphasize the salient features during motion. In addition, the open-closed stationary angle information is modeled by the video key-frame structure feature extraction (KSFE) path. Finally, the features of motion and angle stationary status are aggregated to achieve accurate classification. To verify the effectiveness of DVIA-Net, we collected an AS-OCT dynamic video dataset that records the changes in chamber angle state. Compared with the state-of-the-art methods, experimental results show that the proposed DVIA-Net not only achieves the best overall performance but also makes an improvement on the appositional angle classification.