Multimodal Aspect-Based Sentiment Analysis (MABSA) is a fine-grained sentiment analysis task that aims to identify and analyze sentiment polarities toward specific aspects within multimodal data. Existing approaches to MABSA primarily focus on aspect-oriented image-text alignment as the foundation for aspect detection and sentiment prediction. However, they often overlook the granularity discrepancy between two modalities at the instance level and fail to capture potential inter-aspect correlations. To address these limitations, we propose a novel framework called the Category-Aware Hypergraph Network (CAHN). Specifically, we introduce a Category-Guided Attention Module (CGAM) that aligns image and text features at the instance level with their corresponding semantic categories, effectively mitigating cross-modal noise caused by irrelevant content. In addition, we design an Aspect-Interaction Hypergraph Network (AIHN) that dynamically captures higher-order correlations among aspect terms, enabling more informed sentiment reasoning through structured inter-aspect relationships. Extensive experiments on MABSA benchmarks demonstrate that CAHN consistently outperforms existing state-of-the-art approaches.