Multimodal medical image data provides a comprehensive view of patient conditions, and its effective utilization can enhance clinical decision-making. Traditional clustering methods face challenges in handling complex, complementary, and heterogeneous data. In this paper, we propose a novel method named IMVSC, which provides an integrative architecture for simultaneously optimizing the anchor learning to learn feature representations, graph construction to evaluate data similarity across views, and cluster partitioning to obtain the clustering results. Additionally, we incorporate adaptive variance selection based on the Bayesian Information Criterion, which can retain more informative features during the clustering process. Our IMVSC method exhibits outstanding clustering performance, achieving accuracy, normalized mutual information, purity, and F-score values of 97.09%, 88.38%, 97.16%, and 94.87%, respectively. These results highlight the potential of IMVSC in advancing medical decision-making and facilitating more precise analysis of multimodal medical image data.