<p>Landslide susceptibility assessment is essential for preventing disasters and minimizing casualties and economic losses. However, traditional models rely on static data, which fail to comprehensively capture surface deformation features, thereby compromising evaluation accuracy and reliability. This study introduces a Multi-Scale Convolutional Neural Network (MSCNN) coupling model that integrates time-series Interferometric Synthetic Aperture Radar (InSAR) data, focusing on Zhen’an County, Shangluo City, Shaanxi Province. We combined geohazard survey data with deformation rate data obtained through Small Baseline Subset-InSAR (SBS-InSAR) technology and fifteen evaluation indices, including elevation, slope, aspect, lithology, and rainfall, to construct Information Value (IV) indices. The coupled IV-MSCNN model was developed to enhance the accuracy of landslide susceptibility assessment. Evaluation indices were screened using covariance and correlation analyses, and model performance was assessed using metrics such as AUC, Recall, Precision, and F-score. The MSCNN model (AUC = 0.9434) outperformed independent models including Support Vector Machine (SVM), Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Conventional Convolutional Neural Network (CNN). Furthermore, coupled models (IV-MSCNN, IV-SVM, IV-RF, IV-XGBoost, and IV-CNN) consistently outperform their corresponding independent counterparts, with IV-MSCNN achieving the highest AUC of 0.9835. Upon incorporating InSAR deformation factors, the AUC of IV-MSCNN further improved to 0.9873, resulting in more accurate susceptibility zoning. In summary, the IV-MSCNN dynamic landslide susceptibility assessment model, enriched with InSAR deformation factors, provides a robust reference for local landslide disaster prevention and management.</p>