Accurate and efficient polyp segmentation plays a critical role in colonoscopy and the early diagnosis of colorectal cancer. However, small polyps, which are widely present in clinical settings, often exhibit characteristics such as diminutive size, indistinct boundaries, and low contrast. These characteristics hinder the effectiveness of existing segmentation methods, leading to both high miss rates and poor segmentation performance, thereby increasing the risk of malignant transformation. Therefore, achieving high-precision segmentation of small polyps has become a key technical challenge in enhancing the accuracy and reliability of clinical diagnosis. To address this issue, we propose Interactive fusion and Guided-Selective Enhancement Network (IGSENet). From the perspective of multi-view feature perception, the Interactive Global-Regional Fusion (IGRF) module enables dynamic collaboration across semantic features, enhancing the integration of global context and local boundary information. The Guided Feature Enhancement (GFE) module precisely focuses on critical regional features while suppressing irrelevant information, thereby improving the response to target areas. Additionally, the Selective Detail Enhancement (SDE) module selectively strengthens the representation of uncertain regions, effectively mitigating issues caused by boundary ambiguity in small polyps. Extensive experiments demonstrate that the proposed method effectively overcomes the limitation posed by the underrepresentation of small polyps in training data, compared with the existing approaches, IGSENet achieves a notable improvement in segmentation accuracy for small polyps, with an mDice increase of 5.15% in ETIS, highlighting its superior generalization ability and strong clinical potential.