<p>Wireless Sensor Networks (WSNs) are widely employed for event monitoring and tracking, however energy efficiency remains a primary challenge due to the limited battery capacity of sensor nodes and uneven energy depletion, especially near the sink, causing the energy-hole problem. To address this, an Enhanced Cluster Head Selection and Energy-Efficient Routing approach using Dual-path Multi-scale Attention Guided Network in a WSN environment (DMAGN-CHS-WSN) is proposed. In the proposed method, sensor nodes are randomly deployed in a 2D Cartesian plane. Clustering is performed, and a Cluster Head (CH) is selected using a multi-objective fitness function based on factors such as energy, delay, distance, density, and throughput. DMAGN is employed to enhance CH selection by focusing on critical features through attention mechanisms. For efficient routing, the Leaf in Wind Optimization Algorithm (LWOA) is applied to define the optimal data path from sensor nodes to base station. The proposed method is evaluated based on performance metrics including the number of alive nodes, delay, packet drop, throughput, normalized energy consumption and Packet Delivery Ratio (PDR). Results show that DMAGN-CHS-WSN outperforms existing methods such as ECHS-WSN-LACH, CBRP-WSN-RL, and MOCH-WSN-AMGNN, achieving up to 27.12% higher throughput, 33.02% longer network lifetime, and 31.19% lower energy consumption. These outcomes demonstrate the method’s effectiveness in enhancing WSN performance and energy efficiency.</p>