Accurate segmentation of the medical image is essential for timely clinical diagnosis and treatment planning. However, many high-performing models are computationally intensive, limiting their use in settings with resource scarcity; such as in rural clinics, mobile diagnostic units, or point of care systems. To bridge this gap, we propose EcoMamba-Net, a lightweight yet powerful deep learning (DL) architecture designed for efficient medical image segmentation. At its core are the novel EcoMambaBlocks, which are streamlined modules inspired by the state-space modeling of Mamba. They effectively capture long-range dependencies with minimal overhead. The model integrates depthwise convolutions and a hybrid attention mechanism to combine spatial and channel information. This enables rich feature extraction while maintaining an extremely low parameter count of just 0.5 million. EcoMamba-Net is specifically suitable for medical image segmentation due to its ability to preserve fine-grained anatomical boundaries, handle low inter-class variance, and remain robust to modality-specific noise, while running efficiently on limited clinical hardware. Despite its compact size, EcoMamba-Net achieves high performance – with a Dice Similarity Coefficient of 91.2% in the ISIC 2018 skin lesion dataset and 89.6% in the Kvasir-SEG dataset for polyp segmentation; thereby, making it a practical and scalable solution for real-world clinical environments. GitHub Link: https://github.com/maityanubhab/EcoMamba-Net. git