Automatic stroke lesion segmentation from brain MRI images is crucial for computer-aided diagnosis, enabling physicians to make timely and accurate medical decisions. However, lesion segmentation is challenging due to variations in lesion size, location, number, and irregular boundaries. To address these challenges, this paper proposes a quadruple attention mechanism that leverages global self-attention and multi-scale triplet attention modules to enhance feature extraction. This quadruple module is integrated into the encoder and decoder of the U-Net architecture. The proposed quadruple attention mechanism improves the localization of both local and global features, resulting in better lesion segmentation than existing attention mechanisms. Additionally, a CBAM attention module is incorporated into the U-Net&rsquos bottleneck, improving the retention of small-scale features in the final layer. Evaluation on the publicly available ISLES22 dataset shows that the proposed method outperforms various state-of-the-art methods for stroke lesion segmentation. Furthermore, the lightweight U-Net architecture with approximately 2 million parameters surpassed most existing methods, demonstrating the efficiency of the proposed attention mechanism. Implementing such an efficient method will pave the way for automated and reliable computer-aided diagnosis, ultimately reducing the burden on limited medical facilities and improving patient diagnostic outcomes.