This paper proposes a medical image segmentation network called SEA-NET, which integrates spiral squeeze-and-excitation and attention modules. The model is designed to address the challenges of medical image segmentation, particularly in handling imbalanced foreground and background data, where small target features may be lost during multiple convolutions. SEA-NET uses squeeze-and-excitation modules to adjust channel information and attention modules to adjust spatial information, enhancing the segmentation of small targets. The model also incorporates a hybrid loss function (cross entropy + Tversky loss) to improve convergence and handle imbalanced data. The model was tested on two datasets: the brain MRI dataset LPBA40 and peripheral blood smear images. On LPBA40, the Dice coefficient reached 98.1%, indicating high accuracy. On the peripheral blood smear dataset, the model effectively segmented adhesion cells. Experimental results showed that SEA-NET outperformed U-Net, U-Net++, and other state-of-the-art methods in medical image segmentation. The model's structure includes an encoder-decoder framework with two parallel skip paths. The encoder captures multi-scale features, while the decoder integrates these features using attention and SE-Res paths. The attention path enhances the target area by adjusting the weight ratio of the target and non-target regions, while the SE-Res path adjusts channel information to retain useful semantic features. The hybrid loss function ensures fast convergence and effective handling of imbalanced data. The model was evaluated using metrics such as Dice coefficient, sensitivity, and specificity. Results showed that SEA-NET achieved high accuracy and balanced sensitivity and specificity. The model was also validated through ablation studies, demonstrating that the attention path, SE-Res path, and hybrid loss function significantly improved performance. In conclusion, SEA-NET provides an effective solution for medical image segmentation, particularly for small target segmentation in imbalanced data. The model's design addresses the limitations of existing methods, offering improved accuracy and robustness in medical image analysis.This paper proposes a medical image segmentation network called SEA-NET, which integrates spiral squeeze-and-excitation and attention modules. The model is designed to address the challenges of medical image segmentation, particularly in handling imbalanced foreground and background data, where small target features may be lost during multiple convolutions. SEA-NET uses squeeze-and-excitation modules to adjust channel information and attention modules to adjust spatial information, enhancing the segmentation of small targets. The model also incorporates a hybrid loss function (cross entropy + Tversky loss) to improve convergence and handle imbalanced data. The model was tested on two datasets: the brain MRI dataset LPBA40 and peripheral blood smear images. On LPBA40, the Dice coefficient reached 98.1%, indicating high accuracy. On the peripheral blood smear dataset, the model effectively segmented adhesion cells. Experimental results showed that SEA-NET outperformed U-Net, U-Net++, and other state-of-the-art methods in medical image segmentation. The model's structure includes an encoder-decoder framework with two parallel skip paths. The encoder captures multi-scale features, while the decoder integrates these features using attention and SE-Res paths. The attention path enhances the target area by adjusting the weight ratio of the target and non-target regions, while the SE-Res path adjusts channel information to retain useful semantic features. The hybrid loss function ensures fast convergence and effective handling of imbalanced data. The model was evaluated using metrics such as Dice coefficient, sensitivity, and specificity. Results showed that SEA-NET achieved high accuracy and balanced sensitivity and specificity. The model was also validated through ablation studies, demonstrating that the attention path, SE-Res path, and hybrid loss function significantly improved performance. In conclusion, SEA-NET provides an effective solution for medical image segmentation, particularly for small target segmentation in imbalanced data. The model's design addresses the limitations of existing methods, offering improved accuracy and robustness in medical image analysis.