Biomedical sensor image segmentation algorithm based on improved fully convolutional network
Biomedical sensor image segmentation algorithm based on improved fully convolutional network
复制标题
DOI：
10.1016/j.measurement.2022.111307
复制
发表时间：
2022-05-16
期刊：
影响因子：
5.6
通讯作者：
Yang, Meng
中科院分区：
文献类型：
--
作者：
Li, Hongan;Fan, Jiangwen;Yang, Meng
有效利用生物医学传感器图像，可以清晰地定位病变组织和组织结构，临床诊断和治疗可以帮助医生制定合适的治疗方案。为了对生物医学传感器采集的图像进行高效处理，提出了一种改进的全卷积网络生物医学传感器图像分割方法，该方法首先提取生物医学传感器采集图像的局部空间和频域信息，增强图像的纹理信息；其次，通过增加目标区域权重抑制背景干扰，对图像进行细化处理，增强图像特征，同时减少信息冗余；实验证明，本文模型能有效减少图像分割后细胞粘附现象，具有较好的分割效果和分割精度，能更有效地利用生物医学传感器获取的图像。
Effective use of biomedical sensor image can help locate diseased tissues and tissue structures clearly presented, and clinical diagnosis and treatment can assist doctors in making appropriate treatment plans. In order to efficiently process the images acquired by biomedical sensors, we propose a biomedical sensor image segmentation method with improved fully convolutional network, which firstly extracts the local spatial and frequency domain information of the images acquired by biomedical sensors and enhances the texture information of the images. Secondly, the background interference is suppressed by increasing the target region weights to refine the processing of the image and enhance the features of the image while reducing the information redundancy. It is experimentally proved that the model in this paper can effectively reduce the phenomenon of cell adhesion after image segmentation, has better segmentation effect and segmentation accuracy, and can more effectively utilize the images acquired by biomedical sensors.