Development of U-net Neural Network for Biomedical Images with Big Data

Convolutional neural networks (CNNs) have revolutionized medical image analysis in recent years. Among these architectures, the U-net neural network (UNN) stands out as a widely recognized model in the field of image segmentation and reconstruction. UNN has achieved significant success in handling large volumes of medical images. Initially proposed as a basic architecture comprising standard convolution layers, pooling layers, and up-sampling layers, UNN has evolved to incorporate additional components such as full convolution, residual networks, and attention mechanisms, making it suitable for various medical image applications. However, challenges persist in addressing inverse image problems, reconstruction, segmentation, and quantification in the field of biomedical imaging, proving to be exceedingly complex to resolve. In this review, we outline the present state and progress of UNN in biomedical image analysis while exploring potential research directions that can enhance its clinical applicability utilizing big data.