Self-Supervised Content-Consistent Contrast Enhancement Framework for Ultrafast Plane-Wave Imaging

Ultrafast ultrasound imaging based on plane waves (PWs) has attracted increasing attention owing to its high frame rate. However, the image quality provided by this technology, particularly in image contrast, does not meet clinical standards. Although the compounding method solves the issue of image quality degradation, it compromises the advantage of the high frame rate. Deep-learning-based methods have emerged as a promising avenue for improving the image quality of ultrasound imaging. However, their heavy reliance on high-quality training data poses challenges in practical implementation. Considering the imaging characteristic of B-mode images, in this article, we propose a self-supervised content-consistent contrast enhancement framework to enhance the contrast of B-mode ultrasound PW images. The proposed method consists of a content-preserving module to reconstruct a single image while preserving pixel information by maximizing the mutual information between the pixels of the reconstructed image and the input image, and a contrast-enhancing module to improve the contrast of the reconstructed image using the enhanced lower DR version of the input image by aligning their local perceptual information under multiple receptive fields. Experimental results demonstrate that our method yields improved image quality in terms of contrast ratio (CR), contrast-to-noise ratio (CNR), and generalized CNR (gCNR), when compared with the compounding method and other learning-based reconstruction methods that necessitate high-quality image training. Notably, the assessment conducted on the PICMUS dataset shows the commendable generalization capabilities of the proposed method. The code of our implementation is available at https://github.com/ Wang-aohua/SC3E.