Abstract
Ultrafast Plane-Wave (PW) imaging often produces artifacts and shadows thatvary with insonification angles. We propose a novel approach using ImplicitNeural Representations (INRs) to compactly encode multi-planar sequences whilepreserving crucial orientation-dependent information. To our knowledge, this isthe first application of INRs for PW angular interpolation. Our method employsa Multi-Layer Perceptron (MLP)-based model with a concise physics-enhancedrendering technique. Quantitative evaluations using SSIM, PSNR, and standardultrasound metrics, along with qualitative visual assessments, confirm theeffectiveness of our approach. Additionally, our method demonstratessignificant storage efficiency, with model weights requiring 530 KB compared to8 MB for directly storing the 75 PW images, achieving a notable compressionratio of approximately 15:1.