Spatiotemporal model improves needle tip detection in ultrasound videos
STUNet-Fusion: Spatiotemporal Needle-Tip Localization in Ultrasound Video via Multi-Channel Motion Fusion
Computer Vision and Pattern Recognition
Summary
Finding the needle tip in ultrasound images is hard because it can look faint or be blocked, and other things in the image can look similar. The authors created a method called STUNet-Fusion that combines information about how the picture looks, how things move, and changes between frames over time. Their system uses a neural network to learn both the shape and movement patterns, then predicts exactly where the needle tip is in each video frame. Tests show this approach is more reliable than older methods.
What this means in practice
- •For medical device developers: Integrate accurate needle-tip tracking into ultrasound-guided intervention tools to improve precision during procedures.$Commercial implications: This paper enables creation of ultrasound devices with enhanced needle guidance systems improving procedure safety and efficiency.
- •For medical imaging software engineers: Develop software modules that automatically highlight needle tips in ultrasound video streams to assist clinicians in real time.
Authors
Chia-Chi Hsu, Chia-Hsuan Hsu, Che-Chou Shen
Abstract
Needle-tip localization in ultrasound remains challenging because the needle may appear weak, discontinuous, or partially invisible, while imaging artifacts and anatomical structures can produce similar responses. To address this problem, we propose STUNet-Fusion, a spatiotemporal framework for needle-tip localization in ultrasound videos. The proposed method formulates the input as a tri-channel spatio-temporal fusion tensor, comprising grayscale appearance, grid-based motion feature, and raw frame difference. A shared ResNet-34 encoder extracts spatial features, ConvLSTM integrates temporal dependencies, and a U-Net decoder reconstructs a dense probability heatmap. The final coordinates are extracted via a soft-argmax operation to achieve sub-pixel localization accuracy. Experimental results demonstrate that this spatiotemporal fusion strategy significantly improves localization robustness compared to conventional baselines.