Time-aware deep neural networks for needle tip localization in 2D ultrasound

This paper proposes a novel algorithm for needle tip localization during challenging ultrasound-guided insertions when the shaft may be invisible, and the tip has a low intensity. The approach uses a hybrid deep neural network comprising a convolutional neural network and long short-term memory recurrent units to predict needle tip location using spatiotemporal information from consecutive ultrasound frames. Evaluation on an ex vivo dataset with 17G and 22G needles in bovine, porcine, and chicken tissue yields a tip localization error of 0.52 ± 0.06 mm and shows 30% improvement in accuracy compared to prior methods.