Efficient Medical Instrument Detection in 3D Volumetric Ultrasound Data

Ultrasound-guided procedures have been applied in many clinical therapies, such as cardiac catheterization and regional anesthesia. Medical instrument detection 3D Ultrasound (US) is highly desired, but the existing approaches are far from real-time performance. Our objective to investigate an efficient method US for practical use. We propose a novel Multi-dimensional Mixed Network US, which extracts discriminating features at full-image level by encoder, then applies specially designed dimension reduction block reduce spatial complexity of feature maps projecting space into 2D space. A decoder adopted detect along specified axes. By predicted outputs, detected or visualized volume. Furthermore, enable network better learn discriminative information, we multi-level loss function capture both pixel- image-level differences. carried out extensive experiments on two datasets tasks: (1) catheter RF-ablation (2) needle show that our proposed achieves error 2-3 voxels with efficiency about 0.12 sec per The 3-8 times faster than state-of-the-art methods, leading results has significant value US-guided intervention.