Ultrasound Image Segmentation using Latent Information
Ultrasound (US) imaging is widely employed for biometric measurement and diagnosis of internal organs due to its high availability and the fact it does not emit radiation. It is important to segment the objective anatomies from US images for different applications, e.g., extract the bone surface to do registration for preoperative images and real-time US images, visualize the object in 3D and automatically measure the key parameters for diagnosis.
The current learning-based approach often suffers from generalizability over data recorded from different machines and different patients. To address this problem, we want to classify the image features in the latent space; thereby, improving the understanding of anatomies information rather than pixel intensity.
Keywords: Ultrasound imaging segmentation, blood vessel segmentation, bone surface segmentation
Klinikum rechts der Isar (MRI) Clinic for Vascular and Endovascular Surgery Technical University of Munich (TUM)
- Dr.rer.nat. Angelos Karlas
Klinik für Gefäßchirurgie – Helios Klinikum München West
- Dr. med. Reza Ghotbi
- Prof. Nassir Navab
- Zhongliang Jiang
- Yuan Bi