Deep Learning-assisted Real-time Wall Shear Stress Measurement on Chicken Embryo Heart Using Spectral Domain Optical Coherence Tomography
Baihang Song,Huiwen Jiang,Jian Liu,Yao Yu,Jingmin Luan,Yuqian Zhao,Yi Wang,Jingyuan Zhang,Zhao Liu,Ning Zhang,Xin Zhu,Zhenhe Ma
DOI: https://doi.org/10.1109/tim.2024.3417602
IF: 5.6
2024-01-01
IEEE Transactions on Instrumentation and Measurement
Abstract:Congenital heart disease, the most common birth defect in newborns and children, highlights the significance of understanding heart development. In the early development stage, the biomechanical environment, especially wall shear stress (WSS), plays a crucial role in heart morphogenesis. The outflow tract (OFT) is an important segment of the embryonic heart, and a large portion of congenital heart defects originate in the OFT. However, real-time measurement of WSS in the OFT of animal models remains challenging. We propose an automatic-localization segmentation network (ALSegNet) integrated with spectral domain optical coherence tomography (SD-OCT) to achieve real-time WSS calculation. Our ALSegNet accurately extracts the blood flow area from the SD-OCT structure image by incorporating an automatic localization module. WSS calculation is performed with the combination of extracted flow area and flow velocity information provided by SD-OCT. The network achieves segmentation within 22.2 ms, making the total WSS calculation time approximately 36 ms for each B-scan. Our approach achieves a speed of 27 frames/s, adequate for video display. Using the developed system, we successfully monitor in vivo WSS in the OFT of chicken embryos, providing valuable insights into embryonic heart development. The integration of deep learning with SD-OCT enables real-time and accurate WSS measurements, offering a powerful tool for studying congenital heart disease and heart morphogenesis.