Two-Step Rigid and Non-Rigid Image Registration for the Alignment of Multi-View Three-Dimensional Echocardiography Sequences

• Author / Creator

  Shanmuganathan, Srivathsan

• Ultrasound is a widely used imaging modality, which provides continuous real-time imaging of the human heart, brain, liver, and many other organs. Accurate cardiovascular evaluation plays an important role in early disease diagnosis and echocardiography plays an important role in the assessment of cardiovascular diseases. The lack of ionizing radiation and portability make echocardiography one of the safest imaging modalities. Although two-dimensional echocardiography is widely used to obtain the motion of the heart structures in real-time, real-time 3D echocardiography (RT3DE) imaging allows better 3D imaging by extracting spatial features along with temporal information, thus improving clinical decision making. There have been technological advances; however, the majority of acquired RT3DE images tend to be of low quality, characterized by the absence of anatomical information, decreased spatial and temporal resolution, speckle noise, and a limited field of view. By registering RT3DE images obtained from several sonography windows, it is possible to enhance the recognition of structures and achieve a substantial improvement in image quality as well as it is also useful in the fusion of echo images to image the entire heart.

  This study proposes a fully automatic point-based rigid registration technique, followed by nonrigid B-spline registration, to align 4D echocardiogram images acquired from various sonographic windows. The methodology was evaluated using scans acquired from seven volunteers. The accuracy of registration was visually and quantitatively assessed by delineating the left ventricle in each scan and computing the Dice score overlap metric and the Hausdorff distance mutual proximity measure between the first scan and the rest. The overall findings demonstrate that the suggested registration method improves image alignment compared to the initial scans, which might be helpful in the fusion of echocardiographic images.

• Subjects / Keywords

  • Echocardiography
  • Image registration
  • Left ventricular
  • Reinforcement learning

• Graduation date

  Fall 2024

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/r3-jw0c-eb67

• License

  This thesis is made available by the University of Alberta Library with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.