Registration Methods for IVUS: Transversal and Longitudinal Transducer Motion Compensation

• Published in: IEEE transactions on biomedical engineering Vol. 64; no. 4; pp. 890 - 903
• Main Authors: Talou, Gonzalo D. Maso, Blanco, Pablo J., Larrabide, Ignacio, Bezerra, Cristiano Guedes, Lemos, Pedro A., Feijoo, Raul A.
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.04.2017; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Arteriosclerosis; Atherosclerosis; Axial strain; Cardiac-Gated Imaging Techniques - methods; Catheters; Compensation; Correlation; Frames; Gating; Heart; Humans; Image Enhancement - instrumentation; Image Enhancement - methods; Image segmentation; intravascular ultrasound (IVUS); Kinematics; Maximum likelihood estimation; Maximum likelihood estimators; Medical instruments; Motion; Motion compensation; Optimization; Pattern Recognition, Automated - methods; Phantoms, Imaging; Registration; Reproducibility of Results; Sensitivity analysis; Sensitivity and Specificity; Spatial data; Speckle; Subtraction Technique; Tissues; Transducers; Ultrasonic imaging; Ultrasonography, Interventional - instrumentation; Ultrasonography, Interventional - methods; Ultrasound; ultrasound (US)
• ISSN: 0018-9294; 1558-2531
• DOI: 10.1109/TBME.2016.2581583

Objective: Intravascular ultrasound (IVUS) is a fundamental imaging technique for atherosclerotic plaque assessment, interventionist guidance, and, ultimately, as a tissue characterization tool. The studies acquired by this technique present the spatial description of the vessel during the cardiac cycle. However, the study frames are not properly sorted. As gating methods deal with the cardiac phase classification of the frames, the gated studies lack motion compensation between vessel and catheter. In this study, we develop registration strategies to arrange the vessel data into its rightful spatial sequence. Methods: Registration is performed by compensating longitudinal and transversal relative motion between vessel and catheter. Transversal motion is identified through maximum likelihood estimator optimization, while longitudinal motion is estimated by a neighborhood similarity estimator among the study frames. A strongly coupled implementation is proposed to compensate for both motion components at once. Loosely coupled implementations (DLT and DTL) decouple the registration process, resulting in more computationally efficient algorithms in detriment of the size of the set of candidate solutions. Results and Conclusions: The DTL outperforms DLT and coupled implementations in terms of accuracy by a factor of 1.9 and 1.4, respectively. Sensitivity analysis shows that perivascular tissue must be considered to obtain the best registration outcome. Evidences suggest that the method is able to measure axial strain along the vessel wall. Significance: The proposed registration sorts the IVUS frames for spatial location, which is crucial for a correct interpretation of the vessel wall kinematics along the cardiac phases.