A METHOD for image plane re-registration in a transthoracic three-dimensional echocardiography system

Journal Article

Nearly all three-dimensional echocardiography (3DE) systems described acquire 2D images sequentially, requiring the subject to remain in the same position during data collection. This may be difficult for some patients, and a method for re-registering images despite patient movement during a 3DE study is desirable. METHODS: A 3DE system based on a modified HP Sonos 2500 with an electromagnetic scanhead tracker was used. Data can be displayed as a wireframe of hand-traced contours, and volumes are computed using a minimum-energy deformable shell model. In this study, images could be registered relative to a fixed point on the patient gantry or to a second tracking device affixed to the subject at the xiphoid. The "correction" method (registration of images relative to the subject) was tested on an ellipsoid-target 3D ultrasound phantom (CIRS, Norfolk, VA) and on 8 adult volunteers. Phantom: for each study, 10 images (subset A) widely surveying the target were acquired, the phantom was moved (translation and orientation) and 10 more images (subset B) acquired. After image re-registration, 25 volumes were determined using random combinations of 10 images drawn from subsets A and B. Differences relative to 3DE volume without motion were assessed by Wilcoxon Signed-Rank lest. Volunteers: 10 ECG-triggered ciné loops were acquired, patient was moved, and 10 additional loops acquired. End-diastolic and systolic volumes (EDV, ESV) and ejection fractions (EF) using all 20 loops were computed for uncorrected and corrected datasets and compared to breathhold ciné MRI values by limits of agreement analysis. RESULTS: Phantom: After correction, volumes computed using image planes obtained with translation and rotation of the phantom between image acquisitions did not differ significantly (p>0.10) from volumes without phantom motion. Also, the corrected wireframe models conformed to the known ellipsoidal shape of the phantom. Uncorrected datasets displayed as wireframes were clearly distorted. Uncorrected Corrected Uncorrected Corrected Volunteers MRI Range Mean Bias Mean Bias 95% C.I. 95% C.I. EDV (ml) 109-169 11.0 -3.1 20.6 7.7 ESV (ml) 40-58 3.7 -4.4 10.6 3.3 EF(%) 52-71 0.7 2.5 3.8 3.8 CONCLUSIONS: Registration of 2D image planes despite subject motion can be achieved using a second, subject-based tracking device in a 3DE system with electromagnetic tracking of position and orientation. Accuracy of phantom and human-LV volumes were improved using this method. Limitations have not been fully characterized, but some motions may be difficult to correct, as the relationship between position/orientation of the chest wall-mounted sensor and cardiac motion (particularly tilt/swing) is likely to vary on an individual basis. However, the method can improve accuracy of 3DE in the presence of patient motion.

Duke Authors

Cited Authors

  • Chuang, ML; Beaudin, RA; Riley, MF; Douglas, PS; Hibberd, MG

Published Date

  • December 1, 1997

Published In

Volume / Issue

  • 10 / 4

Start / End Page

  • 391 -

International Standard Serial Number (ISSN)

  • 0894-7317

Citation Source

  • Scopus