Diagnostic Performance of Two‐Dimensional Versus Three‐Dimensional Transesophageal Echocardiography Images of Selected Pathologies Evaluated by Receiver Operating Characteristic Analysis

MAREK BELOHLAVEK, DAVID A. FOLEY, JAMES B. SEWARD, JAMES F. GREENLEAF

Research output: Contribution to journalArticlepeer-review

11 Scopus citations

Abstract

The sensitivity and specificity of 2‐D and 3‐D echocardiographic images for the detection of selected morphological abnormalities were compared using receiver operating characteristic (ROC) analysis. Five experienced clinical echocardiographers blinded to the patients' diagnoses evaluated the 20 original static 2‐D image sets and 20 corresponding 3‐D reconstructions using a five point categorical scale that ranged from definitely abnormal to definitely normal. The ROC curve for the 3‐D images was significantly (P lt; 0.05) closer to the ideal discrimination function than was the ROC curve for the 2‐D transesophageal images (i.e., the sensitivity of the 3‐D images was higher than that of the 2‐D sequential images at the same specificity). In conclusion: 3‐D transesophageal images provided better visual clues for the identification of morphological abnormalities than did serial 2‐D echocardiographic images despite the same input information in both image formats. The use of ROC analysis assisted in the comparison of these two imaging techniques.

Original languageEnglish (US)
Pages (from-to)635-645
Number of pages11
JournalEchocardiography
Volume11
Issue number6
DOIs
StatePublished - Nov 1994

Keywords

  • ROC analysis
  • three‐dimensional echocardiography
  • transesophageal echocardiography

ASJC Scopus subject areas

  • Radiology Nuclear Medicine and imaging
  • Cardiology and Cardiovascular Medicine

Fingerprint

Dive into the research topics of 'Diagnostic Performance of Two‐Dimensional Versus Three‐Dimensional Transesophageal Echocardiography Images of Selected Pathologies Evaluated by Receiver Operating Characteristic Analysis'. Together they form a unique fingerprint.

Cite this