Volume 23, Issue 3 pp. 401-408
Clinical Investigative Study

A New Classification Scheme for Spinal Vascular Abnormalities based on Angiographic Features

Adnan I. Qureshi MD

Adnan I. Qureshi MD

From the Zeenat Qureshi Stroke Research Center, Department of Neurology, University of Minnesota, Minneapolis, Minnesota.

Search for more papers by this author
First published: 10 December 2012
Citations: 6
Adnan I. Qureshi, MD, Department of Neurology, University of Minnesota, 12–100 PWB, 516 Delaware Street SE, Minneapolis, MN 55455. E-mail: [email protected].

J Neuroimaging 2013;23:401-408.

Abstract

ABSTRACT

BACKGROUND AND PURPOSE

To determine the interobserver reliability of a newly proposed classification scheme for angiographic classification of spinal vascular malformations including arteriovenous fistulas (AVFs) and arteriovenous malformations (AVMs).

METHOD

A study was performed done in which 1–2 representative angiographic images of 26 spinal AVFs and/or AVMs were independently classified by five fellows in the ACGME accredited Endovascular Surgical Neuroradiology (ESN) program and two external interventionalists in the absence of any other clinical or imaging data. From these observations the interobserver reliability for each category and the overall scheme were determined in terms of the median weighted kappa statistic.

RESULTS

The overall interobserver reliability for the new classification scheme was a Kappa of 0.53 (Z= 21.3, P= <.0001) among the seven raters. The Kappa for individual grades was as follows: grade I (k= 0.66), grade II (k= 0.50), grade III (k= 0.44), and grade IV (k= 0.58). Three or more raters agreed on 100% of the cases. The interobserver reliability was high among the two practicing interventionalist raters (k= 0.55, 95% confidence interval 0.3–0.8). The interobserver reliability remained high among junior ESN fellows (k= 0.65).

CONCLUSION

The new classification scheme provided satisfactory reliability even in the hands of less experienced observers. The scheme can be used with minimal training and other concurrent data and can be relied upon to provide consistent results.

The full text of this article hosted at iucr.org is unavailable due to technical difficulties.