COMPUTERIZED METHOD OF LESION VOLUME QUANTITATION IN MULTIPLE-SCLEROSIS - ERROR OF SERIAL STUDIES

This study was designed to evaluate a potentially important source of error in T2-hyperintense lesion measurement unique to longitudinal mul tiple sclerosis treatment trials that would not be detected by the sta ndard intraobserver and interobserver error analyses. The effect of th is ''error of serial studies'' was tested by using the standard-of-ref erence manual-outlining approach and a modified bi-feature space (stat istical) approach applied to a database of five consecutive patients. To simulate the conditions of a longitudinal treatment trial, each pat ient had immediate repeat MR studies of the brain with imperfect head repositioning. The study hypothesis was confirmed that with an improve d quantitative methodology, the ''error of serial studies'' (interseri es error) would exceed the intraobserver error.