The accuracy of signal intensity measurements in magnetic resonance imaging as evaluated within the knee.

Quantitative signal intensity measurements are being utilized in both clinical and research magnetic resonance imaging protocols. This paper addresses three questions in quantitative MRI measurements as evaluated within the knee: 1) the accuracy of quantitative measurements; 2) improvement of accuracy by phantom normalization; and 3) the amount of signal change that is clinically significant. Seven normal subjects were imaged on three different days within a 1-wk period. Test-tube phantoms of manganous chloride (MnCl2) were imaged posterior to the knee and were used to normalize each image. The variation in signal intensity within the same subject averaged 20% for both the anterior cruciate ligament (ACL) and the posterior cruciate ligament (PCL). The phantom variation was approximately 18%. Signal intensity normalization by background subtraction, background division, phantom division, or a combination of subtraction and division did not significantly improve either the phantom variation or the ligament variation. Given that an individual ligament intensity will be measured with standard errors of +/- 20% of its value, we calculated the minimum increase in signal intensity to be considered abnormal relative to a normal ligament. A relative signal increase of 46% can be considered pathologic with 95% confidence. These findings emphasize that quantitative measurements must be carefully assessed when being applied in clinical settings.