An oblique cylinder contrast-adjusted (OCCA) phantom to measure the accuracy of MRI brain lesion volume estimation schemes in multiple sclerosis

A new OCCA phantom using Oblique Cylinders and Contrast Adjustment for measuring the accuracy of brain lesion volume estimation schemes is described. It uses obliquely oriented cylinders made from acrylic rod, mounted in a water bath, to give realistic partial volume errors. Image intensities are inverted, scaled, and shifted, and noise is added, to form images that have realistic values of lesion-white matter contrast (5-30%) and contrast-to-noise ratio (3-20%). Artificial gray matter, CSF (cerebrospinal fluid), and scalp lipid are added because these bright areas may determine how the gray level display window is set. The performance of manual and contouring methods for estimating lesion volume was measured for three observers and nine lesions with individual volumes from 0.3 to 6.2 ml. There was a large variation, depending on the choice of method, the observer, and the lesion contrast. Volumes were usually overestimated, with the error increasing at high contrasts. The average error in estimating total lesion volume was 17% (range -16% to +30%). The OCCA phantom may have a role in training observers to improve their accuracy (and hence inter- and intraobserver reproducibility).