A strategy for magnetic resonance imaging of the head: results of a semi-empirical model. Part II.

A semi-empirical model was used to identify specific pulse sequences that cause most lesions to appear distinctly brighter than normal tissues in magnetic resonance (MR) images of the head. Clinical trials confirm the utility of these sequences for patient screening. As a result, a strategy for effective and efficient MR imaging of the head is proposed. The previously described gray-scale model has been modified to account for the effect of image noise. By means of computer simulation, 13,800 different hypothetical cerebral lesions were imaged with a variety of pulse sequences. A number of conclusions resulted. First, two sequences are expected to be sufficient to visualize most intracranial lesions, a "diagonal" SE sequence (e.g., SE 2500/80) and an IR sequence with a short inversion time (e.g., IR 1800/200). These sequences are orthogonal, i.e., lesions missed by one are likely to be detected by the other. Second, signal averaging the screening sequences is expected to be more effective than optimized sequences when lesion tissue parameters differ little from brain. Finally, the effectiveness of unaveraged screening sequences suggests that improved signal-to-noise ratio (SNR) is not necessary for the detection of most large lesions. Therefore, the increased SNR achievable through signal averaging or increased field strength might best be utilized to improve spatial resolution so that smaller lesions can be detected.