Object: The authors' aim was to assess the ability of apparent diffusion coefficient (ADC) ratios in distinguishing brain abscesses from cystic or necrotic neoplasms.
Methods: Fifty-three patients with rim-enhancing masses in the brain observed on T1-weighted MR images were included: 26 had abscesses (14 bacterial, six nonbacterial, and six of unknown origin), 11 had glioblastoma multiforme, and 16 had rim-enhancing metastasis. The ADC values, derived from diffusion-weighted imaging, were measured in the most homogeneous portion of the cystic component of the mass. The ADC ratios were calculated by dividing the ADC values from the nonenhancing cystic portion of the mass by the ADC values from contralateral normal-appearing white matter. Lesions were further differentiated based on presence, absence, or incompleteness of a T2 hypointensity rim. The mean (+/- standard deviation) ADC ratios were significantly higher in neoplasms than in abscesses (2.45 +/- 0.91 compared with 1.12 +/- 0:53, p < 0.01). The accuracy of ADC ratios in discriminating abscesses from neoplasms, determined by the area under the receiver operating characteristic curve (Az), was high: 0.91 +/- 0.04 (mean +/- standard error of the mean [SEM]). The threshold of 1.7 was associated with highest efficiency (87%) in discriminating abscesses from neoplasms. If only bacterial abscesses were analyzed compared with neoplasms, the Az increased to 0.96 +/- 0.03 (SEM). Using ADC ratios and T1 rim characteristics, 50 of 53 lesions were correctly classified (efficiency 94.3%).
Conclusions: The accuracy of ADC ratios in discriminating brain abscesses from cystic or necrotic neoplasms is very high and can be further improved using T2 rim characteristics.