Discrimination of Tumorous Intracerebral Hemorrhage from Benign Causes Using CT Densitometry

Abstract

Background and purpose: Differentiation of tumorous intracerebral hemorrhage from benign etiology is critical in initial treatment plan and prognosis. Our aim was to investigate the diagnostic value of CT densitometry to discriminate tumorous and nontumorous causes of acute intracerebral hemorrhage.

Materials and methods: This retrospective study included 110 patients with acute intracerebral hemorrhage classified into 5 groups: primary intracerebral hemorrhage without (group 1) or with antithrombotics (group 2) and secondary intracerebral hemorrhage with vascular malformation (group 3), brain metastases (group 4), or primary brain tumors (group 5). The 5 groups were dichotomized into tumorous (groups 4 and 5) and nontumorous intracerebral hemorrhage (groups 1-3). Histogram parameters of hematoma attenuation on nonenhanced CT were compared among the groups and between tumorous and nontumorous intracerebral hemorrhages. With receiver operating characteristic analysis, optimal cutoffs and area under the curve were calculated for discriminating tumorous and nontumorous intracerebral hemorrhages.

Results: Histogram analysis of acute intracerebral hemorrhage attenuation showed that group 1 had higher mean, 5th, 25th, 50th, and 75th percentile values than groups 4 and 5 and higher minimum and 5th percentile values than group 2. Group 3 had higher 5th percentile values than groups 4 and 5. After dichotomization, all histogram parameters except maximum and kurtosis were different between tumorous and nontumorous intracerebral hemorrhages, with tumors having lower cumulative histogram parameters and positive skewness. In receiver operating characteristic analysis, 5th and 25th percentile values showed the highest diagnostic performance for discriminating tumorous and nontumorous intracerebral hemorrhages, with 0.81 area under the curve, cutoffs of 34 HU and 44 HU, sensitivities of 65.6% and 70.0%, and specificities of 85.0% and 80.0%, respectively.

Conclusions: CT densitometry of intracerebral hemorrhage on nonenhanced CT might be useful for discriminating tumorous and nontumorous causes of acute intracerebral hemorrhage.

Fig 1.

Flowchart of study population enrollment. IVH indicates intraventricular hemorrhage.

Fig 2.

Representative cases of semiautomatic segmentation of tumorous (A and B) and nontumorous (C and D) ICHs and their relative frequency histogram of hematoma attenuation (E). ICH was segmented with a semiautomatic method based on a voxel-intensity threshold of 40–130 HU. On histogram analysis, the 5th and 25th percentile values were 26 HU and 38 HU in tumorous ICHs and 40 HU and 65 HU in nontumorous ICHs, respectively, and were discriminated correctly by using our histogram analysis.

Fig 3.

Differences among the 5 groups in histogram parameters of ICH attenuation are list on the x-axis of each boxplot: 1) primary ICH without antithrombotics, 2) primary ICH with antithrombotics, 3) secondary ICH due to vascular malformation, 4) secondary ICH due to brain metastasis, and 5) secondary ICH due to primary brain tumor.

Fig 4.

Differences in histogram parameters of ICH attenuation in tumor and nontumorous ICHs. All histogram parameters except maximum value and kurtosis were significantly different between tumorous and nontumorous ICHs. Tumors had lower cumulative histogram parameters and positive skewness. NT indicates nontumorous ICHs; T, tumorous ICHs.

Fig 5.

Receiver operating characteristic curves of histogram parameters for discrimination of tumor and nontumorous ICHs.