Comparison of visual interpretation and BRASS analysis in the localization of Epilepsy SPECT

• Main Authors: Huang Chia-Wen, 黃佳文
• Other Authors: Shih Ko-Nien
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/m5qw9w

碩士 === 元培科技大學 === 放射技術研究所 === 100 === Abstract Purpose: 99mTc-ECD Brain SPECT provides regional cerebral flow distributions in the evaluation of abnormal epileptic foci. In this study, we aimed to compare the sensitivity of detecting epileptic foci in 99mTc-ECD Brain SPECT images between visual interpretation and Hermes-BRASS analysis. The Filters (Butterworth; Metz) applied in reconstruction of SPECT data, and selection models of Hermes-BRASS (NMI; MI; CD) were taken into account as well. Besides, we would like to test the default parameter settings of Hermes-BRASS analysis, such as difference function threshold, template threshold, similarity function, and standard deviation. Materials and methods: In this study, we enrolled 24 patients with intractable epartial seizures. The causes of partial seizures were classified into those with brain tumors (9) and without brain tumors (15, including temporal sclerosis, encephalitis, trauma, heredity etc.).All patients received 99mTc-ECD Brain SPECT during interictal and postictal periods. The reconstructed SPECT images were examined by two nuclear medicine physicians via both visual interpretation and Hermes-BRASS analysis to detect epileptic foci. The standard of sensitivity in detecting epileptic foci was based on the results of other clinical examinations (MRI, EEG/Video, PET) and surgery. Results: (1) The sensitivity of detecting epileptic foci was as followed: visual interpretation, 51.54%, Hermes-BRASS analysis, NMI=82.25%, MI=70.94%, CD=57.99%. (2)Visual interpretation: BW filter (inteictal/postictal): 48.33%; 47.22%; Metz (interictal/ postictal):52.28%; 58.33%. Hermes BRASS analysis: BW filter (interictal/postictal): 65.89%; 63.20%；Metz filter(interictal/postictal): 75.23%; 77.45% (3) It was evident that the sensitivity of detecting epileptic foci in patients with brain tumor is higher than that without brain tumors in Hermes-BRASS analysis. There was significant difference by using Metz filter (p<0.05) (4)The best default parameter settings of Hermes-BRASS analysis were as follows: difference function threshold, 40%-50%; template threshold, 30-40%; standard deviation threshold, from 2 to 5. Conclusion: Hermes-BRASS analysis was more sensitive than visual interpretation in detecting epileptic foci in 99mTc-ECD Brain SPECT images. NMI model in Hermes-BRASS analysis and Metz filter were recommended.