Performance of ROC Curves and LVQ Neural Networks in Detection of Middle Cerebral Artery Spasm
Transcranial color-coded Doppler ultrasonography (TCCD) is used by radiologists for detection of middle cerebral artery spasm. In this study, TCCD examination was performed in 100 patients, refereed by neurosurgeons for arterial angiography. Angiographic images were used to establish the four-grade scale of spasm, which was combined with TCCD blood flow measurements. The overall performance of TCCD in detection of MCA narrowing was estimated by means of Receiver Operating Characteristics (ROC). The ROC analysis proved very high efficiency (over 90%) of TCCD for detection of moderate-to-severe spasm. The multi-parameter classification of TCCD data was also performed by LVQ networks. LVQ networks showed very good accuracy (80–95%), not only for binary outcome, but also for classification of TCCD data in a multi-grade scale of spasm.
KeywordsReceiver Operating Characteristic Middle Cerebral Artery Receiver Operating Characteristic Receiver Operating Characteristic Curve Learn Vector Quantization
Unable to display preview. Download preview PDF.
- 2.Kohonen, T. (1987) Self-Organization and Associative Memory, 2nd edition. Springer-Verlag, BerlinGoogle Scholar
- 5.Sloan, M.A., Wozniak, M.A., Macko R.F. (1999) Monitoring of vasospasm after subarachnoid hemorrhage, in: V.L. Babikian, L.R. Wechsler, eds., Transcranial Doppler ultrasonography. Butterworth-Heinemann, Boston 109–127Google Scholar
- 6.Zweig, M.H., Campbell, G. (1993) Receiver operating characteristics (ROC) plots: a fundamental tool in clinical medicine. Clin. Chem. 39, 561–577Google Scholar