MRI自动图像分割技术对多发硬化（MS）病灶容积测量的初步研究

Objective To improve automated technique in MS lesion volume measurement. MaterialsandMethods To compare lesion volume of multiple channels automated segmentation with that of manual segmentation in different sequences of 3D high resolution MRI: (MPRAGE, T2WI, FLAIR and DIR). Results the sensitivity of M+T2/M+T2+D, M+D/M+D+F, M+F/M+F+T2 and M+F+T2+D were 41.8/40.4%, 33.3/64.8%, 63.4/62.5%, 74.4%; the specificity were 82.2/87.1%, 93.9/99.1%, 99.0/98.6%, 97.0% and the accuracy were 82.1/86.9%, /93.6/98.9%, 98.9/98.5%, 96.9%. Conclusion In automated segmentation, the four channels automated segmentation is the best measurement way in MS lesion volume.