Pattern Recognition of Head MRI Images and It's 3-D Display

点击翻译按钮获取中文摘要

We proposed a pattern recognition of multichannel MR images and it's 3-D display method for supporting diagnosis and surgical plannning. The effectiveness was shown using various types of multichannel MR images. 3-D display of MR images allows better perception of the continuity of structures and spatial relationships than a display of individual slices. But a boundary judgement of soft tissues, like tumor, is difficult even for expert doctor. Proper classification of normal or abnormal soft tissues is sought. We use multiple 3-D MR images of the same anatomical section using different pulse sequences to improve the accuracy of tissue extraction. The extraction methods proposed here consist of nonlinear combination methods of multichannel MR images a fuzzy clustering of the mapped images and a connectivity algorithm of soft tissue clusters of interest. The shading of soft tissue cluster uses a volume rendering. The shading model is improved to realize a powerfully interactive 3-D display. PD-weighted images. TI-weighted images. T2-weighted images and MRA images of human brain are numerically analyzed. Organs of interest are 3-D displayed and 3-D fabricated. A complicated discrimination of tumor, blood vessels, white matter, gray matter, CSF and spatial relationships among the human brain can be recognized for supporting diagnosis and surgical planning.