| Summary: | 博士 === 國立中興大學 === 電機工程學系所 === 103 === This dissertation develops a volume sphering analysis (VSA) approach to tissue classification and volume calculation of multispectral magnetic resonance (MR) brain images. It processes all multispectral MR image slices as an image cube while using only one set of training samples obtained from a single multispectral image slice to perform image analysis. In order to make the selected one slice set of training samples also applicable to other MR image slices an extrapolation algorithm is particularly designed for this purpose. This significantly reduces tremendous burden on radiologists’ selection of training samples as well as computational cost. In this work, we propose the following two different experiments by VSA approach. One is a supervised classification, Supervised Volume Sphering Analysis (SVSA) techniques were considered and analyzed by experiments for MR classification where the required complete knowledge of each MR tissue substance was obtained from the prior knowledge based on ground truth and their anatomical structures. The other is an unsupervised classification, Unsupervised Volume Sphering Analysis (UVSA) techniques could automatically classify healthy brain images with no inputs from operators and the results would be operator-independent. To further resolve instability and inconsistency issues resulting from a single slice set of training samples, an iterative Fisher’s linear discriminant analysis (IFLDA) is also developed to be coupled with SVSA or UVSA to improve the traditional slice-by-slice MR image classification. Experimental results demonstrate that SVSA or UVSA using one set of training samples in conjunction with IFLDA not only performs comparably to approaches using training samples from individual image slices but also saves significant time of selecting training samples and computational cost.
|
|---|
