Statistical Computing 2002: Beiträge
نویسندگان
چکیده
The Heidelberg Retina Tomograph (HRT) provides topographic images of the optic nerve head by scanning the eye with a 670nm laser. As a precondition to medical analysis, the optic nerve head has to be outlined manually in clinical practise. Swindale et al. (2000) suggest automated classification of HRT images by a non-linear approximation of the image using the fitted parameters of a non-linear model. To evaluate the ability of this method to classify glaucomatous and normal eyes, we use a simulation model, which allows to create topography images of the optic nerve head in various forms (normal, glaucomatous and various progressive forms). We compare LDA, classification trees, bagging (Breiman, 1996), and bagging combined classifiers (Hothorn & Lausen, 2002). The methods are illustrated by data of a case control study (Mardin et al., 2002). We discuss possibilities for an improvement and optimization of image analysis for classification.
منابع مشابه
Beiträge zum effizienten Multimediaretrieval
Inhaltsverzeichnis 1 Einleitung 1 2 Suche in polyphoner Musik 9 2.
متن کاملThree New Systematic Approaches for Computing Heffron-Phillips Multi-Machine Model Coefficients (RESEARCH NOTE)
This paper presents three new systematic approaches for computing coefficient matrices of the Heffron-Phillips multi-machine model (K1, …, K6). The amount of computations needed for conventional and three new approaches are compared by counting number of multiplications and divisions. The advantages of new approaches are: (1) their computation burdens are less than 73 percent of that of convent...
متن کاملInvestigating electrochemical drilling (ECD) using statistical and soft computing techniques
In the present study, five modeling approaches of RA, MLP, MNN, GFF, and CANFIS were applied so as to estimate the radial overcut values in electrochemical drilling process. For these models, four input variables, namely electrolyte concentration, voltage, initial machining gap, and tool feed rate, were selected. The developed models were evaluated in terms of their prediction capability with m...
متن کامل