Predicting Pathology in Medical Decision Support Systems in Endoscopy of the Gastrointestinal Tract

Liedlgruber, Michael; Uhl, Andreas

doi:10.5772/16384

Cited by 2 publications

(1 citation statement)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To estimate the classification accuracies we use the leave-one-patient-out cross-validation (LOPO-CV) [45] in order to avoid any bias during classification. In this setup one image out of the database is considered as an unknown image.…”

Section: Methodsmentioning

confidence: 99%

Evaluation of different distortion correction methods and interpolation techniques for an automated classification of celiac disease

Gadermayr

Liedlgruber

Uhl

et al. 2013

Computer Methods and Programs in Biomedicine

View full text Add to dashboard Cite

Due to the optics used in endoscopes, a typical degradation observed in endoscopic images are barrel-type distortions. In this work we investigate the impact of methods used to correct such distortions in images on the classification accuracy in the context of automated celiac disease classification.For this purpose we compare various different distortion correction methods and apply them to endoscopic images, which are subsequently classified. Since the interpolation used in such methods is also assumed to have an influence on the resulting classification accuracies, we also investigate different interpolation methods and their impact on the classification performance. In order to be able to make solid statements about the benefit of distortion correction we use various different feature extraction methods used to obtain features for the classification.Our experiments show that it is not possible to make a clear statement about the usefulness of distortion correction methods in the context of an automated diagnosis of celiac disease. This is mainly due to the fact that an eventual benefit of distortion correction highly depends on the feature extraction method used for the classification.

show abstract

Section: Methodsmentioning

confidence: 99%