Baseline Results for the ImageCLEF 2006 Medical Automatic Annotation Task

Abstract

The ImageCLEF 2006 medical automatic annotation task encompasses 11,000 images from 116 categories, compared to 57 categories for 10,000 images of the similar task in 2005. As a baseline for comparison, a run using the same classifiers with the identical parameterization as in 2005 is submitted. In addition, the parameterization of the classifier was optimized according to the 9,000/1,000 split of the 2006 training data. In particular, texture-based classifiers are parallel combined with classifiers, which use spatial intensity information to model common variabilities among medical images. However, all individual classifiers are based on global features, i.e. one feature vector describes the entire image. The parameterization from 2005 yields an error rate of 21.7%, which ranks 13 th among the 28 submissions. The optimized classifier yields 21.4% error rate (rank 12), which is insignificantly better.

Discriminative cue integration for medical image annotation

“…The run was not submitted to the challenge, but given its results it would have ranked 26th among 28 real submissions. Güld et al (2006) used the texture features proposed in Tamura et al (1978) and Castelli et al (1998) image. They also used a down-scaled representation of the original images evaluating the cross correlation function and the image distortion model value as distance measure.…”

Cue Integration for Medical Image Annotation

“…State of the art approaches used texture-based descriptors as features and discriminative algorithms, mainly SVMs, for the classification step [3,4]. Local and global features, have been used separately or combined together in multi-cue approaches with disappointing results [3,5]. Still, years of research on visual recognition showed clearly that multiple-cue methods outperform single-feature approaches, provided that the features are complementary.…”