Results of the ImageCLEF 2006 - Medical Automatic Annotation Task

Here are the results of the medical automatic annotation task of ImageCLEF 2006.

In 2006, 27 groups registered for the medical automatic annotation task and 12 of these submitted a total of 27 runs (maximal number of submissions per group 5).

The task description can be found on the task website.

The following table gives the error rates (i.e. number of misclassified images/number of classified images, i.e. 1-(classification accuracy)) for the submitted runs.

Group ID	Runtag	Error rate
RWTHi6	SHME	16.2
UFR	UFR-ns-1000-20x20x10	16.7
RWTHi6	SHSVM	16.7
MedIC_CISMeF	local+global_PCA335	17.2
MedIC_CISMeF	local_PCA333	17.2
MSRA	WSM-msra_wsm_gray	17.6
MedIC_CISMeF	local+global_PCA450	17.9
UFR	UFR-ns-800-20x20x10	17.9
MSRA	WSM-msra_wsm_patch	18.2
MedIC_CISMeF	local_PCA150	20.2
RWTHi6	IDM	20.4
rwth_mi	opt	21.5
rwth_mi	baseline	21.7
cindi	cindi-svm-sum	24.1
cindi	cindi-svm-product	24.8
cindi	cindi-svm-ehd	25.5
cindi	cindi-fusion-KNN9	25.6
cindi	cindi-svm-max	26.1
OHSU	OHSU_iconGLCM2_tr	26.3
OHSU	OHSU_iconGLCM2_tr_de	26.4
NCTU	dblab-nctu-dblab2	26.7
MU	I2R_refine_SVM	28.0
OHSU	OHSU_iconHistGLCM2_t	28.1
ULG	SYSMOD-RANDOM-SUBWINDOWS-EX	29.0
DEU	DEU-3NN-EDGE	29.5
OHSU	OHSU_iconHist_tr_dev	30.8
UTD	UTD	31.7
ULG	SYSMOD-RANDOM-SUBWINDOWS-24	34.1

if you want your group id or your runtag or the link to your site to be changed, please contact me with the necessary details.

Further analysis and details of the results will be made available soon. So far, we provide the correct classification for the test data.

Thomas Deselaers

Last modified: Fri Jun 30 11:17:57 CEST 2006

;