Congress:
ECR24
Poster Number:
C-11624
Type:
EPOS Radiologist (scientific)
Authorblock:
A. Hertel1, F. Haag1, F. Tollens1, S. O. Schönberg1, M. Frölich1, S. Waldeck2, D. P. Overhoff1; 1Mannheim/DE, 2Koblenz/DE
Disclosures:
Alexander Hertel:
Nothing to disclose
Florian Haag:
Nothing to disclose
Fabian Tollens:
Nothing to disclose
Stefan Oswald Schönberg:
Nothing to disclose
Matthias Frölich:
Nothing to disclose
Stephan Waldeck:
Nothing to disclose
Daniel Pasqual Overhoff:
Nothing to disclose
Keywords:
Artificial Intelligence, Computer applications, Urinary Tract / Bladder, CT, CT-Quantitative, Computer Applications-Detection, diagnosis, Calcifications / Calculi
Especially in the monoenergetic reconstructions with low kev values (40 and 70kev) a differentiation of the different kidney stone types is possible based on the mean HU Values. Using a random forest classifier, an accuracy of 78% could be demonstrated with regard to the differentiation of the various kidney stone types based on the extracted Radiomics Parameters. Original_NGTDM_Strength was identified in the random forest feature importance analysis as the most important parameter for differentiating between the various types of kidney stone.