OBJECTIVES: The size estimation in CT images of an obstructing ureteral calculus is important for the clinical management of a patient presenting with renal colic. The objective of the present study was to develop a reader independent urinary calculus segmentation algorithm using well-known digital image processing steps and to validate the method against size estimations by several readers. METHODS: Fifty clinical CT examinations demonstrating urinary calculi were included. Each calculus was measured independently by 11 readers. The mean value of their size estimations was used as validation data for each calculus. The segmentation algorithm consisted of interpolated zoom, binary thresholding and morphological operations. Ten examinations were used for algorithm optimisation and 40 for validation. Based on the optimisation results three segmentation method candidates were identified. RESULTS: Between the primary segmentation algorithm using cubic spline interpolation and the mean estimation by 11 readers, the bias was 0.0 mm, the standard deviation of the difference 0.26 mm and the Bland-Altman limits of agreement 0.0 ± 0.5 mm. CONCLUSIONS: The validation showed good agreement between the suggested algorithm and the mean estimation by a large number of readers. The limit of agreement was narrower than the inter-reader limit of agreement previously reported for the same data. KEY POINTS: The size of kidney stones is usually estimated manually by the radiologist. An algorithm for computer-aided size estimation is introduced. The variability between readers can be reduced. A reduced variability can give better information for treatment decisions.
OBJECTIVES: The size estimation in CT images of an obstructing ureteral calculus is important for the clinical management of a patient presenting with renal colic. The objective of the present study was to develop a reader independent urinary calculus segmentation algorithm using well-known digital image processing steps and to validate the method against size estimations by several readers. METHODS: Fifty clinical CT examinations demonstrating urinary calculi were included. Each calculus was measured independently by 11 readers. The mean value of their size estimations was used as validation data for each calculus. The segmentation algorithm consisted of interpolated zoom, binary thresholding and morphological operations. Ten examinations were used for algorithm optimisation and 40 for validation. Based on the optimisation results three segmentation method candidates were identified. RESULTS: Between the primary segmentation algorithm using cubic spline interpolation and the mean estimation by 11 readers, the bias was 0.0 mm, the standard deviation of the difference 0.26 mm and the Bland-Altman limits of agreement 0.0 ± 0.5 mm. CONCLUSIONS: The validation showed good agreement between the suggested algorithm and the mean estimation by a large number of readers. The limit of agreement was narrower than the inter-reader limit of agreement previously reported for the same data. KEY POINTS: The size of kidney stones is usually estimated manually by the radiologist. An algorithm for computer-aided size estimation is introduced. The variability between readers can be reduced. A reduced variability can give better information for treatment decisions.
Authors: Brian H Eisner; Avinash Kambadakone; Manoj Monga; James K Anderson; Andrew A Thoreson; Hang Lee; Stephen P Dretler; Dushyant V Sahani Journal: J Urol Date: 2009-02-23 Impact factor: 7.450
Authors: Jianfei Liu; Shijun Wang; Evrim B Turkbey; Marius George Linguraru; Jianhua Yao; Ronald M Summers Journal: Med Phys Date: 2015-01 Impact factor: 4.071