Juan Cerrolaza PhD, Abstract Co-Author: Nothing to Disclose

Nabile M. Safdar MD, Abstract Co-Author: Shareholder, Montage Healthcare Solutions, Inc

Emmarie Myers, Abstract Co-Author: Nothing to Disclose

Craig A. Peters MD, Abstract Co-Author: Nothing to Disclose

James Jago, Abstract Co-Author: Employee, Koninklijke Philips NV

Marius George Linguraru DPhil, MS, Presenter: Nothing to Disclose

To evaluate the potential of a new ultrasound (US) based 3D Hydronephrosis Index (HI) to assess the severity of hydronephrosis from the semi-automatic quantification of renal parenchyma (P) and collecting system (CS).

The retrospective dataset (IRB approved) consists of 10 patients (0-7 years) whose severity varies from grade 1 to 3 according to the Society for Fetal Urology hydronephrosis scale (SFU-HS). 3DUS images were acquired from an iU22 system with a X6-1xMatrix transducer (Philips Healthcare). The kidney (K) was segmented using semi-automated active shape models, including new shape and texture models tailored to US physics. The CS was obtained with a new extension of 3D graph cuts that includes brightness and contrast normalization, and anatomical prior information. K and CS were delineated manually by a board certified radiologist to evaluate the accuracy through volume overlap (VO), relative volume difference (RVD) and average surface distance (ASD) (leave-one-out cross-validation). A 2D slice containing a longitudinal section of the K and its CS was manually segmented to compute the 2DHI. Finally, we analyze the Spearman correlation of both 3DHI and 2DHI to the SFU-HS.

For the segmentation of the K and CS, VO was 86% and 62%, RVD was 3% and 26%, and ASD was 3.06mm and 0.35mm, respectively. The error in the estimation of the 3DHI was 1.5 percentage points. For the groups with SFU-HS 1, 2, and 3, the average (and range) 3DHI was 0.83 (0.82-0.83), 0.71 (0.56-0.80) and 0.43 (0.35-0.48), respectively. The values for the 2DHI were 0.80 (0.73-0.83), 0.62 (0.55-0.66) and 0.49 (0.35-0.64) for SFU-HS 1, 2, and 3 respectively. The correlation coefficient between the HI and the SFU-HS was 0.92 (3DHI) and 0.82 (2DHI).

The quantification of 3DUS images shows the potential of 3DHI for the assessment of severity of hydronephrosis with better agreement with the SFU-HS than the classic 2DHI. Further studies will focus on developing rigorous correlations with more detailed functional parameters.

Automated analysis of 3D kidney US data demonstrates higher correlation with SFU-HS than manual 2D measures, demonstrating the potential for 3D US image processing techniques.

Cerrolaza, J, Safdar, N, Myers, E, Peters, C, Jago, J, Linguraru, M, The Value of 3D Hydronephrosis Index in the Assessment of Pediatric Hydronephrosis.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14047425.html