MP7-20 NOVEL AUTOMATED STONE DETECTION SYSTEM TO MEASURE RENAL CALCULI WITH ULTRASOUND

Abstract

You have accessJournal of UrologyImaging/Radiology II1 Apr 2014MP7-20 NOVEL AUTOMATED STONE DETECTION SYSTEM TO MEASURE RENAL CALCULI WITH ULTRASOUND Franklin Lee, Bryan Cunitz, Barbrina Dunmire, Mathew Sorensen, Ryan Hsi, Michael Bailey, and Jonathan Harper Franklin LeeFranklin Lee More articles by this author , Bryan CunitzBryan Cunitz More articles by this author , Barbrina DunmireBarbrina Dunmire More articles by this author , Mathew SorensenMathew Sorensen More articles by this author , Ryan HsiRyan Hsi More articles by this author , Michael BaileyMichael Bailey More articles by this author , and Jonathan HarperJonathan Harper More articles by this author View All Author Informationhttps://doi.org/10.1016/j.juro.2014.02.343AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookTwitterLinked InEmail INTRODUCTION AND OBJECTIVES Ultrasonography is underutilized in the diagnosis and follow-up of nephrolithiasis due to its poor accuracy and variability in detecting stones. We sought to decrease ultrasound user variability by developing a computer algorithm to automate stone sizing. METHODS Calcium oxalate monohydrate renal calculi were imaged in a water bath using a research ultrasound system and C5-2 transducer. Stones were imaged at depths of 6, 10, and 14 cm. A 10 cm tissue phantom was utilized to mimic the attenuating effects of human tissue. The raw ultrasound data was saved and analyzed using three different automation programs including: ray line imaging (RL), flash angle imaging (FA) and harmonic imaging (HI). RL imaging is equivalent to conventional B-mode. FA averages ultrasound signals captured over multiple angles. HI is a ultrasound technique that utilizes higher frequencies to improve ultrasound resolution. Voltage and gain were automatically adjusted within each mode to the minimum value needed to accurately detect the stone. Stone width was measured using an edge detection algorithm based on gray scale intensity. Groups were compared using one-way ANOVA and post-hoc Tukey analysis. RESULTS A total of 45 stones and 180 measurements were obtained. Stones < 4mm were excluded as they were not reliably detected. With the automated program, at least 88% of stones were able to be measured overall. Mean absolute differences between true stone size and automated stone size are shown below (Table 1). For each automation program, overestimation increased with depth. HI was found to be a better predictor of true stone size for depths of 6 and 14 cm (p<0.001) while there was no difference between HI, FA, and RI at 10 cm (p=0.29). Using the tissue block, RL was found to better predict stone size compared to FA and HI (p<0.01). CONCLUSIONS Automated stone detection is a feasible technology that could minimize user variability. Our automated detection algorithm was able to measure stone size for the majority of stone depths. Further refinement of the algorithm is needed to capture smaller stones and to reduce variability. Measured Stone Size (Table 1) Depth Ray Line Imaging Flash Angle Imaging Harmonic Imaging 6 cm 2.1±2.3 mm 2.4±2.9 mm 0.4±3.7 mm 10 cm 2.9±2.5 mm 2.6±3.9 mm 2.2±2.6 mm 14 cm 3.3±2.8 mm 3.6±3.3 mm 1.8±2.7 mm Tissue Block 0.4±4.0 mm 1.5±4.4 mm 1.2±3.7 mm © 2014FiguresReferencesRelatedDetails Volume 191Issue 4SApril 2014Page: e74 Advertisement Copyright & Permissions© 2014MetricsAuthor Information Franklin Lee More articles by this author Bryan Cunitz More articles by this author Barbrina Dunmire More articles by this author Mathew Sorensen More articles by this author Ryan Hsi More articles by this author Michael Bailey More articles by this author Jonathan Harper More articles by this author Expand All Advertisement Advertisement PDF downloadLoading ...