A novel approach to classify urinary stones using dual-energy kidney, ureter and bladder (DEKUB) X-ray imaging

Detection of urinary stone composition before treatment can help in its management. The purpose of this work is to study the feasibility of classifying the kidney stone compositions in vivo by dual-energy kidney, ureter, and bladder (DEKUB) X-ray imaging.

Six urinary stone compositions with nine diameters were simulated in a water phantom, and two 70- and 120-kVp images were acquired by radiography tally of the Monte Carlo code. Six image features among 10 were selected for classification of the kidney stones. Four classification algorithms were applied to the dataset using MatLab software. Five-fold cross-validation was applied to the most accurate algorithm for 1000 times and the true and false detection rates were reported.

The obtained accuracy of kidney stone classification was 96 ± 2% and this decreased with increasing noise level. The DEKUB was successful in distinguishing brushite, calcium oxalate monohydrate, cystine, and calcium phosphate stones from other types.

Acceptable results achieved by the low-cost, low-dose DEKUB system in detection of kidney stone composition not only obviates a need for complicated imaging systems such as dual-energy computed tomography, but also provides an available and useful aid for physicians to choose between treatment approaches.