Jeffry S. Kriegshauser MD, Presenter: Nothing to Disclose

Alvin C. Silva MD, Abstract Co-Author: Nothing to Disclose

Robert Gene Paden, Abstract Co-Author: Nothing to Disclose

Miao He, Abstract Co-Author: Nothing to Disclose

Eric Wisenbaugh MD, Abstract Co-Author: Nothing to Disclose

Mitchell Humphreys MD, Abstract Co-Author: Nothing to Disclose

Steven Ilan Zell MD, Abstract Co-Author: Nothing to Disclose

Evaluate predictability of renal stone composition using values for 53 variables obtained with ssDECT and by applying selected data analysis algorithms. 

Thirty-two ex vivo stones found to be >90% pure by IR spectroscopy (IRS) were evaluated with ssDECT using up to 53 variables, including size, effective Z, density (in HU) at 11 monochromatic keV values, and 40 material density pairs. Data was evaluated using several algorithms, including ANN, Random Tree, and NB Tree models. A subset of 23 stones, which excluded stones less than 5 mm, also was evaluated using up to 26 variables. Seventeen stones measuring 5 mm or larger removed from 13 patients were evaluated in vivo with ssDECT using up to 11 variables, focusing on commonly available variables: density (70 keV), effective Z, and iodine and water pairs. IRS determined true composition.  

In the 32 stone dataset, 14 were uric acid (UA) and 18 non-UA stones. Non-UA stones were 7 cystine (CYS), 7 struvite (STR) and 4 calcium oxalate (CaOx) stones. In the 23 stone dataset, 12 were uric acid (UA) and 11 non-UA stones. Non-UA stones were 5 CYS, 4 STR and 2 CaOx stones. In vivo stones included 2 UA, 2 CYS, 12 calcium-based (90-100% CaOx Monohydrate (COM) or mixtures of COM with CaOx Dehydrate and calcium phosphonates), and 1 N4-Acetyl-Sulfomethoxazol. Several algorithms could predict UA versus non-UA with 100% accuracy in all sets and water-iodine material density pairs were also 100% accurate. Non-UA stones were determined less accurately, with the best models 64% accurate for all 32 ex vivo stones, improving to 82% for stones >5mm (23 stone dataset). Both errors in the latter set were misclassified STR stones. In vivo, 1 small (5 mm) calcium-based ureteral stone was misclassified as CYS and the N4-Acetyl-Sulfomethoxazol stone was classified as a calcium-based stone. 

Using ssDECT, UA stones can be consistently distinguished from non-UA stones, most simply using the iodine-water material density pairs. Considerable overlap in parameters is seen with non-UA stones, although CYS and calcium-based stones are more accurately predicted than STR stones.

Accurate in vivo prediction of renal stone composition is important for determining cause and best management, and can aid surgical planning.

Kriegshauser, J, Silva, A, Paden, R, He, M, Wisenbaugh, E, Humphreys, M, Zell, S, Single-source Dual Energy CT (ssDECT) Renal Stone Characterization: A Multi-Parametric Approach    .  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14016298.html