Non-enhanced single-energy computed tomography of
urinary stones
av
Johan Jendeberg
Akademisk avhandling
Avhandling för medicine doktorsexamen i medicinsk vetenskap med inriktning mot kirurgi,
som kommer att försvaras offentligt fredagen den 5 februari 2021 kl. 09.00,
Hörsal C3, X-huset, Campus Universitetssjukhuset Örebro Opponent: Professor Anders Magnusson
Uppsala Universitet
Örebro universitet
Institutionen för Medicinska Vetenskaper 701 82 ÖREBRO
Abstract
Johan Jendeberg (2021): Non-enhanced single-energy computed tomography of urinary stones. Örebro Studies in Medicine 229.
Computed tomography (CT) is the mainstay imaging method for urinary stones.
The aim of this thesis was to optimize the information obtained from the initial CT scan to allow a well-founded diagnosis and prognosis, and to guide the clinician as early and as far as possible in the further treatment of urinary stone disease.
We examined CT scan parameters with regards to their importance for prediction of spontaneous ureteral stone passage, the impact of inter-reader variability of stone size estimates on this prediction, and the pre-dictive accuracy of a semi-automated, three-dimensional (3D) segmenta-tion algorithm. We also developed and tested the ability of a machine learning algorithm to classify pelvic calcifications into ureteral stones and phleboliths.
Using single-energy CT, three quantitative methods for classification of stone composition into uric acid and non-uric acid stones in vivo were prospectively validated, using dual-energy CT as reference.
Our results show that spontaneous ureteral stone passage can be pre-dicted with high accuracy, with knowledge of stone size and position. The interreader variability in the size estimation has a large impact on the pre-dicted outcome, but can be eliminated through a 3D segmentation algo-rithm. Which size estimate we use is of minor importance, but it is im-portant that we use the chosen estimate consistently. A machine learning algorithm can differentiate distal ureteral stones from phleboliths, but more than local features are needed to reach optimal discrimination.
A single-energy CT method can distinguish uric acid from non-uric acid stones in vivo with accuracy comparable to dual-energy CT.
In conclusion, single-energy CT not only detects a urinary stone, but can also provide us with a prediction regarding spontaneous stone passage and a classification of stone type into uric acid and non-uric acid.
Keywords: Diagnostic, CT, urinary stone, kidney stone, urolithiasis,
phlebolith, uric acid, spontaneous passage, CNN, artificial intelligence. Johan Jendeberg, Department of Radiology, Faculty of Medicine and Health, School of Medical Sciences
Örebro University, SE-701 82 Örebro, Sweden, johan.jendeberg@regionorebrolan.se
