Radiomics and Image Segmentation of Urinary Stones by Artificial Intelligence (RISUS_AI)

Kidney stone disease causes significant morbidity, and stones obstructing the ureter can have serious consequences. Imaging diagnostics with computed tomography (CT) are crucial for diagnosis, treatment selection, and follow-up. Segmentation of CT images can provide objective data on stone burden and signs of obstruction. Artificial intelligence (AI) can automate such segmentation but can also be used for the diagnosis of stone disease and obstruction.

In this project, the aim is to investigate if:

Manual segmentation of CT scans can provide more accurate information about kidney stone disease compared to conventional interpretation.

AI segmentation yields valid results compared to manual segmentation. AI can detect ureteral stones and obstruction or predict spontaneous passage.

Study Overview

• Status: Active, not recruiting
• Conditions: Renal Colic; Urolithiasis; Urinary Stone; Kidney Stone; Ureter Stone; Urosepsis; Obstruction Ureter

Detailed Description

Background:

Goals and Objectives:

The project aims to contribute to personalized and improved treatment and follow-up of patients with kidney stones using radiomics and the development of an artificial intelligence tool for CT examination assessment. The objectives are to assess:

• Whether manual segmentation of CT images of the urinary tract provides equivalent or more accurate information about kidney stone disease compared to conventional interpretation and reporting.
• Whether segmentation performed with AI yields valid results compared to manual segmentation.
• Whether AI can detect ureteral stones and obstruction and/or predict spontaneous passage of stones.

Method:

Cohort:

Patients are recruited to the study at Oslo University Hospital, Radiology Department, Section Aker, which performs approximately 1350 CT examinations for urinary tract stones in approximately 1000 patients each year. Approximately 500 patients with a new episode or newly occurring colic pain and clinical suspicion of kidney stones are expected to be included.

Clinical data (where available):

• Baseline CT: date and image data
• Initial treatment (conservative, URS, PCN, ESWL) decision after baseline CT
• Follow-up CT: date and image data
• Time to spontaneous stone passage (negative control CT) or completed surgical intervention (URS)
• Any other surgical/invasive procedure
• Stone chemical analysis
• Clinical biochemistry: creatinine/eGFR, CRP, leukocytes (at baseline and follow-ups).

Image data:

Clinical radiology report:

• Stone: (largest calculus and any obstructing calculus): largest diameter in any plane, density (ROI set by clinical judgment, largest possible ROI - in the slice where the stone is largest), location (upper ureter: above crossing of vessels, lower ureter: below crossing of vessels, ostial: in bladder wall)
• Renal pelvis: largest diameter of calyx neck lower calyx, clinical assessment of dilation (not dilated/slight/moderate/severe).
• Segmentation:
• Stone: total segmented stone volume, largest diameter, and density of segmented stone.
• Collecting system: total segmented volume of the collecting system and renal pelvis.

• Study Type: Observational
• Enrollment (Actual): 522

Contacts and Locations

Study Locations

• Norway
  • Oslo, Norway, 0586
    • Oslo University Hospital, Aker

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Non-Probability Sample

Study Population

Patients referred for CT for new/acute episode of renal colic and suspicion of /or known urinary stone disease.

Description

Inclusion Criteria:

• Referral for CT due to episode of renal colic and clinical suspicion of urinary stone disease or
• Referral for CT due to new episode of pain in patient with known urinary stone disease
• Age ≥ 18 years

Exclusion Criteria:

• Referral for control CT of asymptomatic patients with known urinary stone disease
• Referral for control CT after treatment
• Referral for control CT for spontaneous passage of stone.
• Lack of informed consent for any reason.

Study Plan

How is the study designed?

Number of groups / cohorts

1

Cohorts and Interventions

Group / Cohort
Adults investigated with CT for suspected urinary stone disease; Newly occurring colic pain and clinical suspicion of kidney stones or known kidney stone with new/increasing symptoms. Age ≥ 18 years

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Comparison of stone diameter from manual segmentation with radiology report	Stone diameter (in mm) compared between manual segmentation and radiology report (paired t-test or wilcoxon rank sum test if non-normally distributed data)	At time of CT examination (inclusion and follow up - expected average 12 weeks)
Comparison of AI-segmentation of stones (DICE-score) with manual segmentation	DICE-score for AI-segmentation of stones, compared to manual segmenation (gold standard)	At time of CT examination (inclusion and follow up - expected average 12 weeks)
Prospective performance (diagnostic accuracy) of AI detection of ureteral stone (compared to radiology report (gold standard)	Comparison of differences in dicotomous proportions in paired data according to Newcombe	At time of CT examination (inclusion and follow up - expected average 12 weeks)

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Comparison of stone density from manual segmentation with radiology report	Stone density (in Hounsfield Units) compared between manual segmentation and radiology report (paired t-test or wilcoxon rank sum test if non-normally distributed data)	At time of CT examination (inclusion and follow up - expected average 12 weeks)
Comparison of distention of renal pelvis from manual segmentation with radiology report	Distention of renal pelvis (in mm) compared between manual segmentation and radiology report (paired t-test or wilcoxon rank sum test if non-normally distributed data)	At time of CT examination (inclusion and follow up - expected average 12 weeks)
Comparison of AI-segmentation of stones (Hausdorff distance) with manual segmentation	Haussdorff distance for AI-segmentation of stones, compared to manual segmenation (gold standard)	At time of CT examination (inclusion and follow up - expected average 12 weeks)
Comparison of AI-segmentation of stones (diagnostic accuracy) with manual segmentation	Diagnostic accuracy for AI-segmentation of stones compared to manual segmenation (gold standard)	At time of CT examination (inclusion and follow up - expected average 12 weeks)
Comparison of AI-segmentation of renal pelvis (Dice-score) with manual segmentation	DICE-score for AI-segmentation of renal pelvis compared to manual segmenation (gold standard)	At time of CT examination (inclusion and follow up - expected average 12 weeks)
Comparison of AI-segmentation of renal pelvis (Hausdorff distance) with manual segmentation	Hausdorff distance for AI-segmentation of renal pelvis compared to manual segmenation (gold standard)	At time of CT examination (inclusion and follow up - expected average 12 weeks)
Comparison of AI-segmentation of renal pelvis (diagnostic accuracy) with manual segmentation	Diagnostic accuracy for AI-segmentation of renal pelvis compared to manual segmenation (gold standard)	At time of CT examination (inclusion and follow up - expected average 12 weeks)
Comparison of AI-segmentation of renal parenchyma (DICE-score) with manual segmentation	DICE-score for AI-segmentation of renal parenchyma compared to manual segmenation (gold standard)	At time of CT examination (inclusion and follow up - expected average 12 weeks)
Comparison of AI-segmentation of renal parenchyma (Hausdorff distance) with manual segmentation	Hausdorff distance for AI-segmentation of renal parenchyma compared to manual segmenation (gold standard)	At time of CT examination (inclusion and follow up - expected average 12 weeks)
Comparison of AI-segmentation of renal parenchyma (diagnostic accuracy) with manual segmentation	Diagnostic accuracy for AI-segmentation of renal parenchyma compared to manual segmenation (gold standard)	At time of CT examination (inclusion and follow up - expected average 12 weeks)
Prospective performance (diagnostic accuracy) of AI detection of ureteral obstruction (compared to radiology report (gold standard)	Comparison of differences in dicotomous proportions in paired data according to Newcombe	At time of CT examination (inclusion and follow up - expected average 12 weeks)

Collaborators and Investigators

• Sponsor: Oslo University Hospital
• Investigators: Principal Investigator: Peter M. Lauritzen, MD, PhD, Oslo University Hospital

Study record dates

Study Major Dates

• Study Start (Actual): May 21, 2024
• Primary Completion (Estimated): August 2, 2025
• Study Completion (Estimated): March 28, 2028

Study Registration Dates

• First Submitted: April 30, 2024
• First Submitted That Met QC Criteria: May 8, 2024
• First Posted (Actual): May 14, 2024

Study Record Updates

• Last Update Posted (Actual): August 11, 2025
• Last Update Submitted That Met QC Criteria: August 5, 2025
• Last Verified: August 1, 2025

More Information

Terms related to this study

• Keywords: Computed tomography; Urolithiasis; Artificial intelligence; Ureteral obstruction; Image segmentation
• Additional Relevant MeSH Terms: Urogenital Diseases; Pain; Neurologic Manifestations; Male Urogenital Diseases; Pathological Conditions, Anatomical; Kidney Diseases; Urologic Diseases; Female Urogenital Diseases; Female Urogenital Diseases and Pregnancy Complications; Ureteral Diseases; Nephrolithiasis; Kidney Calculi; Calculi; Urinary Calculi; Urolithiasis; Renal Colic; Ureteral Obstruction
• Other Study ID Numbers: 31347039; 660399 (Other Identifier: Regional Committees for Medical Research Ethics (in Norway))

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No