Bladder Cancer Detection Using Convolutional Neural Networks

Recruiting

• Conditions: Bladder Cancer
• Interventions: Diagnostic Test: Al_bladder

• Registration Number: NCT05193656
• Lead Sponsor: Zealand University Hospital

Brief Summary

The investigators aim to experiment and implement various deep learning architectures to achieve human-level accuracy in Computer-aided diagnosis (CAD) systems. In particular, the investigators are interested in detecting bladder tumors from CT urography scans and cystoscopies of the bladder in this project.

Detailed Description

The investigators aim to experiment and implement various deep learning architectures to achieve human-level accuracy in Computer-aided diagnosis (CAD) systems. In particular, the investigators are interested in detecting bladder tumors from CT urography scans and cystoscopies of the bladder in this project. The investigators want to classify bladder tumors as cancer, non cancer, high grade and low grade, invasive and non-invasive, with high sensitivity and low false positive rate using various convolutional neural networks (CNN). This task can be considered as the first step in building CAD systems for bladder cancer diagnosis. Moreover, by automating this task, the investigator scan significantly reduce the time for the radiologists to create large-scale labeled datasets of CT-urography scans and reduce the false-negative and positive that can happen due to human evaluation cystoscopies.

Study Timeline

• Study Start: 2021-06-01
• Study First Submitted: 2021-11-17
• Study First Submitted QC: 2022-01-14
• Study First Posted: 2022-01-18
• Status Verified: 2024-01
• Last Update Submitted: 2024-01-28
• Last Update Posted: 2024-01-30
• Primary Completion: 2026-06-01
• Study Completion: 2026-06-01

Recruitment & Eligibility

• Status: RECRUITING
• Sex: All
• Target Recruitment: 5000

Inclusion Criteria

• Patients with first time hematuria
• Patients with the control program for previous bladder cancer

Exclusion Criteria

• Patients with control cystoscope for noncancer suspected disease

Study & Design

• Study Type: OBSERVATIONAL
• Study Design: Not specified

Arm && Interventions

Group	Intervention	Description
Detecting bladder tumor	Al_bladder	Patients with hematuria, or previous bladder tumor

Primary Outcome Measures

Name	Time	Method
Comparing standard technique to Machine Learning	5 years	The accuracy of Machine learning to detect bladder cancer compared to standard cystoscopy

Secondary Outcome Measures

Name	Time	Method
Detecting accuracy of subtypes of bladder cancer	5 years	The abelity of Machine Learning to identify high grad bladder cancer from low grad bladder cancer

Trial Locations

Locations (1)

Zealand University Hospital; 🇩🇰 Roskilde, Denmark