CNN-based diagnosis model of children's bladder compliance using a single intravesical pressure signal

Area of Science:

• Urology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Bladder compliance assessment is vital for diagnosing bladder functional disorders.
• Urodynamic study (UDS) is the primary method but is complex and time-consuming, particularly for pediatric patients.
• An efficient pre-UDS screening method for bladder compliance is needed.

Purpose of the Study:

• To develop and validate a 1D-CNN model for accurate and efficient bladder compliance screening.
• To create a faster, more accessible method for identifying potential bladder functional disorders before UDS.

Main Methods:

• A dataset was constructed using urodynamic study (UDS) data.
• A 1D-convolutional neural network (1D-CNN) model was designed, optimized, and trained.
• The trained model was applied to data from a novel perfusion experiment.

Main Results:

• The 1D-CNN model demonstrated superior performance compared to other evaluated algorithms.
• The model accurately and efficiently identified abnormal bladder compliance.
• The perfusion experiment data, when analyzed by the model, yielded promising results.

Conclusions:

• The developed 1D-CNN model shows significant potential as an accurate and efficient screening tool for bladder compliance.
• This approach could streamline the diagnostic process for bladder functional disorders, reducing reliance on time-consuming UDS in initial evaluations.
• Further validation in clinical settings is warranted to confirm its utility in pediatric urology.