Automated quantification of penile curvature using artificial intelligence

Area of Science:

• Medical imaging
• Artificial intelligence
• Urology

Background:

• Penile curvature (PC) assessment traditionally suffers from subjective variability among surgeons.
• Accurate PC measurement is crucial for diagnosing conditions like Peyronie's disease and guiding treatment.
• Existing methods lack standardized, objective quantification.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI)-based algorithm for automated penile curvature (PC) measurement.
• To improve the precision and objectivity of PC assessments using 2D images.
• To provide a tool for surgeons and researchers in hypospadiology and related fields.

Main Methods:

• A dataset of 900 images from 3D-printed penile models with varying curvatures was created.
• A framework involving penile area localization (Yolov5), shaft segmentation (Unet), and curvature estimation (Hough Transform) was developed.
• Smartphone cameras were used for image acquisition under diverse conditions.

Main Results:

• The AI framework achieved high accuracy in penile area localization (mAP 99.4%) and shaft segmentation (DSC 98.4%).
• The curvature angle estimation demonstrated excellent performance with a mean absolute error (MAE) of 8.5°.
• The system proved robust across various imaging conditions.

Conclusions:

• The developed AI algorithm offers a significant improvement in the precision of penile curvature measurements.
• This automated approach can reduce inter- and intra-surgeon variability in PC assessments.
• The framework has the potential to enhance clinical diagnosis and research in urology and hypospadiology.