Novel hypospadias phenotype urethral plate description determined by pixel cluster mapping.
Nicolas Fernandez, MD, PhD1, Adam Maxwell, PhD2.
1Seattle Children's Hospital, Seattle, WA, USA, 2University of Washington, Seattle, WA, USA.
Background: Hypospadias affects the formation of the urethra, ventral skin, and corporeal bodies. Location of the urethral meatus has historically been the phenotypic landmark that defines hypospadias. Nonetheless, classifications following location of the urethral meatus fail to consistently predict outcomes and have no correlation with the genotype. There is no universal classification that has intended to explore the phenotype beyond anatomical landmarks. Description of the urethral plate is very subjective and difficult to reproduce for basic science or non-surgical research. We hypothesize that the use of digital pixel cluster analysis proposes a novel method to describe the phenotype of patients with hypospadias and differentiates the tissues affected by the anomaly and involved in the repair.Methods: A total of 1200 anonymous images of patients with hypospadias that included the image of the penis from the ventral aspect were classified by 3 independent experts following the GUMS score. Statistical pixel k-means cluster analysis, a machine-learning method to categorize pixels based on color groups, was performed on images classified following the 4 types of urethral plate descriptions (1. Healthy, deeply grooved urethral plate, 2. adequate urethral plate, grooved, 3. urethral plate narrow with some fibrosis, flat, 4. urethral plate indistinct very narrow or flat). Images were compared against adjacent regions of normal skin in proximity to the urethral plate defect and were labeled as "healthy" tissue. Results: Analysis was used to generate intensity maps of areas corresponding with high intensity for each color group to differentiate tissue types. The k-means cluster analysis identified differences for healthy vs. abnormal tissues surrounding the urethral plate for two color clusters with p<0.0001 for both groups (Figure 1). Image maps showing proximity of each pixel to different k color groups. The red outline in k1 indicates abnormal tissue and yellow is a region of healthy tissue. No statistical differences were identified between classified urethral plate GUMS groups 1-4.Conclusions: Subjective description of the urethral plate does not evaluate objectively differences in the quality of the tissue, which may have an impact on surgical outcomes. Statistical k-means cluster analysis can detect significant differences between "healthy and abnormal tissue surrounding the urethral plate. This approach, combined with automated feature detection, may be useful to objectively quantify urethral plate condition to aid in decision-making for hypospadias interventions. 
Back to 2021 Posters