Identification of Hypospadias Parameters Using Digital Photography and Artificial Intelligence

Overview

The goal of this observational study is to develop and validate a digital pattern recognition system based on artificial neural network to determine various parameters in hypospadias. The main question it aims to answer is: How accurate is the digital pattern recognition system based on artificial neural network to determine various parameters in hypospadias? Participants in this study are hypospadias patients aged \< 18 years old. The guardian (and the patient, if applicable) will be informed about the study and asked for consent. The digital picture of participants' penis will be taken from different angles according to the predetermined angle. The clinical characteristics of the photographed penis are then inputted and used to train a customized artificial neural network (ANN). The machine is then used to predict various hypospadias parameters presenting at the patients' penis. The accuracy of the machine is then compared to the measurement done by pediatric urologists.

This is a development protocol for a digital pattern recognition program for identification of various hypospadias parameters using artificial neural network. The population for this study are children suspected of having hypospadias who are admitted or referred to hospitals in Indonesia. The consent for the capture of the images needed in the study will be obtained from the parents of hypospadias patients as part of the standard of care to be used as a clinical reference. The photograph taken for this study were kept in an encrypted database. The inclusion criteria of this study are children aged \<18 years old who are suspected of having hypospadias, with age-matched control. Those with history of hypospadias repair or those who refuse to participate in the study are excluded. The clinical outcomes measured in this study are: 1. Hypospadias status: hypospadias or non-hypospadias 2. Meatal location: glanular, coronal, distal shaft, proximal shaft, penoscrotal 3. Meatal shape: normal, abnormal 4. Quality of the urethral plate: good, bad 5. Glans diameter: in mm 6. Glans shape: normal, abnormal The clinical outcomes were measured preoperatively and 1 month postoperatively A hypospadias and normal penis image with matched age database was used to train the artificial neural network. Three image aspects of the penis (ventral, dorsal, and lateral aspect which include the glans, shaft, and scrotum) were taken from each subjects. The data was labeled with hypospadias parameters: hypospadias status, meatal location, meatal shape, quality of the urethral plate, glans diameter, and glans shape. The data were uploaded to train the artificial neural network. The statistical analysis plan will be followed for all clinical outcomes analyses. Standard error, 95% confidence intervals, and P values will be reported whenever possible. Intra and inter rater analysis will be performed using the Fleiss Kappa statistical analysis. The data is deemed to be statistically significant if the p-value is less than 0.05. In addition, accuracy, precision, recall, and f1 score values will be computed to measure the performance of recognition model.