Since 2017, IsletNet has been developed and tested during actual transplantations. The produced results are consistent and the final reports are easy to process and save. The whole application environment is user-friendly with a short learning curve.
Data collected from multiple experts shows that the islet volume predictions made by IsletNet are highly correlated — the coefficient of determination (R^2) is more than 0.99. Furthermore, we compare the IsletNet predictions with expert predictions, pixel by pixel. An aggregated F1-score greater than 0.9 shows that the system has a human-like precision.