Artificial Intelligence in Functional Urology: How It may Shape the Future

Imad Bentellis; Sonia Guérin; Zine-Eddine Khene; Rose Khavari; Benoit Peyronnet


Curr Opin Urol. 2021;31(4):385-390. 

In This Article

Limitations and Future Developments

AI tools are increasingly being incorporated into routine practice. It is true for training, simulation, diagnosis, or therapeutics. However, this is not a one-way change. Indeed, we do not simply add those models to our practice. They also change the way we learn, diagnose and treat. The main limitation to robustness of these algorithms is the quantity and quality of the data on which the models are developed. This is especially the case for rare diseases, with patterns that are difficult to predict or based on subjective elements. Thus, in order to obtain more powerful tools, the first change we are being asked to make is the automation of data collection and data quality. We should work on the collaboration of different medical software, electronic medical records, care platforms, ambulatory devices, patient parameters, questionnaires, and patients-related outcomes measurement. Lastly, we need to find ethical solutions for the storage of operative and perioperative data. The implementation of standardized processes over large territorial groups and cooperation between institutions is one of the keys to this collection of quality data which will ultimately benefit to the patients.

These changes also concern a new way of diagnosing. The predictive tool does not replace the physician, but provides statistics for decision-making and, above all, patients' information. The physician thus acts as an informed mediator within this information ecosystem. But it also questions the pathophysiology explanations. Most AI models are 'black boxes' and hidden variables are very often not clinically relevant.

Another change is underway in training and education. The simulation and monitoring of skills enable the creation of effective training programs and enables to export and share knowledge more easily. This may favor a more dynamic and personalized learning process for the surgeon. It is also true about the way we do science, and the concerns about the peer-reviewing model. It is not excluded that in a near future, peer-review could be AI driven,[38] at least partly.