Session

Overview: Artificial intelligence (AI) has permeated all aspects of our daily life. Nowadays, AI-driven products are being used to inform important decisions. A particular example in radiotherapy is auto-segmentation based on deep learning (DL), a sub-field of AI, which is actively used to plan for hundreds of patients. Depending on the way AI models have been trained, they can reproduce biases present in the training data, impacting their application. When AI is applied in the healthcare setting, it is essential to recognize and address this algorithmic bias from the development phase to the application. Beyond model training, fair AI should also consider the impact on other aspects: implementation replacing 'human' jobs, and costs attached to these new technologies that can be out of reach for low- and middle-income countries. In this symposium, we will explore the concept of AI fairness and its possible impact in radiotherapy, examining it from the training data to its global-scale impact.