The article introduces a novel approach called Progressive Growing of Patch Size, designed to enhance 3D medical image segmentation through automatic curriculum learning. This method involves gradually increasing the patch size during the training process, which helps improve class balance and accelerates training. Evaluations of the approach were conducted in both resource-efficient and performance-focused modes, demonstrating its effectiveness. The technique is positioned as a curriculum learning strategy that systematically adjusts training complexity to optimize outcomes. Claims associated with the method indicate positive impacts on segmentation quality, supported by empirical evaluation. This approach aligns with ongoing research efforts to improve medical image analysis by leveraging adaptive training schemes. Overall, Progressive Growing of Patch Size represents a promising advancement in the application of machine learning to medical imaging tasks.