[Artificial intelligence in diagnostics-a pathology perspective]

The increasing complexity and individualization of oncologic diagnostics and therapy present new challenges for pathology. At the same time, artificial intelligence (AI) is evolving from a futuristic concept into a core area of digital medicine. With the availability of digital whole-slide images (WSIs) and increasingly powerful deep learning architectures, the number of publications in digital pathology has risen almost exponentially since around 2019.At the algorithmic level, numerous innovations have emerged in recent years: convolutional neural networks (CNNs), which initially dominated the field, are increasingly being replaced by Vision Transformer (ViT)-based models. Since 2023, foundation models have gained rapid importance due to their broad applicability and generalizability.Proof-of-concept studies have repeatedly demonstrated that AI-based solutions can improve the efficiency and sensitivity of diagnostic workflows. Several AI algorithms for histopathology have already been approved by U.S. and European regulatory agencies. More recent developments, such as vision-language models (VLMs), enable the multimodal integration of text and image data, opening up new interactive possibilities in diagnostics.Overall, the field is at the transition from proof-of-concept studies toward clinical implementation. In particular, foundation models have the potential to fundamentally reshape the structure of histopathological diagnostics in the near future. However, technical, legal, and socio-psychological barriers must still be overcome before widespread clinical adoption can be achieved.