Automatische Bildanalyse und künstliche Intelligenz in der Nephropathologie

BACKGROUND: The digital transformation of pathology through the widespread use of so-called whole slide scanners offers numerous opportunities for nephropathology, especially with respect to the implementation of computer assistance. Currently, the possibilities of systems based on the use of deep learning, a special technique of information processing, are being intensively explored. OBJECTIVE: The aim is to determine the current state of research regarding applications of deep learning methods for image analysis in nephropathology. MATERIAL AND METHODS: A literature search was carried out in Web of Science (WOS) and PubMed. For figure one the following search query in WOS was used: ALL= (digital pathology AND AI OR deep learning OR machine learning). RESULTS: There are numerous applications of deep learning-based methods to assist in nephropathology. These focus largely on segmentation and quantification of renal histology, although diagnostic classification and synthetic data generation are also increasingly being explored. The translation of these systems into everyday diagnostic practice has not yet taken place. For example, prospective evidence demonstrating the utility of these methods in clinical care is lacking. CONCLUSION: The implementation of digital nephropathology with assistance from deep learning-based methods has great potential. The translational gap should be filled by multicenter prospective interdisciplinary studies in the future.