Intelligent, cooperative support for diagnosis and therapy in ophthalmology
Today’s imaging technologies in ophthalmology are so advanced that retinal and vascular structures in the eye can be resolved with unprecedented precision in two, three and even four dimensions. However, interpreting the image material and deriving a therapy decision taking into account the patient's history is a complex task that requires a lot of experience. Treatment errors may have severe consequences for patients. The recently launched joint project »Ophthalmo-AI«, coordinated by the Fraunhofer Institute for Biomedical Engineering IBMT, aims to create an intelligent and interactive assistance system that supports ophthalmologists with methods of explainable artificial intelligence to provide comprehensible diagnoses and treatment suggestions.
The results of the »Ophthalmo-AI« project will be an intelligent assistance system which supports ophthalmologists by using image data and clinical data in order to arrive at a correct medical diagnosis and the best possible treatment. For understandable suggestions to the medical staff, the AI system will first mark biological structures and pathological features in the image data. Then, special AI models will derive diagnoses based on the image findings as well as additional information from the patient record. Finally, the system will suggest the optimal treatment and predict the chances of success of the suggested therapy. Using interactive machine learning methods, doctors' knowledge about the respective case will also be integrated into the process. Development will be based on treatment data collected at large scale and processed on a special data integration platform. The utility of the resulting augmented intelligence system will be demonstrated in a clinical setting for macular degeneration and diabetic retinopathy. All data-driven processing will fully comply with GDPR data protection requirements.
The three-year joint project »Ophthalmo-AI«, which commenced in March 2021, is coordinated by the Fraunhofer Institute for Biomedical Engineering IBMT and brought to implementation by an internationally experienced, multidisciplinary team of experts from the fields of medicine, medical technology, text analytics, medical informatics and artificial intelligence. In the project, the Fraunhofer IBMT provides the data integration platform for sharing and processing relevant clinical data for the AI model development and develops the data layer of the assistance system. As a specialist for text analytics, LangTec from Hamburg shall develop tools for extracting diagnostically relevant information from patient records, which will be needed as additional input in training and executing the machine learning models. LangTec is also responsible for implementing the pseudonymisation component for personal data. The German Research Centre for Artificial Intelligence GmbH (DFKI), represented by the research units Interactive Machine Learning (IML) and Cognitive Assistance Systems (COS), is developing the required AI machine learning tools for diagnosis and therapy support. With the help of interactive machine learning, ophthalmologists are given an interface to explanatory tools and visualisations that permit them to integrate their expert knowledge into the process (human-in-the-loop). In addition, Heidelberg Engineering GmbH, as a supplier of medical technology, shall develop methods for feature extraction and segmentation in so-called OCT retina images. The resulting AI assistance system is expected to score significantly higher with doctors and patients in acceptance and trustworthiness. These aspects shall be investigated by Saarland University. The contributing ophthalmic providers, the Sulzbach Eye Clinic and the Eye Centre at the St. Franziskus Hospital in Münster, shall provide access to extensive clinical data for training the machine learning models, defining application scenarios and testing the demonstrators created in clinical use.
»Ophthalmo-AI« is supported by the Federal Ministry of Education and Research as part of the funding focus "Adaptive Technologies for Society - Intelligent Interaction of Humans and Artificial Intelligence".