DOG Forum digital
Längst ist die Digitalisierung aus dem ärztlich Alltag nicht mehr wegzudenken. Doch der Bedarf nach Informationen zu diesem komplexen wie auch sensiblen Thema ist groß. Mit dem Forum digital bietet der DOG Kongress 2019 eine Plattform um Fragen zu digitalen Anwendungen, Datensicherheit und Big Data zu diskutieren. Neben wissenschaftlichen Sitzungen greifen Vorträge auch Service-Themen auf, die den Teilnehmern praktische Empfehlungen für den Alltag in Klinik und Praxis vermitteln werden.
Donnerstag, 26. 9. 2019
Freitag, 27. 9. 2019
Samstag, 28. 9. 2019
engl
| Forum Digital | 10:15 - 11:30 | 28.09.2019 |
| Symposien | Sa12 |
| Academic Deep Learning Projects from the Technical Perspective | |
Academic Deep learning projects in ophthalmology have shown to be promising approaches to improve patient care. Joint technological research teams - informatics and ophthalmologists - continue to work on a variety of topics. In this symposium, computer scientists will present their latest projects in regard to ophthalmology.
Karsten Kortüm
(München/Ludwigburg)
Hrvoje Bogunovic
(Wien)
Deep learning has a great potential as a precision medicine instrument for providing individualized prognosis of disease progression. In this presentation, we focus on predicting individual conversion to a late stage disease in eyes with early/intermediate age-related macular degeneration (AMD) using longitudinal optical coherence tomography (OCT) imaging. We discuss supervised and unsupervised learning approaches that allow learning representations that go beyond conventional imaging biomarkers.
Deep learning has a great potential as a precision medicine instrument for providing individualized prognosis of disease progression. In this presentation, we focus on predicting individual conversion to a late stage disease in eyes with early/intermediate age-related macular degeneration (AMD) using longitudinal optical coherence tomography (OCT) imaging. We discuss supervised and unsupervised learning approaches that allow learning representations that go beyond conventional imaging biomarkers.
Holger Langner
(Mittweida)
The development of prediction models for VEGF-therapy outcome and future visual acuity in the context of Age-related Macula Degeneration (AMD) is a challenging task, especially when it is applied to real-life data that contains rich variants and cross diagnoses of degenerative eye diseases for patients who have not undergone surgery under thoroughly planned study conditions. The more classical methods for time series prediction cannot be applied here in a straightforward manner. In our talk we propose an approach of combining OCT-image analysis based on Deep-Learning methods with Survival / Hazard modeling of textual features derived from clinical notes, and present some of our current results.
The development of prediction models for VEGF-therapy outcome and future visual acuity in the context of Age-related Macula Degeneration (AMD) is a challenging task, especially when it is applied to real-life data that contains rich variants and cross diagnoses of degenerative eye diseases for patients who have not undergone surgery under thoroughly planned study conditions. The more classical methods for time series prediction cannot be applied here in a straightforward manner. In our talk we propose an approach of combining OCT-image analysis based on Deep-Learning methods with Survival / Hazard modeling of textual features derived from clinical notes, and present some of our current results.
Philipp Prahs
(Regensburg)
The use of machine learning and neural networks in ophthalmic imaging has increased in recent years. A common approach is to employ a seperate neural network classifier for every clinical entity that is to be investigated. We present common solutions when similar clinical questions are analyzed and clinical images for neural network training are scarce.
The use of machine learning and neural networks in ophthalmic imaging has increased in recent years. A common approach is to employ a seperate neural network classifier for every clinical entity that is to be investigated. We present common solutions when similar clinical questions are analyzed and clinical images for neural network training are scarce.
Olivier Morelle
(Bonn)
Automated image analysis has become an indispensable tool in ophthalmology due to the increasing amounts of data. Algorithms are already being used in studies and will soon be incorporated into our clinical routine. This talk will highlight the technical aspects of a deep learning algorithm for segmenting the ellipsoidal zone in OCT.
Automated image analysis has become an indispensable tool in ophthalmology due to the increasing amounts of data. Algorithms are already being used in studies and will soon be incorporated into our clinical routine. This talk will highlight the technical aspects of a deep learning algorithm for segmenting the ellipsoidal zone in OCT.
Christoph Kernstock
(Tübingen)
In Germany, only 10-12% of pedestrian traffic lights are equipped with assistive signals for visually impaired or blind people. Developing a smartphone-app to assist these people poses some additional challenges compared to diagnostic applications for deep learning in ophthalmology.
In Germany, only 10-12% of pedestrian traffic lights are equipped with assistive signals for visually impaired or blind people. Developing a smartphone-app to assist these people poses some additional challenges compared to diagnostic applications for deep learning in ophthalmology.