Abstract

The development of AI models for the diagnosis of diseases in Ophthalmology is accelerating rapidly. These models, however, mostly refer to a single or only a few diseases. In addition, only a few examination techniques are considered in a given model. For example, many current AI models only consider fundus images for their diagnosis. Thus, general clinical information that is also available about a patient is often neglected. This information includes, for example, age, sex and secondary diagnoses.In most cases, diseases have different frequencies depending on age and sex. This leads to a different a priori probability for a disease diagnosis depending on the patient's age. When considering several diseases in one model, this class imbalance must be considered in the model. Large longitudinal data sets of disease histories of individual persons are rare or under high data privacy requirements.With the aim of more holistic AI models that include multiple diagnoses, a novel AI model was developed that is able to simulate longitudinal disease histories of individual (synthetic) patients. This was made possible based on data from the German Federal Statistical Office and a newly developed AI algorithm. In this present study, the model is used to simulate different diagnoses of the German population that lead to inpatient hospital admissions. The data were generated as completely synthetic data and have no data privacy requirements. Information on epidemiology and disease incidence can be derived from this model. Furthermore, it is possible to generate forecasts about future hospital admissions and their underlying diagnoses by extrapolating the model into the future.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call