Challenges with Human Digital Twins - Harald Rietdijk

Описание: Within the lectoraat Digital transformation one of the research fields is the implementation of Human Digital Twins in personal coaching applications that can be used for example in rehabilitation trajectories or coaching apps for people working in sedentary jobs. When trying to implement a Human Digital Twin, we face several challenges that relate specifically to the fact that we are dealing with human data sets. One of those being that, as common in many clinical studies, we are working with a small datasets with a relatively high number of features. In a recent study, we investigated the possibility of identifying features that are important in determining the duration of rehabilitation after knee arthroplasty among people of working age, expressed in the Return To Work period, by using feature selection tooling. Several models were used to classify the patient’s data into two classes, and the results were evaluated based on the accuracy and the quality of the ordering of the features, for which we introduced a ranking score. A selection of estimators was used in an optimization step, reorganizing the feature ranking. The results showed that for some models, the proposed optimization results in a better ordering of the features. The ordering of the features was evaluated visually and identified by the ranking score. Furthermore, for all models, higher accuracy, with a maximum of 91%, was achieved by applying the optimization process. The features that were identified as relevant for the duration of the Return To Work period are discussed and provide input for further research.