Federated learning for secure development of AI models for Parkinson's disease detection using speech from different languages

Among automatic PD assessment methods, Recently, deep learning (DL)-based methods have particularly deep learning models have gained particular interest. Recently, gained a lot of attention for analyzing PD speech signals the community has explored cross-pathology and crosslanguage [7, 8]. However, a major impediment to developing such models which can improve diagnostic accuracy even robust DL models is the need for accessing lots of training further. However, strict patient data privacy regulations largely data, which is challenging for many institutions. Thus, benefiting prevent institutions from sharing patient speech data with each from data from different external institutions could solve other. In this paper, we employ federated learning (FL) for PD this issue. However, strict patient data privacy regulations in detection using speech signals from 3 real-world language corpora the medical context make this infeasible in most cases in realworld of German, Spanish, and Czech, each from a separate institution.