Ontology population using CNN model: Application to COVID-19 diagnosis

AI models and especially deep learning models have found applications across various medical domains. While many studies focus on using electronic health records (EHRs) data to train AI models, implementing the reverse process poses several challenges. Also, the heterogeneous nature of the used data in EHRs, such as diagnoses, clinical reports, test results, prescriptions, and medical images, makes their interpretation difficult for the practitioners which require a unified understanding. Medical ontologies are useful to improve data interpretability and semantics. In this paper, we aim to create a synergy between ontologies and deep learning techniques to enhance efficiency, data understanding, and provide healthcare quality. We propose a system in which our developed ontology is populated automatically using the predictions of an AI model. We have developed and trained a CNN model to aid in the COVID-19 diagnosis. The initial obtained results are promising with an accuracy of 98%.