Fusion of multimodal temporal clinical data for the retrieval of similar patient cases

The temporal evolution of the patient is a key factor in providing an effective healthcare. Moreover, comparing a patient case with other cases having a similar progress in time can prove valuable for medical decision making. However, the large amounts of complex temporal clinical data are not always considered by clinicians in the clinical information process. In this work, a computational framework is proposed for the comparison of multimodal temporal clinical data obtained from different patients. The purpose of the proposed framework is to use a patient's temporal evolution so as to retrieve the most similar profiles from large repositories which can be used to compare treatments, diagnosis, test results and other information. This is achieved by retrieving similar temporal patterns from the data by being based on a sequence similarity scheme. The similarity between the patient cases is assessed by a novel fusion scheme that involves the estimation of multiple dynamic time warping distances between the temporal clinical sequences. The results obtained from its application on a reference dataset of hepatic infections demonstrated high precision even for low recall rates.