Using neural networks for differential diagnosis of Alzheimer disease and vascular dementia

摘要

Differential diagnosis among different types of dementia, mainly between Alzheimer (AD) and Vascular Dementia (VD), offers great difficulties due to the overlapping among the symptoms, and signs presented by patients suffering these illnesses. A differential diagnosis of AD and VD can be obtained with a 100% of confidence through the analysis of brain tissue (i.e. a cerebral biopsy). This gold test involves an invasive technique, and thus it is rarely applied. Besides these difficulties, to get an efficient differential diagnosis of AD and VD is essential, because the therapeutic treatment needed by a patient differs depending on the illness he suffers. In this paper, we explore the use of artificial neural networks technology to build an automaton to assist neurologists during the differential diagnosis of AD and VD. First, different networks are analyzed in order to identify minimum sets of clinical tests, from those normally applied, that still allows a differential diagnosis of AD and VD; and, second, an artificial neural network is developed, using backpropagation and data based on these minimum sets, to assist physicians during the differential diagnosis of AD and VD. Our results allow us to suggest that, by using our neural network, neurologists may improve their efficiency in getting a correct differential diagnosis of AD and VD and, additionally, that some tests contribute little to the diagnosis, and that under some combinations they make it rather more difficult.