Abstract
There are a number of clinical questions for which there are no easy answers, even for welltrained doctors. The diagnostic tool commonly used to assess cognitive impairment in neurodegenerative diseases is based on established clinical criteria. However, the differential diagnosis between disorders can be difficult, especially in early phases or atypical variants. This takes on particular importance when it is still possible to use an appropriate treatment.
To solve this problem, physicians need to have access to an arsenal of diagnostic tests, such as neurofunctional imaging, that allow higher specificity in clinical assessment. However, the reliability of diagnostic tests may vary from one to the next, so the diagnostic validity of a given investigation must be estimated by comparing the results obtained from “true” criteria to the “gold standard” or reference test. While pathological analysis is considered to be the gold standard in a wide spectrum of diseases, it cannot be applied to neurological processes. Other approaches could provide solutions, including clinical patient follow-up, creation of a data bank or use of computer-aided diagnostic algorithms.
In this article, we discuss the development of different methodological procedures related to analysis of diagnostic validity and present an example from our own experience based on the use of I-123-ioflupane-SPECT in the study of patients with movement disorders. The aim of this chapter is to approach the problem of diagnosis from the point of view of the clinician, taking into account specific aspects of neurodegenerative disease.
Keywords: Alzheimer’s disease, computer-aided diagnosis systems, dementia, diagnostic accuracy, machine learning, neuro-imaging, support vector machines.