Autor(es): Bajouco, Miguel ; Mota, David ; Coroa, Manuel ; Caldeira, Salomé ; Santos, Vítor ; Madeira, Nuno
Data: 2017
Identificador Persistente: https://hdl.handle.net/10316/47464
Origem: Estudo Geral - Universidade de Coimbra
Assunto(s): Schizophrenia; Machine-Learning; Biomarkers; Neuroimaging
Schizophrenia is a severe mental disorder and one of the leading causes of disease burden worldwide. It represents a source of signi cant su ering and disability to the a ected individuals, and is associated with substantial societal and economical costs. The diagnosis of schizophrenia still depends exclusively on the detection of symptoms that are also present in other mental disorders. This situation causes overlapping of the boundaries of the diagnostic categories and constitutes a source of diagnostic errors. Moreover, current treatment algorithms do not take into account the substantial interindi- vidual variability in response to antipsychotic drugs. As a result, around one-third of patients are treatment-resistant to rst line antipsychotic drugs. This deleterious consequence is associated with poor individual outcomes and elevated healthcare costs. Neuroimaging research in schizophrenia has shed some light in a vast array of structural and functional connectivity abnormalities and neurochemical (dopamine and glutamate) imbalances, which may constitute ‘organic surrogates’ of this disorder. However, the neuroimaging eld, so far, has not been able to identify biomarkers that could facilitate early detection and allow individualised treatment management. This paper reviews neuroimaging studies from di erent modalities that may provide relevant biomarkers for schizo- phrenia. We discuss how the current application of novel Machine Learning methods to the analyses of imaging data is allowing the translation of such ndings into potential biomarkers enabling the prediction of clinical outcomes at the individual level, towards the development of innovative and personalised treatment strategies.
