Neuroimaging research revealed the mechanisms involved in neuropsychiatric disorders during the last decade. The DSM-5 suggested incorporating biomarkers in diagnosing neuropsychiatric diseases. Although large amounts of imaging data have been accumulated to date, a decisive step has still to be done: Translating group studies’ results into individualized regimens. Early prediction of diagnosis and therapy are in particular relevant for neuropsychiatric disorders. Diseases shall be diagnosed as early as possible to predict their course and enable disease-specific treatment. The symposium will focus on predicting diagnosis and treatment in neuropsychiatric disorders with cutting-edge pattern recognition algorithms in neuroimaging data. The first part of the symposium will focus onto neurodegenerative diseases. Matthias Schroeter will show how pattern classification in multimodal imaging data can be used to predict the second most frequent dementia syndrome, frontotemporal lobar degeneration and its subtypes. Franziska Albrecht will discuss how machine learning in imaging data can separate atypical Parkinsonian syndromes and identify their symptoms. Juergen Dukart will present a novel approach of linking spatial alteration patterns in resting state fMRI to underlying neurotransmitter systems as an individualized biomarker of Parkinson’s disease. The second part of the symposium will extend the view to psychiatric disorders such as schizophrenia and mood disorders. Here, Nikolaos Koutsouleris will show what neuroimaging data can contribute to identification and prediction of treatment response in depression and schizophrenia. The symposium discusses the potential of pattern recognition algorithms / machine learning on the way to individualized treatment regimens in the framework of personalized medicine, and the importance of biomarkers for disease classification in new systems such as ICD-11.
Predicting frontotemporal lobar degeneration with pattern recognition algorithms based on multimodal imaging and meta-analyses
Matthias Schroeter, Leipzig (Germany)
Details anzeigen
Autor:in:
Matthias Schroeter, Leipzig (Germany)
Neuroimaging research revealed mechanisms involved in neuropsychiatric disorders. New diagnostic criteria including imaging biomarkers were proposed for frontotemporal lobar degeneration (FTLD), in particular for its behavioral variant and language subtypes, i.e. primary progressive aphasias. We validated imaging biomarkers by conducting comprehensive meta-analyses across studies from the literature and investigating the potential to translate group studies’ results into individualized regimens. Anatomical likelihood estimate meta-analyses identified the neural correlates for each FTLD subtype underlining disease-specificity (Europ J Neurol 2016;23:704-12; Cortex 2014;57:22-37; JNNP 2015;86:700-1). Analyses were conducted separately for atrophy (MRI) and glucose metabolism (FDG-PET) revealing specific patterns. Results open the road to method-specific imaging criteria as already suggested for Alzheimer’s disease. If new imaging criteria are valid they shall enable early diagnosis in single patients. To prove the potential for individual diagnosis we applied cutting-edge pattern recognition algorithms in neuroimaging data. Support vector machine classification (SVM) with multimodal imaging (MRI & FDG-PET) enabled early individual detection and discrimination of FTLD subtypes (Neuroimage Clin 2017;14:334-43 & 14:656-62). Limiting SVM classification regionally to meta-analytically identified disease networks improved discrimination accuracy. Analyses were reliable in multi-centric data. Finally, we investigated multi-syndrome classifiers containing up to eight different groups and the potential of machine learning to predict treatment efficacy, here shown for Parkinson’s disease (Neuroimage Clin 21:101636). Results support and refine the application of imaging criteria and suggest that pattern classification algorithms enable early individual diagnosis and differential diagnosis of FTLD subtypes, a precondition for translation to personalized medicine.