Dementia Classification and Recognition Based on Neuropathological, Hematological, and Genetic Data

Despite numerous advances in modern medical research, clinical diagnosis and correct classification of dementia types are still very challenging during a patient’s life time, as a decent diagnosis of dementia can only be done by performing neuropathological brain examinations after the decease of the patient. Therefore, a majority of diseased patients is not correctly diagnosed in an early state or in the worst case at no time. We have developed an in-vivo classification system for dementia that combines data sources and relates dementia types to disease-related processes in the brain. In detail, the classification model is based on post-mortem data, namely microscopy images of brain slices of patients (currently used for the diagnosis and classification of dementia), hematological data from patients (blood samples), and genetic data of patients (SNPs). We use post-mortem data as training data for supervised machine learning algorithms and so identify relationships between these features and dementia classifications (which are known post-mortem). The so generated mathematical models will be applied on new data from living patients in order to assign a dementia type and state by only using data available at the patient’s lifetime. In our study we analyzed data of more than 200 patients suffering from Alzheimer’s disease, Parkinson's disease, or Amyotrophic lateral sclerosis, and more than 100 control cases. Using this in-vivo classification system novel correlations between blood parameters, neuropathological features and the state of the disease are detected, and variable interaction networks between the different data collections are identified.