Drying Patterns of Cerebrospinal Fluid as Indicator for Alzheimer’s Disease by a Machine Learning Framework

Cerebrospinal fluid contains specific biomarkers of Alzheimer’s disease that include amyloid beta peptides and tau proteins. In this work, we present for the first time possible evidence that the formation of the constituents of cerebrospinal fluid during drying is related with Alzheimer’s. We use machine learning to examine optical microscope images of dried cerebrospinal fluid patterns from patients with Alzheimer’s and healthy controls to create a diagnostic model. To analyze the images, the histogram of oriented gradients is used as a feature descriptor. Each image is mapped into the corresponding feature space, and principal component analysis is applied for dimensionality reduction. A machine-learning prediction model with a sensitivity of 82% was built. These promising preliminary results show great potential for new rapid and low-cost diagnostic pathways in the detection of Alzheimer’s disease.