Neuritic plaques and neurofibrillary tangles are crucial morphological criteria for the definite diagnosis of Alzheimer's disease. We evaluated 12 unstained frontal cortex and hippocampus samples from 3 brain donors with Alzheimer's disease and 1 control with hyperspectral Raman microscopy on samples of 30 × 30 μm. Data matrices of 64 × 64 pixels were used to quantify different tissue components including proteins, lipids, water and beta-sheets for imaging at 0.47 μm spatial resolution. Hierarchical cluster analysis was performed to visualize regions with high Raman spectral similarities. The Raman images of proteins, lipids, water and beta-sheets matched with classical brain morphology. Protein content was 2.0 times, the beta-sheet content 5.6 times and Raman broad-band autofluorescence was 2.4 times higher inside the plaques and tangles than in the surrounding tissue. The lipid content was practically equal inside and outside. Broad-band autofluorescence showed some correlation with protein content and a better correlation with beta-sheet content. Hyperspectral Raman imaging combined with hierarchical cluster analysis allows for the identification of neuritic plaques and neurofibrillary tangles in unstained, label-free slices of human Alzheimer's disease brain tissue. It permits simultaneous quantification and distinction of several tissue components such as proteins, lipids, water and beta-sheets.