A new explainable machine learning framework combining EEG signals with clinical data distinguishes Alzheimer's disease from frontotemporal dementia with 89.8 percent accuracy.