Using CNN Saliency Maps and EEG Modulation Spectra for Improved and More Interpretable Machine Learning-Based Alzheimer’s Disease Diagnosis

<div>Optimal regions and clusters found for each of the five tasks using experiment 1 settings. In each subplot, the y-axis corresponds to conventional frequency (unit: Hz) and the x-axis to modulation frequency (unit: Hz). (a) N vs. AD1 vs. AD2, (b) N vs. AD, (c) N vs. AD1, (d) AD1 vs. AD2, and (e) N vs. AD2.</div>

Computational Intelligence and Neuroscience

fig6

Figure 6

Figure 6: Using CNN Saliency Maps and EEG Modulation Spectra for Improved and More Interpretable Machine Learning-Based Alzheimer’s Disease Diagnosis 

Figure 6 | Using CNN Saliency Maps and EEG Modulation Spectra for Improved and More Interpretable Machine Learning-Based Alzheimer’s Disease Diagnosis 