Table 8 Comparison between the proposed model and the other pretrained models based on various aspects.
From: A hybrid filtering and deep learning approach for early Alzheimer’s disease identification
Aspect | VGG19 | ResNet152V2 | InceptionV3 | EfficientNetB0 | Proposed Model |
|---|---|---|---|---|---|
Parameter | More parameters | More parameters | More parameters | Less parameters | Fewer parameters |
(\(\tilde{1}\)43.7M) | (\(\tilde{6}\)0.4M) | (\(\tilde{2}\)3.9M) | (\(\tilde{5}\).3M) | (\(\tilde{1}\)4.5M) | |
Model | Larger, generating | Larger model size, | Larger, leading to | Smaller model than | Smaller, facilitating |
Size | slower inference | slower inference | slower inference | EfficientNetV2B3 | faster inference |
Training | Longer due to | Moderate training | Longer due to | Shorter due to | Typically shorter for |
Time | increased number | time | more parameters | fewer parameters | fewer parameters |
of parameters | |||||
Feature | Effective but less | Effective at | Effective but not | Efficient, but less | More effective at |
Extraction | efficient than | extracting | as efficient | so compared | obtaining relevant |
EfficientNetV2B3 | relevant features | as EfficientNetV2B3 | to EfficientNetV2B3 | features from MRIscans | |
Heatmap | Generate less precise | Detailed heatmaps | Detailed heatmaps | Generates precise | Generates more |
Precision | heatmaps than | but less refined | but less refined | heatmaps but less | refined and |
EfficientNetV2B3 | detailed than V2B3 | precise heatmaps | |||
Interpret- | Good; but, heatmaps | High interpretability | High, but heatmaps | High, but less refined | Higher due to refined |
ability | can have less detail | with | can be less detailed | than V2B3 | heatmaps, making it easier |
detailed heatmaps | to interpret brain regions | ||||
Scalability | Less scalable compared | Good scalability | Good scalability | Good scalability | Better scalability with |
to EfficientNetV2B3 | compound scaling |
