An Enhanced Diagnosis of Monkeypox Disease Using Deep Learning and a Novel Attention Model Senet on Diversified Dataset

Round 1

Reviewer 1 Report

In mti-2522429, Surati et al. attempted to use deep learning to diagnise monkeypox.

S1. They applied existing Squeeze and Excitation Network (SENet) Attention model [46] to classify monkeypox from two other diseases (chickenpox and measles).

S2. The use of SENet together with InceptionV3 led to high accuracy.

W1. More exhaustive evaluation (esp. comparisons with other related works mentioned in Table 2) seems to be appropriate.

Author Response

Thank you very much for your comments!

Reviewer 2 Report

The work aims to achieve an improved diagnosis and classification of Monkeypox, the results of which can be used by health experts and researchers to deal with the spread of this disease. For this, starting from the image data set, the paper performs the classification for Monkeypox, Chikenpox and Measles with the help of trained autonomous DL models (InceptionV3, EfficientNet, VGG16) and the Squeeze and Excitation Network (SENet) attention model. Thus, the SE-block of an unexplored SENet architecture is combined with VGG16,  EfficientNet  and InceptionNet models of CNN to achieve remarkable improvement in the accuracy of classification. 

The authors began by searching, collecting and aggregating existing and verified datasets. Additionally, they identified two different data sets from distinct sources that were aggregated to improve performance metrics.

Different available datasets collected from the existing authors/resources are shown in table 1 and the details of the data sets used in this work for experimentation are given in table 3, where the third data set used during training is an aggregation of two sets of individual data.

The images of all 3 datasets are scaled to 224x224 before feeding the dataset in the DL architecture of EfficientNet, VGG16, SENet+EfficientNet and SENet+VGG16 and to 299x299 before feeding the dataset in the DL architecture of InceptionV3 and SENet+InceptionV3.

Figures 1, 2, 3 and 4 show the VGG16 architecture, respectively the InceptionV3 architecture, the EfficientNet architecture, and the SENet architecture. Figure 5 shows the entire workflow of the explored DL architectures.

Section 3.2.2 presents the metrics used for comparison and Table 4 presents the results of the comparative analysis in relation to these metrics for the analyzed DL models

In section 3.3. a detailed discussion of the obtained results is made. 

In section 3.3. there is a detailed discussion of the results obtained, and the result of their comparison with existing approaches (Table 7)

The obtained results demonstrate the validity of the proposed solution.

Weakness

- I suggest placing all figures and tables before the conclusions section

Author Response

Thank you very much for your comments! Figures are now placed before the conclusion section.