COVID-19 and Artificial Intelligence: An Approach to Forecast the Severity of Diagnosis
, Alice Elena Ghenea 2,*,†, Ștefan Udriștoiu 1,†, Manuela Neaga 1, Ovidiu Mircea Zlatian 2, Corina Maria Vasile 3, Mihaela Popescu 4, Eugen Nicolae Țieranu 5, Alex-Ioan Salan 6, Adina Andreea Turcu 7, Dragos Nicolosu 8, Daniela Calina 9,* and Ramona Cioboata 10Round 1
Reviewer 1 Report
The presented paper deals with an artificial intelligence-based method for grading the Covid-19 severity based on multimodal features taken at admission. The proposed technique in this work provided fast and powerful assistance to physicians in forecasting the severity of the Covid-19 diagnosis evolution. - The paper deals with the topic of actuality. - The paper is clear, well written, and the organization is very good. - The references are up to date, and they are well organized according to the format required by the journal. Nevertheless, I have some recommendations to improve the quality of the manuscript further: The conclusion is very short and should be enlarged by recapping the work's main contribution and adding some perspectives. - The results (or findings) should be compared to some recently published papers (works) under the same configurations (i.e., the same database and protocol of evaluation). - Why the authors divided the dataset to 380/95 for train/test sets? In general, we use 2/3 of the population for training and the remaining 1/3 for testing. - Couples of grammatical errors and misleading sentences have been found. It would be good if authors perform carefully another revision, I can cite: • Line 86: remove 'as'. • Line 174: remove 'the'. • Line 296: correct '…is learning…' --->'…is learned…' • …. - I recommend authors cite the following paper as it contains very interesting information about the principle of deep learning and its different architectures: I. Adjabi et al. 'Past, Present, and Future of Face Recognition: A review'. Electronics, 2021.Author Response
"Please see the attachment."
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The work done is of current interest, but the manuscript is not publishable as it stands, as it still has significant gaps and therefore needs improvement.
The main weakness is the choice of the imaging modality which is the X-ray scan.
Why was the CT scan not used?
Especially since the images shown in Figure 6 are blurred and appear noisy. Such images would have required pre-processing in order to have a targeted COVID-19 diagnosis. Various denoising techniques exist and we recommend wavelet-based denoising. It would therefore be necessary to talk about it and to cite the following reference at least in the introduction:
A review of wavelet denoising in medical imaging . DOI: 10.1109/WoSSPA.2013.6602330
On the technical side, other weaknesses are visible:
- The number of patients used during the training and test phases are low (see Table 1) to be able to generalize the adopted approach.
- Figures 2, 3 and 4 do not make sense statistically speaking
To determine the Covid-19 severity, a more technically reliable approach would have been to estimate the volume of contaminated lung, for example by performing a semantic segmentation (see reference below). In any case, the authors should open the discussion and mention that there are other solutions by citing the reference below:
Deep learning for real-time semantic segmentation: Application in ultrasound imaging. Pattern Recognit. Lett. 2021, 144, 27–34. https://0-doi-org.brum.beds.ac.uk/10.1016/j.patrec.2021.01.010
- The proposed algorithms should have been represented by Deep Learning architecture schemes.
- La figure 11 pose un problème expérimental.
- To enhance their manuscript, authors should cite more recent references on COVID-19, such as:
A Review on Deep Learning Techniques for the Diagnosis of Novel Coronavirus (COVID-19). DOI: 10.1109/ACCESS.2021.3058537.
COVID-19 diagnosis -A review of current methods. DOI: 10.1016/j.bios.2020.112752
There are still some typos in the manuscript like in figure 1 where instead of "pacient" discharge , write "patient" discharge.
Author Response
"Please see the attachment."
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The authors followed our recommendations, which significantly improved the manuscript.
