Summary: The authors in this work extracted a dataset from EEG records aiming to predict prognosis of a premature newborn. The authors made use of the concept of Artificial Neural Network (ANN) to validate their results. In this work, the authors suggest to divide EEG records according to more than two intervals based on predefined amplitude thresholds. The used methodology on the amplitude of bursts intervals of EEG signal yields 26 inputs parameters instead of 14 as published by the authors in previous work. Using a specific analyzer, named by the authors as SegEEG, a single output parameter predicts the normal, risky and abnormal cases of premature newborn.

Comments Format and structure Authors should review the formatting of the references. Moreover, the references should be numbered in view o their appearance in the text. For instance, after citing reference [1], the authors cite references [5, 6, 7, 19] while references [2] appear later on in the second paragraph. There is a problem in the paragraph organization in the last paragraph of section 2. There is problem in using the font in the first paragraph of the fifth section. Using correct formatting while referring to figures in text. Clarity and language The authors tend to give names for parameters, terms and processes without illustration of their choice. This is not a major problem, however it decreases the quality of the paper and leaves the reader a bit uncomfortable. For instance, the choice of the term BDD-IBI as well as the use of faible, moy and fort and other terms. Sometimes, the authors change terms in referring to the same condition or observation. Again, this is not wrong but it makes the reader feel that there is little coherency. For instance, authors sometimes use the term healthy other times use the term normal to refer to normal cases. The same issue exists for the terms interval and phase. Figures 3 and 6 are not clear. Replacing them with tables is much more appreciated. Figures 4 and 5 are of low quality and all the other figures are of moderate quality. It is vital that the authors enhance the quality of the figures. More clarifications can be provided on how the recordings are treated in order to obtain the 26 parameters. More clarifications using block diagrams should be provided for the ANN system used in the validation section. It is important as well to explain the methodology used in choosing and adjusting the weights and the threshold. Typing mistake of the term BDD_SEG in table 3. The language (structure and grammar) through out the whole paper needs to be reviewed. In section 2, the authors recalled the result of their work in [11]. Recheck the percentages given because different values appear in the abstract of the published paper: “Finally, they found that decision tree gave best result with performance of 100% for sick records, 76.9% for risky and 69.1% for normal ones.” If it is the case, the new results don’t differ a lot from the already published results.

Commenting and interpretation The results are well commented but there is a lack of interpretation of certain things. The authors presented in section 3 the amplitude thresholds according to which the signal was segmented into three intervals (low, medium and high). However, the authors didn’t precise how these values were chosen. This should be tackled in the modified version and parametric study could be presented in order to convince the readers more.