Gradient-Guided Convolutional Neural Network for MRI Image Super-Resolution
Round 1
Reviewer 1 Report
Summary evaluation:
This paper describes the use of convolutional neural networks for super-resolution MRI images and propose residual block network in combination with gradient-guided sub network.
Experiments are done on three MRI public database.
The results show that gradient-guided CNN approaches may work better than other methods.
While the topic and the proposed approach are interesting, the presentation has to be improved.
Specific comments:
(Major comments)
The presentation of "introdution" and "related works" of the paper should be improved, it is necessary to explain traditional approaches versus CNN approaches in the field of super-resolution MRI images, and discuss the salient features of the state of the art;the training details are very poor: no hiperparameters tuning are provide; why ADAM and not SGDM? or why size "64" and not tuning on other size (... 32, 128, ...); a graph should be produced of the performances / epochs.
(Minor revision)
It is opportune to improve the bibliography, especially the part related to CNN and their use in biomedical imaging (Recommended: "A State-of-the-Art Survey on Deep Learning Theory and Architectures" Electronics, 2019; "Deep CNN for IIF Images Classification in Autoimmune Diagnostics" Applied Sciences, 2019; "Deep Convolutional Neural Network for HEp-2 Fluorescence Intensity Classification" Applied Sciences, 2019); in 2.1 section, the first sentence should be improved; line 100 duplicate word: "for for".Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
The authors describe a technology of a deep gradient-guided residual network for restoring a high-resolution image from its input low-resolution version.
The topic is very important and interesting to IT technology, AI developers, and also to Radiology community, as well as Radiology industry - it feels like the further, future goal could be shortening of scanning times (benefit to the patients) and simplifying (cost reducing direction) of MRI-scanners as well - and that is of extreme importance and significance.
The text is easy to understand, the style nearly perfect; the pictures and formulas are informative and their ammount does not exceed reasonable number.
The major drawback of the paper in my opinion is lack of presentation of it's novelty. Lines 48-57 describe the novelty of the method, but it is very important to compare the results of the study with other authors's results and bring the novelty points of the current study forward.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The article has been revised as indicated, excellent work....