All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
Thank you for your valuable contribution.
[# PeerJ Staff Note - this decision was reviewed and approved by Xiangjie Kong, a PeerJ Section Editor covering this Section #]
no comment
no comment
no comment
The authors followed the suggestions received during the first phase of the revision process, improving the quality and the clarity of the manuscript.
Please address all the requests and suggestions of the reviewers in detail.
**PeerJ Staff Note:** Please ensure that all review and editorial comments are addressed in a response letter and that any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.
**Language Note:** The review process has identified that the English language must be improved. PeerJ can provide language editing services - please contact us at [email protected] for pricing (be sure to provide your manuscript number and title). Alternatively, you should make your own arrangements to improve the language quality and provide details in your response letter. – PeerJ Staff
The paper presents a review of deep learning models that utilize handwriting samples for detecting learning disabilities. While the topic is timely and potentially impactful, I have several concerns regarding the quality and organization of the manuscript. While there are a few published literature reviews on the application of deep learning techniques for detecting learning disabilities such as dyslexia and dysgraphia, it is not clear what sets this review apart. Also, the writing quality and organization need significant improvement. Clearer structure and language refinement are required to enhance readability.
While Table 2 presents several performance metrics, it lacks clarity on key aspects such as the type of handwriting samples (e.g., digital or scanned), data availability (public, private, or available upon request), and source. Additionally, the paper does not clearly articulate the insights derived from this comparison. It would be valuable to identify which model(s) are most suitable for analyzing children's handwriting in the context of learning disabilities, and to discuss the specific challenges deep learning models face in this task.
Figure 4 lacks visual clarity. It outlines a range of features, such as stroke-based, temporal, texture, edge, structural, and geometric, that are relevant to handwriting analysis. However, it is unclear whether these features are universally applicable across all types of learning disabilities or if certain features are more indicative of specific conditions. This distinction is not clearly addressed in either the figure or the accompanying explanation in Section 3.1. A more detailed discussion on the relevance and specificity of these features to different learning disabilities would strengthen the analysis.
Lines 221–225 mention that deep learning approaches such as CNNs can extract features like pressure and trajectory from handwriting images. However, the manuscript does not clearly explain how these features are derived from image data. Clarification or elaboration on this process would strengthen the reader’s understanding.
The manuscript would benefit from improved logical flow and coherence across all sections. Transitions between ideas are often abrupt, and the progression of the narrative does not consistently guide the reader toward the key arguments or objectives.
This systematic review examines 24 studies on deep learning models applied to handwriting images to detect dyslexia, dysgraphia, and related learning disabilities. It identifies CNNs as the dominant architecture, with emerging interest in Vision Transformers (ViTs). The paper discusses datasets, features used, model performance, and ethical concerns. Major challenges include dataset diversity, computational constraints, and lack of model interpretability. It outlines future directions for inclusive, explainable, and efficient DL solutions in educational and clinical settings.
- Improve English Language, Style, several sections contain grammatical and syntactical errors. A professional language edit is strongly recommended.
- Reword awkward phrases
- Not all readers may be familiar with terms like “Vision Transformer”, “Grad-CAM”, or “Procrustes analysis”; a brief explanation or reference would improve accessibility.
- through Literature Coverage.
- It has novelty:
- It specifically focuses on deep learning models applied to handwriting images for detecting multiple learning disabilities, rather than only one condition.
- It combines both technical analysis and ethical/social dimensions
- It emphasises the potential of multimodal integration and Explainable AI
- The article reflects the scope of the journal
- Improve clarity in the Abstract—split long sentences and remove repetition of terms
- References are numerous and current, but some lack a DOI or a journal context
- Well-organized sections, Critical Reflection, and Forward-Looking
- There is a good discussion on ethical and social concerns
The manuscript is written in clear, precise English, using technical terms that suit its academic audience. In the introduction and literature review, the authors lay out a thorough background on learning disabilities, such as dyslexia and dysgraphia, and discuss how deep learning models are increasingly being used in this area. The methods section details how studies were chosen for review, and the main findings are presented both in narrative form and in tables, making the results easy to follow.
However, to further strengthen the work, the authors should make sure that all raw data and any supplementary files are not only accessible but also clearly explained within the main text, including any necessary metadata to support reproducibility in future research. While the paper does touch on ethical considerations and the challenges of implementing these models in real-world settings, a more thorough discussion with specific recommendations would be beneficial. This would help readers better understand the complexities involved and how they might be addressed moving forward.
This article presents a literature review examining how deep learning models are being used to detect and classify learning disabilities such as dyslexia and dysgraphia by analyzing handwriting samples. The subject matter fits well within the scope and objectives of PeerJ Computer Science, and the choice of a review article is appropriate for the journal’s format.
The paper is well-structured, moving logically from the abstract and introduction through to the background, methodology, main findings, and final conclusions. Visual aids like figures and tables are effectively used to help summarize and clarify the results for readers.
One area that could be improved is the explanation of how the studies were selected for review. Providing more detail about the criteria and process used to choose the included research would enhance the transparency and reliability of the review.
While the review does a good job explaining why further research is needed, it could go a step further by pointing out which aspects of current studies would benefit most from being replicated by other researchers. Additionally, the authors could strengthen their recommendations by offering more practical guidance, such as outlining best practices for collecting data in an ethical manner or suggesting ways to make deep learning models easier to interpret and understand. This would help future researchers address the gaps identified in the review and move the field forward in a meaningful way.
The manuscript is generally well-written, but careful proofreading could further improve clarity and flow. While the process of selecting papers is mentioned, providing more detail about the selection criteria and search strategy would enhance transparency. Important points about ethical challenges, such as data privacy and algorithmic bias, are highlighted. However, this section can be enhanced with more concrete examples or recommendations to provide practical guidance for future researchers. For example, outlining best practices for ethical data collection, or offering strategies to improve model interpretability and generalizability, would make the conclusions more impactful.
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.