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.
Dear authors, we are pleased to verify that you meet the reviewer's valuable feedback to improve your research.
Thank you for considering PeerJ Computer Science and submitting your work.
Kind regards
PCoelho
[# PeerJ Staff Note - this decision was reviewed and approved by Jyotismita Chaki, a PeerJ Section Editor covering this Section #]
This manuscript is well-structured and written in clear, professional English. It is supported by adequate background information, relevant literature references, raw data, figures, and tables, making it self-contained and directly aligned with the stated hypotheses. The results are presented rigorously, with explicit definitions and detailed proofs, ensuring clarity, reproducibility, and compliance with academic publishing standards.
This study offers original primary research that clearly defines a relevant question, addresses a known knowledge gap, and aligns well with the journal’s aims and scope. AI investigation is thorough, ethically conducted, and supported by reproducible methods described in enough detail for others to replicate.
The manuscript shows strong validation via internal cross-validation, real-world test datasets, and agreement checks with expert diagnoses, confirming its accuracy and broad applicability. The results strongly support that the methodology is reliable, clinically relevant, and prepared for wider use.
Dear authors,
After the previous revision round, some adjustments still need to be made. As a result, I once more suggest that you thoroughly follow the instructions provided by the reviewers to answer their inquiries clearly.
You are advised to critically respond to all comments point by point when preparing a new version of the manuscript and while preparing for the rebuttal letter. All the updates should be included in the new version of the manuscript.
Kind regards,
PCoelho
The article needs grammatical corrections
Key Points:
Other segmentation/classification models (such as U-Net and EfficientNet) are not included in the baseline comparisons of the paper. Proof of YOLOv8's claimed superiority depends on this.
Simplify the Annotation Process. More details on the quantity of images tagged for every class and the handling of inter-annotator agreement are needed.
Absence of comparison between ENPAT and other models on the same dataset renders the abstract's assertion that it "stands out in terms of accuracy and comprehensive performance" untenable.
Measurements of Performance:
Confidence intervals or statistical significance tests comparing junior dentists' performance with artificial intelligence are absent.
No baseline models—like ResNet or U-Net—are compared in order to show YOLOv8's superiority.
Issues related to overfitting:
Though early stopping is discussed (Line 278), this work does not use a strong validation technique (such as external validation or cross-validation), which raises issues regarding generalisability.
Interpretability is:
Beyond bounding boxes and confidence scores, it's not clear how readily non-expert users can grasp the model's outputs.
Using Grad-CAM, SHAP, or other visual aids is not mentioned.
The study aims to develop a novel YOLOv8-based diagnostic model, named the Esthetic-zonal Non-invasive Periodontal Assessment Tool (ENPAT), to comprehensively evaluate periodontal health conditions in the anterior esthetic zone using intraoral images.
This is a novel approach, as previous studies have primarily focused on binary classification of periodontal diseases. This study introduces a more nuanced multi-classification system for Oral Health Grading (OHG), Modified Gingival Index (MGI), and Papillae Filling Index (PFI).
- The study's comprehensive evaluation of the model's performance using various metrics, including Precision, Recall, Accuracy, F1-Score, mAP50, Specificity, and Negative Predictive Value, is a significant strength.
-
The authors could consider providing more details on the dataset, such as the distribution of different periodontal conditions and the rationale behind the selected sample size.
- It would be beneficial to further compare the ENPAT model's performance with the current clinical gold standard (i.e., periodontal clinical examination) to highlight its advantages.
- The authors may want to consider discussing the potential limitations of using intraoral images, such as the impact of lighting conditions and image quality, and how these factors may influence the model's performance.
- The authors could explore incorporating additional clinical parameters, such as plaque index or bleeding on probing, to enhance the model's diagnostic capabilities.
consider reading these articles for better clinical understanding
Classification and prediction of smoker melanosis in gingiva using SqueezeNet algorithms
Deep learning-based dental plaque detection on primary teeth: a comparison with clinical assessments
ALL APPROPRIATE RAW DATA MUST BE AVAILABLE
NO COMMENTS
The manuscript does not clearly highlight the novelty of the work. Please provide a clear explanation of what differentiates this study from existing research in the field.
1. The manuscript does not provide details about the source of the dataset used in the study. For transparency and reproducibility, please include information on where the dataset was obtained, including relevant citations or links. If the dataset is proprietary or custom-collected, please specify the methodology and context of data collection. Additionally, if a custom-collected dataset is used, kindly include information about the associated copyright policies and permissions to ensure compliance with data usage regulations
2. The manuscript mentions the use of the YOLO model, which requires a labeled dataset. However, it is unclear which labels were annotated in the dataset images (e.g., class labels, bounding box coordinates). Please provide detailed information about the labeling process, including the types of labels used. Additionally, since image labeling is a tedious process, clarify whether the labeling was done manually or if any automation techniques were used to facilitate the process.
3. It is not clear whether the existing YOLO models were used as they are or if any modifications were made to the model architecture or parameters. If modifications were made, please provide a detailed explanation of the changes, including the rationale behind them and how they impact the model's performance.
4. The manuscript mentions the use of the F1 confidence curve to analyze the model's performance. Please provide a detailed explanation of how the F1 confidence curve is constructed and interpreted. Additionally, clarify how this curve helps in concluding that the proposed model performs well compared to existing models. Including insights on the significance of different regions of the curve and how threshold selection impacts the F1 score would enhance the understanding of the performance analysis.
5. Running deep learning models typically requires substantial time for training and inference. The manuscript does not provide details about the computational efficiency of the proposed model. Please include information on the training time, the hardware specifications used (e.g., GPU model, RAM), and the time taken for prediction. Additionally, compare the time complexity and inference speed of the proposed model with existing models to give a clearer perspective on its efficiency and practical applicability.
6. Uploading and making available resources such as the dataset, code, and programs used in the research is crucial for enabling other researchers to build upon and validate the findings. Sharing these resources can significantly contribute to the advancement of the field and help researchers replicate and evolve the work in the coming years. Please consider providing access to these resources, either through a repository or supplementary materials, to enhance the transparency and reproducibility of the research
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.