Review Reports - Three-Dimensional Analysis of Facial Skeleton Textures in CBCT as an Early Warning Sign of Osteoporosis—A Pilot Study

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

Thank you for submitting your manuscript. The comments related to the study are described in the attached Word file.

Comments for author File: Comments.pdf

Author Response

We sincerely thank the Reviewer for the positive and thoughtful evaluation of our manuscript, as well as for the constructive suggestions. We appreciate the recognition of the originality and clinical potential of the proposed opportunistic CBCT-based approach.

Due to the pilot and exploratory nature of the study, no formal correction for multiple comparisons was applied, as the aim of the analysis was to identify potentially relevant radiomic features for further validation rather than to confirm definitive associations. Nevertheless, the increased risk of Type I error should be acknowledged, and the findings should be considered preliminary and requiring confirmation in independent datasets.

We agree that the moderate correlations observed between tube current (mA) and several top-ranked radiomic features may indicate a potential acquisition-related effect. Unfortunately, detailed information on scanner models was not available for all examinations in this retrospective dataset. We have therefore revised the manuscript to explicitly acknowledge this as a limitation. We also expanded the Discussion to note that variability in acquisition parameters may have influenced some radiomic features and that the observed associations should be interpreted with caution. Given the pilot exploratory nature of the study, these findings will be considered preliminary and require confirmation in future studies with better-controlled acquisition settings and dedicated harmonization strategies.

We agree that assessment of inter-observer variability would further strengthen the reliability of the segmentation procedure. In the present study, ROI delineation was performed twice by the same observer at different time points, which allowed us to assess intra-observer reproducibility and resulted in a high ICC value (0.95). However, inter-observer variability was not evaluated and has now been explicitly acknowledged as a limitation of the study. Future work will include multiple observers to assess segmentation robustness more comprehensively.

Table 3 has been revised to improve readability in grayscale print, and the legend has been clarified accordingly.

The Abstract has been revised to provide a more balanced summary of the results by indicating that the top-performing features included not only GLCM-derived parameters, but also ARM- and Gradient-based features

Reviewer 2 Report

Comments and Suggestions for Authors

This study is an interesting and current pilot study aiming to opportunistically assess bone structure using routine CBCT images. The topic is original and has clinical potential. The approach of obtaining information about bone health from imaging data without requiring additional processing is particularly valuable.

However, the study could be significantly strengthened if certain points were clarified in its current form. Firstly, since the study is pilot-based, it would be appropriate to present the results with a bit more caution. The sample size is limited, and numerous texture parameters were examined; therefore, it should be emphasized that the findings are more exploratory than confirmatory.

Another important point is the number of figures. I believe there are too many figures in the main text, which somewhat distracts from the main message. In particular, the large number of graphs showing similar correlations could be included in the Supplementary Data; leaving only the most selected figures showing the strongest and most significant findings in the main text would improve readability.

In the discussion section, it would be beneficial to better define the study's place within the opportunistic bone assessment literature. In this context, it would be appropriate to add two recent articles (2025 and 2026) focusing on CT-based opportunistic bone mineral density assessment in adult and pediatric liver transplant patients to your references. Such opportunistic approaches are particularly important in liver transplant recipients or other patient populations from many perspectives.

Overall, the study is valuable, but I believe it could be improved with simplification of the figures, clarification of the message, and strengthening of the literature framework.

Author Response

We clarified more explicitly that, given the pilot design, limited sample size, and the large number of analysed texture features, the findings should be interpreted as exploratory and hypothesis-generating rather than confirmatory.
To improve readability, we reduced the number of figures in the main text by moving selected correlation plots with overlapping information to the Supplementary Material, while retaining the most representative figures in the main manuscript.
The Discussion was expanded to better position the study within the broader literature on opportunistic bone assessment and was updated with more recent relevant references.
We also further emphasized the potential clinical role of the proposed method as a supplementary opportunistic screening strategy based on imaging already performed for other diagnostic purposes.
We believe these changes have improved the clarity, focus, and overall clinical context of the manuscript.

Reviewer 3 Report

Comments and Suggestions for Authors

General Comments:

This pilot study investigates the potential of 3D texture analysis of maxillofacial bones using CBCT as an early warning sign for osteoporosis. The integration of AI-generated synthetic phantoms with clinical data is an innovative approach. While the results are promising, several methodological and statistical points need to be addressed to strengthen the validity of the findings.

Major Revisions

Anatomical and Clinical Justification of ROI Selection

The study extracts 309 ROIs from seven different anatomical sites. However, there is a lack of discussion regarding why certain sites were chosen or how the results varied between them. For instance, the authors mention that the ramus is narrow and highly variable. Please provide a more detailed analysis of the correlation strength for each specific site and justify which anatomical region is most reliable and clinically practical for osteoporosis screening.

Texture Feature Selection and Multiple Testing Correction

The reduction of 399 features to 36 significant ones is a key part of the analysis. However, in a radiomics study of this scale, the risk of false positives is high. Please clarify whether any multiple testing corrections (e.g., Bonferroni or Benjamini-Hochberg) were applied. Furthermore, please provide a brief biological interpretation of these 36 features—specifically, what aspects of bone micro-architecture (e.g., spatial heterogeneity) they represent.

Integration of Synthetic Phantoms and Clinical Data

The use of AI-generated phantoms is a strength of this paper. To enhance this, the authors should provide a more direct quantitative comparison between the "trends" observed in the phantoms and those in the clinical patient data (e.g., comparing the slopes of regression lines). This would more robustly demonstrate that the selected texture features are truly sensitive to structural changes in actual human bone.

Minor Revisions

Detailed Patient Metadata: Please clarify the exclusion criteria regarding medications that affect bone metabolism (e.g., bisphosphonates) or other systemic comorbidities, as these can significantly influence the results.
Standardization of Imaging Protocols: Confirm whether the CBCT device settings (kVp, mA, FOV, and reconstruction algorithms) were identical for all patients. If not, describe how these variations were accounted for.
Reproducibility of ROI Delineation: While the ICC of >95% is excellent, it appears to reflect intra-observer reliability. Please comment on inter-observer reliability or the potential for automated segmentation in the future.
Clarification of Bone Density Ranges: In the classification of T-scores, some notations (e.g., "2 -1.0") are confusing. Please use clear mathematical inequalities (e.g., $\ge -1.0$) to define Normal, Osteopenia, and Osteoporosis groups.
Figure Quality and Labeling: The scatter plots in Figures 6–18 are difficult to read. Please improve the resolution and ensure that axis labels, units, and regression equations are clearly legible.
Artifact Management: The manuscript mentions that metal artifacts affected ROI size. Please specify if any specific metal artifact reduction (MAR) filters or software were used during processing.
Adjustment for Age and Gender: Given that age and gender are primary drivers of bone density, please consider including a multivariate analysis (e.g., multiple linear regression) using these as covariates.
Consistency in Terminology: The terms "3D texture analysis" and "Radiomics" are used interchangeably. Please provide a clear definition or ensure consistent usage throughout the text.
Clinical Cut-off Values: For a screening tool, preliminary diagnostic performance metrics (e.g., Sensitivity, Specificity, or ROC curves) for the most significant features would greatly enhance the paper’s impact.
Nuance in Conclusion: Since this is a pilot study, the conclusion should avoid definitive statements like "CBCT can diagnose osteoporosis." Instead, please frame it as a "potential supplementary tool for risk assessment."
Updating References: As this is a 2025 submission, I recommend including more recent literature from 2023–2024 regarding AI and radiomics in maxillofacial imaging.
Typographical Errors: Please perform a thorough check for formatting errors in symbols (e.g., $\times$ signs and inequalities) and ensure all units are consistently provided in the figures.

Comments on the Quality of English Language

The English language and style are generally good and the paper is easy to follow. Only minor spell-checking and a few adjustments to the formatting of mathematical symbols and units are required.

Author Response

We sincerely thank the Reviewer for the careful reading of our manuscript and for the constructive comments. We appreciate the positive assessment of the study concept, particularly the combination of clinical CBCT data with AI-generated synthetic phantoms. The manuscript has been revised accordingly, with clarification of several methodological aspects and further emphasis on the exploratory nature of this pilot study.

We clarified the rationale for selecting the seven anatomical ROI sites and added a more explicit discussion of their practical limitations.
We clarified that no formal correction for multiple comparisons was applied because this was a pilot exploratory study intended to identify candidate radiomic features for further validation. We also added a brief interpretation of the main feature groups, including GLCM-, ARM-, and Gradient-derived parameters.
We clarified that the synthetic phantoms were used as a controlled validation framework and acknowledged that a direct quantitative comparison between phantom-derived and clinical regression trends was beyond the scope of the present pilot study.
We acknowledged more explicitly that, due to the retrospective design, detailed information on medications affecting bone metabolism and systemic comorbidities was not available in a complete and standardized form for all patients.
We clarified that the study was based on retrospective CBCT data acquired under routine clinical conditions. Although image intensities were normalized, scanner-related effects could not be systematically assessed and are now discussed as a limitation.
We clarified that the reported ICC reflects intra-observer rather than inter-observer reproducibility and acknowledged the lack of inter-observer assessment as a limitation.
The definitions of normal bone density, osteopenia, and osteoporosis were revised for clarity using explicit mathematical inequalities.
To improve readability, selected correlation plots were moved to the Supplementary Material, while the most representative figures were retained in the main text.
We clarified that no dedicated metal artifact reduction filter or MAR software was used and added this point to the methodological limitations.
We acknowledged that no multivariable adjustment for age and gender was performed in this pilot analysis and indicated this as an important direction for future studies.
We clarified the use of the term “3D texture analysis” in the context of radiomics-based analysis of CBCT data and improved terminological consistency throughout the manuscript.
We acknowledged that diagnostic cut-off values and ROC-based performance measures were beyond the scope of this exploratory pilot study and should be addressed in larger validation cohorts.
The conclusion was revised to maintain a more cautious tone and to present CBCT-based 3D texture analysis as a potential supplementary tool rather than a stand-alone diagnostic method.
The reference list was updated with more recent literature on radiomics, AI, and opportunistic bone assessment.
The manuscript was checked for typographical and formatting inconsistencies.

Reviewer 4 Report

Comments and Suggestions for Authors

Your manuscript tackles a really intriguing and clinically important topic. The whole idea of using CBCT-based radiomics for early osteoporosis detection feels fresh and full of promise. The study design is solid overall, and I like the concept a lot.
That said, a few things could use some polishing to make it stronger:
In the Results section, there’s just a ton of data—parameters and tables everywhere—which can make it tough to follow. I’d suggest streamlining it a bit, highlighting the key findings that matter most clinically, and wrapping it up with a strong “take-home message.” That would make it way more readable and punchy.
On the statistical analysis, I’m not totally clear why you didn’t lean more into correlation analysis. Bringing that in prominently could really sharpen the links between your radiomic features and bone density, helping readers (and clinicians) grasp the implications better.
The Discussion needs some tightening too. Focus more on stacking your results up against what’s already out there in the literature, and dig deeper into a critical take on what you’ve found. Especially, spell out the real-world clinical value—why this approach beats the usual stuff in daily practice.

The manuscript has clear potential, but revisions are needed to improve clarity, interpretation, and clinical applicability of the results.

Author Response

We sincerely thank the Reviewer for the positive and encouraging assessment of our manuscript and for the helpful suggestions. We appreciate the recognition of the originality and potential clinical relevance of the proposed CBCT-based radiomics approach.

We streamlined the Results section by emphasizing the most clinically relevant findings and moving selected figures with overlapping information to the Supplementary Material.
We clarified that correlation with DXA T-score constituted the central analytical framework of this pilot study and highlighted the strongest associations among the radiomic features.
The Discussion was revised to provide a more focused comparison with the available literature and to better place the findings in the context of opportunistic bone assessment.
We further clarified the potential clinical value of the method as a supplementary opportunistic screening strategy based on CBCT/CT examinations already performed for other indications, rather than as a replacement for DXA.
We believe these changes have improved the clarity, interpretation, and clinical relevance of the manuscript.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have made the necessary corrections to my comments. I congratulate them on this successful and detailed work. Sincerely.

Author Response

Thank you very much for your positive evaluation and kind words. We greatly appreciate your time, thoughtful review, and supportive feedback.

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have carefully addressed the majority of the reviewer’s previous comments, and the manuscript has been substantially improved in terms of clarity, methodological transparency, and scientific balance.

In particular, the authors appropriately clarified the exploratory nature of the study, acknowledged the absence of multiple testing correction, and revised the conclusions to reflect the preliminary character of the findings. The addition of explanations regarding ROI selection and study limitations is also appreciated.

The integration of synthetic phantom analysis with clinical data remains a strong and innovative aspect of the study, and the revised manuscript presents this component more clearly.

I have only a few minor comments that could further strengthen the manuscript:

Although the exploratory nature is now clearly stated, the manuscript would benefit from a brief forward-looking statement outlining how the identified features could be incorporated into a predictive model in future studies.
The influence of imaging parameters, particularly tube current (mA), is now described in the Results. However, a short discussion paragraph explicitly addressing this as a potential confounder and its implications for standardization in future radiomics studies would improve the manuscript.
While the ROI selection rationale has been clarified, it may still be useful to briefly suggest which anatomical region might be most practical for clinical implementation in future screening settings.

Overall, this is a well-conducted pilot study with an innovative methodological approach. After these minor revisions, the manuscript would be suitable for publication.

Comments on the Quality of English Language

The English language and style are generally good and the paper is easy to follow. Only minor spell-checking and a few adjustments to the formatting of mathematical symbols and units are required.

Author Response

We sincerely thank the Reviewer for the very positive and thoughtful evaluation of our revised manuscript. We greatly appreciate the recognition that the manuscript has been substantially improved in terms of clarity, methodological transparency, and scientific balance. We are also grateful for the positive assessment of the integration of synthetic phantom analysis with clinical data, which we also consider a key and innovative component of the study.

In response to the remaining minor comments, we made several additional clarifications in the revised manuscript. We added a brief forward-looking statement indicating that the most informative radiomic features identified in this pilot study may serve as candidate variables for the future development of a predictive osteoporotic/osteopenic model in larger validation cohorts. We also expanded the discussion of tube current (mA) as a potential confounder and further emphasized the need for scanner and protocol standardization in future radiomics studies. In addition, we added a brief comment on the practical clinical implementation of ROI selection, indicating that anatomically accessible regions that are consistently visible on routine CBCT examinations may be the most suitable for future screening applications, although this will require dedicated validation. Minor spell-checking and formatting corrections were also made throughout the manuscript. We thank the Reviewer again for the constructive feedback and for the positive final assessment of our work.

Reviewer 4 Report

Comments and Suggestions for Authors

Thank you for the revisions made. I suggest that the manuscript be accepted

Author Response

Thank you very much for your positive evaluation and recommendation. We sincerely appreciate your time, careful review, and supportive feedback.