Automatic Method for Bone Segmentation in Cone Beam Computed Tomography Data Set

Round 1

Reviewer 1 Report

In this study, the authors propose a new method for automatic segmentation of craniomaxillofacial bone structures from cone beam
computed tomography datasets. Based on the manuscript, the proposed method provides accurate and fast segmentation. The automated
segmentations were compared to manual segmentation by a surgeon. The methods are well described and clearly visualized, and the the
manuscript is nicely structured. The language is ok in most parts, but there are some grammatic issues that repeat throughout the
manuscript and have to be corrected for.

Below is a (probably incomplete) list of wordings/sentences that need to be corrected. Moreover, the manuscript would benefit from a
review of the grammar by a native English speaker.

1) Wordings "were got" and "is got" should be changed to something more descriptive and scientific
- at least on lines: 52, 53, 184, 198, 199, 218

2) When listing things, the word "and" should be used before the last item
- at least on lines: 27, 54, 75

3) Following sentences need to be revised for grammar and/or clarity
- line 46: Remove period after "min": "(20 min.)" -> "(20 min)"
- lines 48-49: "..however, they also as previous authors, tested..."
- line 78: "Therefore mentioned anatomical..." Add comma after "Therefore"
- line 103: "...make analysed histograms a bimodal." Remove "a"
- lines 168-170: This sentence is unclear.
- line 195: "...of DSC were similar then comparing with other studies..." Word "then" is used improperly
- line 219: "It was caused that the coordinates..." Not clear
- line 220: "...an equal as a segmented..." Not clear
- line 227: "not accurate" change to "inaccurate"
- line 239: "The presented study revealed a new automatic..." The word "revealed" is used improperly. Consider changing to "proposed"
or something similar
- line 240 "An important features of the..." Mixture of singular and plural forms.

4) What does "voxel side" mean on line 131?

Author Response

Hello, we are very grateful for your comments.

We corrected our manuscript by your list: 1), 2), 3), 4). Each note are corrected and marked by a different color in PDF file:

1) yellow
2) red
3) green
4) blue

All manuscript has been reviewed by the authors again. Incorrect sentences/wordings have been corrected. Please see the attachment (corrected PDF file).

 

Kind regards,

Authors

Author Response File: Author Response.pdf

Reviewer 2 Report

Cone-beam computed tomography (CBCT) scans have been commonly used in diagnosing and planning surgical or orthodontic treatment to correct craniomaxillofacial (CMF) deformities.  This process comprises two main tasks (i) bone segmentation and (ii) the precise identification of anatomical landmarks to generate an accurate 3D model of CMF structures and the identification of clinical landmarks. In this sense, this work can be considered one more approach for 3D bone segmentation on CBCT.

From the practical point of view the proposed method appear to me valid and feasible to be implemented. The proposed paper is well-written following the current rules for scientific works, mathematical formulas have been well-used / defined, graphical figures show a good quality, and its are not needed important corrections of the English language . However, in simple search at the ScienceDirect database with the keywords “Bone Segmentation Cone Beam Computed Tomography”, I had found 1249 results (published papers), including different approaches from classical image processing algorithms and methods to computer vision machine (deep) learning based approaches. Therefore, before to accept this paper for publication, I ask authors to answer the following questions / issues:

If it is possible, I think it will be not a problem to identify the Clinic of Orthonagthic Surgery where this work was piloted. Consider to extend the dataset with more patients' cases, due to the fact that 20 patients' cases is not sufficient to guarantee the validity of the proposed method (i.e., I guess that are 20 cases where sets of CBCTs images were capture before and after the surgery), it is true?. In my opinion the main handicap of this work the novelty of the proposed method, due to it appear to be a proper integration of previous developed algorithms and methods. Please, can you explain better, which are the major contributions of the authors beyond the state of the art?

Author Response

Hello, we are very grateful for your comments. All answers to your questions are presented bellow. Also corrected manuscript is in attachment. Please see the attachment. 

Cone-beam computed tomography (CBCT) scans have been commonly used in diagnosing and planning surgical or orthodontic treatment to correct craniomaxillofacial (CMF) deformities.  This process comprises two main tasks (i) bone segmentation and (ii) the precise identification of anatomical landmarks to generate an accurate 3D model of CMF structures and the identification of clinical landmarks. In this sense, this work can be considered one more approach for 3D bone segmentation on CBCT.

From the practical point of view the proposed method appear to me valid and feasible to be implemented. The proposed paper is well-written following the current rules for scientific works, mathematical formulas have been well-used / defined, graphical figures show a good quality, and its are not needed important corrections of the English language. However, in simple search at the ScienceDirect database with the keywords “Bone Segmentation Cone Beam Computed Tomography”, I had found 1249 results (published papers), including different approaches from classical image processing algorithms and methods to computer vision machine (deep) learning based approaches. Therefore, before to accept this paper for publication, I ask authors to answer the following questions / issues:

1) If it is possible, I think it will be not a problem to identify the Clinic of Orthonagthic Surgery where this work was piloted.

Answer:

1.1 Simonas Grybauskas' Orthognathic Surgery, S'OS, Vytenio str. 22 / Naugarduko str. 41, Vilnius (2.1 Subsection - Data acquisition)

2) Consider to extend the dataset with more patients' cases, due to the fact that 20 patients' cases is not sufficient to guarantee the validity of the proposed method (i.e., I guess that are 20 cases where sets of CBCTs images were capture before and after the surgery), it is true?

Answer:

2.1. Yes, it is true, this sample is not yet sufficient to guarantee the validity of the proposed method. However we think, that this sample makes a possibility to call proposed method as accurate for segmentation of bones from CBCT data. At this moment we don‘t have possibility increase the amount of CBCT data sets. Also a lot of studies used similar amount of CBCT data sets in order to prove segmentation accuracy. In further study, we will increase the sample of CBCT data sets. Due to language barrier, the word validity or validation, that was previously used in present study, was too ‘intense‘ or too ‘strong‘, so we changed it to more appropriate word (2.4 Subsection - Evaluation of method accuracy).

3) In my opinion the main handicap of this work the novelty of the proposed method, due to it appear to be a proper integration of previous developed algorithms and methods. Please, can you explain better, which are the major contributions of the authors beyond the state of the art?

Answer:

3.1 In this work, we are proposing an automatic method for bone segmentation from CBCT data. Proposed method of segmentation is based on locally found values of global threshold. It means that more than one global value of threshold is used to perform a bone segmentation. We have supplemented our introduction with a recently published article [11] in order to base that fully automatic methods are in demand. We want to highlight two main benefits of the proposed method. The first – simplicity of the implementation of the bone segmentation by using our method. Proposed method does not require to have the computer with a high computation power also it doesn‘t require a special training for algorithm. Every time performing the segmentation is enough just to apply proposed method for the CBCT data set without additional settings. The second – proposed method performs segmentation very rapidly (~46 s/case) comparing with other methods. Segmentation is an automatic and doctor can save a time. Saved time can be used for the planning of surgery. Due to these reasons proposed method is convenient for user. These two basic benefits provide an opportunity to integrate proposed method in the most popular software designed to support oral and maxillofacial surgery.

Kind regards,

Authors

Author Response File: Author Response.pdf