Classification of Metallic Powder Morphology Using Traditional and Automated Static Image Analysis: A Comparative Study

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The work is a comparative study between platforms aiming to detect and measure particle features. In particular morphological indexes are taken into consideration. The introduction and paper structure are well formed and methodology is well clarified. Results are well presented. Some points are suggested in the following:

For the sake of completeness please describe the automatic bridge removal tool indicated at page 13 line 304.

At page 14, please deepen the issues indicated in the paragraph between line 364 and 369.

Authors claim the image resolution sensitivity is affected by the resolution (page 15 line 392): please deepen this aspect by adding numerical elements to judge and recover from possible lack of accuracy. In particular try to determine which indexes could be mainly affected. The resolution is recalled also at line 410: add an explaination why this is a special case.

Some items were found in the manuscript and are listed in the following.

Page 4 line 124: there is a ";" in place of ":"

Page 5 line 148: the reference is missing

Figure 5: the title of each element is cut and the information related to the scale is unreadable

Page 10 line 247: a typo "xtend"

Page 13 line 308 the line is broken

 

 

Author Response

We sincerely thank the reviewers for their insightful comments and valuable feedback, which have greatly contributed to enhancing the quality of our manuscript.

Comment 1:
For the sake of completeness please describe the automatic bridge removal tool indicated at page 13 line 304.

Response 1:
Clemex does not provide detailed information on the separation process. However, we have included additional information indicating that the tool applies instructions to differentiate and isolate binary objects that are juxtaposed or partially overlapping.

Comment 2:
At page 14, please deepen the issues indicated in the paragraph between line 364 and 369.

Response 2:
We have elaborated on the issues by emphasizing that 2D imaging captures only the height and width of particles, omitting depth information. This limitation is critical because depth is necessary for accurately representing the true geometry of irregularly shaped particles.

Comment 3:
Authors claim the image resolution sensitivity is affected by the resolution (page 15 line 392): please deepen this aspect by adding numerical elements to judge and recover from possible lack of accuracy. In particular try to determine which indexes could be mainly affected. The resolution is recalled also at line 410: add an explanation why this is a special case.

Response 3:
We apologize for the oversight in not initially referring to previous work that addresses this issue. We have now included the sentence: "As described in previous work, increasing the resolution decreases the pixel size, which can lead to a higher measured perimeter length due to the more detailed capture of the object's edges [13]." In this work, we have already demonstrated the impact of magnification on the sphericity index by using different magnifications.

Comment 4:
Some items were found in the manuscript and are listed in the following.
Page 4 line 124: there is a ";" in place of ":"
Page 5 line 148: the reference is missing
Figure 5: the title of each element is cut and the information related to the scale is unreadable
Page 10 line 247: a typo "xtend"
Page 13 line 308 the line is broken

Response 4:
Line 124: The paragraph originally beginning at line 124 has been rewritten and now starts at line 142, ending at line 151.
Line 148: We have added the missing reference to the ISO standard at line 166. Please verify if this addresses your concern regarding the reference.
Figure 5: The figure has been revised to ensure that all titles and scale information are complete and legible, providing clarity for the reader.
Line 247: The typo "xtend" was corrected at line 266.
Line 308: The line break was removed to adjust the formatting at line 333.

Reviewer 2 Report

Comments and Suggestions for Authors

 

Overall, this paper needs major revision. A long list of comments is provided but mostly the authors need to be clearer and more concise on wording and visuals. The purpose of the paper was well laid out and easily followed throughout the paper. The authors were able to sort out key differences and similarities between the software as their focus intended, but it is not sure how it supports the objective of “implementing a methodology in several software and analyzers.” This should be better explained/connected.

 

 

• List references before period
• What scientific value does Fig. 1 add? All it tells the reader is that two different softwares return different values on the same data.
• In section 2.1, what does “schematized particles” mean? “Segmented particles”, perhaps? “Representative particles”, perhaps?
• Document the resolution of the images used as input to the softwares and any manipulations that are performed on the images before analyzing shapes etc.
• Missing Oxford comma multiple times throughout the document. When listing items, make sure to include the final comma before the final listed item. Example: “Powder qualification is based on various factors such as chemistry, flowability, particle size distribution, density, and shape.”
• The new standard acronym for laser powder bed fusion is “PBF-LB” not “LPBF”
• There are some grammatical errors throughout the document, so be sure to double check.
• Line 76-88: Are you making the argument that there isn't an option of choosing sphericity vs circularity in each software? Are you saying both software only provides these individual options, assuming they mean the same thing?
• Line 92-97: You say that the “objective” is to implement a new methodology but the paper concentrates on checking the extent of agreement (or otherwise) between the different softwares used. Is the focus of this paper just the first step in achieving this objective and there are more steps to follow?
• Line 104: convex hull perimeter
• Section 2.1: I would describe the different parameters before listing the threshold values. That way, I understand what you are talking about before a number is listed.
• Line 101: You say that “the threshold for each descriptor is established first through an analysis of schematized particles, then validated on metallic powder samples” and then the threshold values are listed. Are these the values you found through this analysis? Are they from previous work/literature? It’s not entirely clear why one should trust these values for each variable.
• Line 109: It is stated that Er is “based on the second moment of the object.” What does this mean?
• Section 2.2: Will all particles be identified with one of these descriptions or will some be left over and labelled as "unknown"? Can particles fall into multiple groups? Or if it previously identified with one group, does that eliminate it from the other groups? What if a particle has more than nine satellites or more than nine facets, etc.?
• Figure 3: I’m not sure if there are scale bars in each group, but it is too small and never mentioned/explained.
• Section 2.3: This section can be further explained in more detail. Example: What size grit was used? What size suspensions were used? How long was each polishing step? Did all the metal materials follow the same procedure?
• Section 2.4, Line 148: Error message for reference
• Figure 5: Title of each image is covered/not fully seen. Bottom row of images overlap with the top row. These could be expanded. It is hard to see/distinguish any changes in the images going from the Gray Thresholding to Frame Delineation to Particle Splitting.
• Line 162: Be sure to state “best-fit ellipse length”
• Line 158-179: Be careful with how distinguishing which formulas are for which software. I would state what the software is before the formula/equation so that the reader knows which one it is for before understanding the formula/equation. I would also suggest reformatting the equations so that it is clear when the variables are identifying one value and is being multiplied by something else. Example:
• Line 170 is one example (but it happens throughout the paper) where if you are listing two things and their respective values or variables, be sure to state “…. , respectively.”
• Section 2.4: It might be good to include visuals/schematics of what these variables measure, especially for the ones where the metrics are measured differently.
• Equation 10: Can this, or something similar, also be used to compare with the measurements from Morphologi with Clemex and ImageJ?
• Table 1: Image scale bars are blurry and too small. For the particles with facets, it looks like the images are rearranged compared to when they first appear earlier in the paper in Figure 3. Does this affect any of the results? I would be careful and try to keep everything consistent. You also do not mention what “SD (%)” is anywhere in the paper. Is it standard deviation? Is it something else? I would also add more horizontal lines in the table to break up the different groups of information. It is all to close and confusing.
• Line 191-203: This section could be better explained. I’m not sure I completely follow the reasoning of why there is differences in the perimeter calculations. Visuals are provided in Figure 6, but I don’t think they are well explained in the paper.
• Line 214: What value is considered "numerous"? Is it a value from the list of 1 - 9 number of satellites?
• Line 215: You state “Cx must be lower than 0.95 for Clemex and 0.90 for ImageJ”; is this seen on the graph in Figure 7? It is unclear how you came to this conclusion.
• Line 236-239: How do you know that "particle 1" in Clemex is the same "particle 1" in ImageJ? Are you checking if the identification of each particle is the same for both software?
• Line 242: Was this specific adjustment thresholding procedure described?
• Line 249: The graphs show white bars, not dotted.
• Line 252-255: How did you discover this "corrected threshold" where it is okay to not use the convexity index? Why are these values of 0.955 and 0.95 more reliable if you are no longer including the convexity index? I'm not convinced how this approach allows for consistency based on Figure 11. What percent difference is considered acceptable?
• Figures should not have boxes around them, e.g., Figs. 11, 12
• Line 296: You say the threshold was adjusted. Adjusted to what? Is there a new value? How was it adjusted?
• Line 302: You only mention analyzing the Aluminum powder. Was this an issue for the other powder materials?
• Line 308-309: Format error. Is there an approximation for how much it overestimates?
• Line 334: You mention comparing results with Clemex. What about ImageJ? Any comparisons to be made?
• Figure 16: What is the y-axis measuring? Units?
• Line 350-353: Any figures to show this?
• Section 4.1:
  • Algorithm Variability: specify “perimeter of the powder particle.”
  • Image Resolution Sensitivity: Is there a minimum resolution limit that is preferred for these analyses?
• Abbreviations Section: MIC = Minimum Inscribed Circle
• Discussion and Conclusion sections should be laid out in a table to make it more clear and visually easier to read/understand discrepancies found from measuring certain descriptors from one software to the next and any recommendations that the authors may have.
• Suggest adding references to one or more of the classical texts on stereology that describe many of the basic geometric analyses, e.g., by Underwood or Russ & DeHoff.

 

Author Response

We sincerely thank the reviewers for their insightful comments and valuable feedback, which have greatly contributed to enhancing the quality of our manuscript.

Comment 1:
List references before period

Response 1:
Periods were moved to follow the references.

Comment 2:
What scientific value does Fig. 1 add? All it tells the reader is that two different softwares return different values on the same data.

Response 2:
Figure 1 was removed from manuscript.

Comment 3:
In section 2.1, what does “schematized particles” mean? “Segmented particles”, perhaps? “Representative particles”, perhaps?

Response 3:
Section 2.2 explains that "schematized particles" are software-designed representations based on common morphologies found in metallic powders, used for validation and calibration of image analysis methods. We have added a sentence in section 2.1 to clarify their purpose.

Comment 4:
Document the resolution of the images used as input to the softwares and any manipulations that are performed on the images before analyzing shapes etc.

Response 4:
In Section 2.3, we clarify that while image resolution is not the ultimate criterion, achieving a sufficient number of pixels per particle is crucial for accurate analysis. We have added details about the minimum number of pixels per particle required in our study to ensure precision. We added all image manipulations performed before analyzing shapes in Section 2.4

Comment 5:
Missing Oxford comma multiple times throughout the document. When listing items, make sure to include the final comma before the final listed item. Example: “Powder qualification is based on various factors such as chemistry, flowability, particle size distribution, density, and shape.”

Response 5:
We have reviewed the document and added the Oxford comma to all relevant lists, ensuring clarity and consistency in accordance with standard punctuation guidelines.

Comment 6:
The new standard acronym for laser powder bed fusion is “PBF-LB” not “LPBF”

Response 6:
The acronym has been updated to "PBF-LB/M" throughout the document, in accordance with the ISO/ASTM 52911 standard.

Comment 7:
There are some grammatical errors throughout the document, so be sure to double check.

Response 7:
Thank you for your feedback regarding grammatical errors in the document. We have thoroughly reviewed and revised the manuscript to address these issues. This involved correcting typographical errors, ensuring consistent use of technical terminology, and improving sentence structure for clarity.

Comment 8:
Line 76-88: Are you making the argument that there isn't an option of choosing sphericity vs circularity in each software? Are you saying both software only provides these individual options, assuming they mean the same thing?

Response 8:
The paragraph aims to highlight the potential discrepancies in measurements due to differences in how circularity and sphericity are defined across different software. Our intention is not to suggest that users cannot choose between sphericity and circularity in each software, but rather to emphasize the need for standardized definitions and methodologies to ensure consistency across different systems. We have clarified the paragraph by explicitly defining circularity and sphericity and explaining how these terms are used differently in systems like Clemex and ImageJ. The paragraph was modified accordingly to better communicate this distinction.

Comment 9:
Line 92-97: You say that the “objective” is to implement a new methodology but the paper concentrates on checking the extent of agreement (or otherwise) between the different softwares used. Is the focus of this paper just the first step in achieving this objective and there are more steps to follow?

Response 9:
The objective of this paper is indeed the first step in achieving a larger goal. The paragraph has been revised to clarify this focus.

Comment 10:
Line 104: convex hull perimeter

Response 10:
Typo was corrected.

Comment 11:
Section 2.1: I would describe the different parameters before listing the threshold values. That way, I understand what you are talking about before a number is listed.

Response 11:
The list of the different descriptors was transferred to section 2.1 before discussing the thresholds.

Comment 12:
Line 101: You say that “the threshold for each descriptor is established first through an analysis of schematized particles, then validated on metallic powder samples” and then the threshold values are listed. Are these the values you found through this analysis? Are they from previous work/literature? It’s not entirely clear why one should trust these values for each variable.

Response 12:
The values of the threshold were indeed established in our previous work, where we conducted a thorough analysis of schematized particles and validated these thresholds on metallic powder samples. The paragraph has been reedited to clearly reflect this.

Comment 13:
Line 109: It is stated that Er is “based on the second moment of the object.” What does this mean?

Response 13:
The line stating that Er is "based on the second moment of the object" has been removed because the definitions of descriptors have been moved to an earlier section for clarity, in line with the reviewer's suggestion.

Comment 14:
Section 2.2: Will all particles be identified with one of these descriptions or will some be left over and labelled as "unknown"? Can particles fall into multiple groups? Or if it previously identified with one group, does that eliminate it from the other groups? What if a particle has more than nine satellites or more than nine facets, etc.?

Response 14:
In the revised paragraph, we clarified the selection and testing of particle shapes using image analysis software.

Comment 15:
Figure 3: I’m not sure if there are scale bars in each group, but it is too small and never mentioned/explained.

Response 15:
You are correct that the scale bars in Figure 3 were generic bars generated by AutoCAD, and they do not provide pertinent information for the analysis presented. To prevent any confusion and enhance clarity, these scale bars have been eliminated from the figure.

Comment 16:
Section 2.3: This section can be further explained in more detail. Example: What size grit was used? What size suspensions were used? How long was each polishing step? Did all the metal materials follow the same procedure?

Response 16:
While the detailed procedure for polishing is not the primary focus of this paper, we have added information to the paragraph about the grit size, polishing time, and suspension used.

Comment 17:
Section 2.4, Line 148: Error message for reference

Response 17:
Hyperlink was eliminated to avoid this issue.

Comment 18:
Figure 5: Title of each image is covered/not fully seen. Bottom row of images overlap with the top row. These could be expanded. It is hard to see/distinguish any changes in the images going from the Gray Thresholding to Frame Delineation to Particle Splitting.

Response 18:
The image has been adjusted to address the reviewer's concerns.

Comment 19:
Line 162: Be sure to state “best-fit ellipse length”

Response 19:
The term "best-fit ellipse length" has been added to the text to ensure clarity.

Comment 20:
Line 158-179: Be careful with how distinguishing which formulas are for which software. I would state what the software is before the formula/equation so that the reader knows which one it is for before understanding the formula/equation. I would also suggest reformatting the equations so that it is clear when the variables are identifying one value and is being multiplied by something else. Example:

Line 170 is one example (but it happens throughout the paper) where if you are listing two things and their respective values or variables, be sure to state “…. , respectively.”

Response 20:
We have restructured the text to clearly indicate which formulas correspond to each software before presenting the equations. We have also reformatted the equations for clarity and used "respectively" where applicable to ensure clear identification of values and variables.

Comment 21:
Section 2.4: It might be good to include visuals/schematics of what these variables measure, especially for the ones where the metrics are measured differently.

Response 21:
We have restructured the text to provide clearer explanations of the different variables and their measurement across software platforms. Given the detailed explanations provided, we believe the text sufficiently supports and clarifies the different variables.

Comment 22:
Equation 10: Can this, or something similar, also be used to compare with the measurements from Morphologi with Clemex and ImageJ?

Response 22:
Equation 10 is designed to facilitate one-on-one comparisons of each measurement for the same particle, which is not applicable to Morphologi because the particles are spread on a surface rather than being part of metallographic samples.

Comment23:
Table 1: Image scale bars are blurry and too small. For the particles with facets, it looks like the images are rearranged compared to when they first appear earlier in the paper in Figure 3. Does this affect any of the results? I would be careful and try to keep everything consistent. You also do not mention what “SD (%)” is anywhere in the paper. Is it standard deviation? Is it something else? I would also add more horizontal lines in the table to break up the different groups of information. It is all to close and confusing.

Response 23:
Thank you for your detailed feedback. Table 1 has been adjusted to address the issues you raised.

Comment 24:
Line 191-203: This section could be better explained. I’m not sure I completely follow the reasoning of why there is differences in the perimeter calculations. Visuals are provided in Figure 6, but I don’t think they are well explained in the paper.

Response 24:
We have revised the section to provide a clearer and more detailed explanation of the distinct methodologies used by each software.

Comment 25:
Line 214: What value is considered "numerous"? Is it a value from the list of 1 - 9 number of satellites?

Response 25:
Thank you for pointing this out. In the context of our study, "numerous" refers to particles with more than one satellite. For clarity, we have updated the text

Comment 26:
Line 215: You state “Cx must be lower than 0.95 for Clemex and 0.90 for ImageJ”; is this seen on the graph in Figure 7? It is unclear how you came to this conclusion.

Response 26:
Thank you for your observation. The thresholds for the convexity index—0.95 for Clemex and 0.90 for ImageJ—are derived from the data presented in Figure 6 (initially Figure 7). We have clarified this connection in the text to ensure that the rationale behind our threshold selection is clear.

Comment 27:
Line 236-239: How do you know that "particle 1" in Clemex is the same "particle 1" in ImageJ? Are you checking if the identification of each particle is the same for both software?

Response 27:
Yes, we ensure that "particle 1" in Clemex corresponds to the same "particle 1" in ImageJ by comparing each particle individually. ImageJ provides the option to number each particle, which facilitates direct comparison. On the other hand, Clemex highlights particles individually, allowing us to cross-reference and verify that we are analyzing the same particle across both software. This careful alignment ensures consistency in our analysis and results.

Comment 28:
Line 242: Was this specific adjustment thresholding procedure described?

Response 28:
You are correct we did not specify the adjustment procedure, where we tested different thresholds with Image J and compared which particles were classified as containing more than one satellite. We were not able to obtain comparable results. We decided to delete the passage “even after adjusting the threshold”, since it does not provide pertinent information and the outcome of not using the convexity index is based on the unreliability of the perimeter measurement.

Comment 29:
Line 249: The graphs show white bars, not dotted.

Response 29:
We have updated the text to accurately reflect this.

Comment 30:
Line 252-255: How did you discover this "corrected threshold" where it is okay to not use the convexity index? Why are these values of 0.955 and 0.95 more reliable if you are no longer including the convexity index? I'm not convinced how this approach allows for consistency based on Figure 11. What percent difference is considered acceptable?

Response 30:
Thank you for your questions and for highlighting the need for further clarification regarding the "corrected threshold." The threshold values of 0.955 and 0.95 were determined through an iterative trial-and-error process, where we tested various thresholds to find the values that maximized the similarity in particle categorization between Clemex and ImageJ. This method involved comparing the percentage of particles with satellites identified by both software tools and adjusting the thresholds until the results were most consistent. The decision to forego the convexity index was based on the observation that the perimeter measurement's unreliability affected its calculation. Regarding the acceptable percent difference, we considered variations within a few percentage points to be acceptable, as this range maintained the integrity of the analysis and provided a robust comparison between the software outputs. We acknowledge the limitations illustrated in Figure 11 and have aimed to address these through the chosen methodology.

Comment 31:
Figures should not have boxes around them, e.g., Figs. 11, 12

Response 31:
We have corrected the layout to remove the boxes around Figures

Comment 32:
Line 296: You say the threshold was adjusted. Adjusted to what? Is there a new value? How was it adjusted?

Response 32:
Thank you for pointing out the ambiguity. The original sentence was not clear. It should convey that adjusting the threshold for each powder type might be necessary to achieve accurate particle categorization.

Comment 33:
Line 302: You only mention analyzing the Aluminum powder. Was this an issue for the other powder materials?

Response 33:
The issue with the bridge removal tool potentially impacts all particle types, including those from other powder materials. However, we specifically mentioned aluminum particles to highlight that the impact of this tool might be more pronounced for spherical particles, like those found in aluminum, compared to more irregularly shaped particles from other powders. This distinction was adjusted in the text to ensure clarity.

Comment 34:
Line 308-309: Format error. Is there an approximation for how much it overestimates?

Response 34:
The line break was removed to correct the formatting error. Regarding the overestimation, we did not quantify the extent of overestimation in this study. However, this is an important aspect that we acknowledge could be explored in future work to provide a more precise understanding of the impact on particle classification.

Comment 35:
Line 334: You mention comparing results with Clemex. What about ImageJ? Any comparisons to be made?

Response 35:
While the original text specifically mentioned comparing results with Clemex, it is important to note that the number of particles with satellites was previously found to be similar between Clemex and ImageJ. We have updated the text to include ImageJ in the comparison

Comment 36:
Figure 16: What is the y-axis measuring? Units?

Response 36:
The y-axis measures the percentage of particles. We have added a title to the figure to clearly indicate this information and ensure clarity for readers.

Comment 37:
Line 350-353: Any figures to show this?

Response 37:
The information discussed in lines 350-353 is represented in Figure 15 (initially labeled as Figure 16). In this figure, the particles were identified as polygonal instead of facetted. The figure has been corrected to reflect the appropriate terminology and ensure consistency with the text.

Comment 38:
Section 4.1:

Algorithm Variability: specify “perimeter of the powder particle.”

Response 38:
The text in Section 4.1 has been updated to specify "perimeter of the powder particle" to clarify the context of algorithm variability.

Comment 39:
Image Resolution Sensitivity: Is there a minimum resolution limit that is preferred for these analyses?

Response 39:
The minimum resolution limit preferred for these analyses is guided by the ISO 9276-6 standard, which specifies the minimum number of pixels needed for an accurate analysis of the smallest particles. This information has been added to the text to provide clarity and ensure that the resolution requirements are clearly understood.

Comment 40:
Abbreviations Section: MIC = Minimum Inscribed Circle

Response 40:
The typo in the Abbreviations Section regarding MIC has been corrected.

Comment 41:
Discussion and Conclusion sections should be laid out in a table to make it more clear and visually easier to read/understand discrepancies found from measuring certain descriptors from one software to the next and any recommendations that the authors may have.

Response 41:

The discussion and conclusion sections were rewritten to address this comment.
Comment 42:
Suggest adding references to one or more of the classical texts on stereology that describe many of the basic geometric analyses, e.g., by Underwood or Russ & DeHoff.

Response 42:
Thank you for your suggestion. We have addressed your comment by adding a reference to Underwood's work in the discussion section. This inclusion highlights classical texts on stereology that describe fundamental geometric analyses, thereby enriching the context and depth of our study.

Reviewer 3 Report

Comments and Suggestions for Authors

This paper presents an experimental campaign to compare automated image analysis results, pointing at classifying defects in metal spherical particles to be used in additive Manufacturing processes. The experiments consisted of submitting the tomographic SEM images of powder samples embedded in epoxy resins to two different image analysis software. The same powders were analysed in motion using Malven Morfology instruments. Particle shapes are classified using specific shape indexes for which classification thresholds are inferred from experimental results.

The topic of the paper is of interest in the effort to provide a standardised approach to compare results obtained in different laboratories with different image analysis differences. This scope still makes an approach based on deterministic methods. I think, however, that, at least in the future direction, a mention of the potential of using AI tools to classify powder shapes would be advisable.

Given the above, I suggest the publication of this paper, but I strongly recommend improving the quality of the written text that suffers from internal inconsistency in the use of Symbols, the naming of variables and indexes, as well as the sequentiality of the information provided that presently make the reading of the paper uselessly complicate. In the following, a list of detailed notes is reported:

• Line 75. “…commercially…” -> “…commercial...”
• Lines 90-91. Figure 1 and caption. This figure and caption are confusing. Commonly, sphericity (i.e. 10.1016/j.partic.2020.09.006) is the ratio between the area of the surface of a sphere having the same volume of the particle related and the surface of the particle. Similarly, circularity is the perimeter of a circle having the same projected area of a particle and the perimeter of the particle’s projected area. A “perfect sphere”, indicated in the caption, would produce a value of 1 for both definitions. The equation provided for both indices is the one of circularity. So, the caption should explain that the numerical results provided correspond to the indexes named by the two software and that the deviation from unity is due to differences arising in the application of the automated image analysis procedures, which have different methods to determine the variables appearing in Figure 1. The meaning of A and P, non-used elsewhere, should be given in the caption.
• Section 2.1 and 2.2. Where they are given, Sections 2.1 and 2.2 can hardly be followed. The understanding would be immediate if these sections were placed after section 2.4.
• Line 148. A referencing error appears.
• Section 2.4. The use of different equations to define the same symbol is confusing. I suggest adding the context (Clemex, ImagJ or Morfologi) in which that equation is valid to the same line of the equation.
• Line 176. It should be specified for which parameter (projected area, perimeter or else) the equivalent diameter is calculated.
• Line 223-224, Figure 8 and its caption. The name “Extent C” has never been used before. Please use “Extent ratio (EC)” in the caption and Extent ratio in the axis label.
• Line 230, caption of Figure 9. “…Ellipse ratio …”-> “Ellipse ratio (Er)”
• Line 232, caption of Figure 19. “…Roundness index …”-> “Roundness index (Rn)”
• Lines 244-247. This paragraph is unclear and should be rewritten.
• Lines 249. The term “original methodology” is confusing. I suggest referring to a “Convexity threshold approach” and an “Extent ratio threshold approach”.
• Line 250 and following. The sentence “captured without using the convexity index” is very confusing. Perhaps it is not necessary to understand Figure 11. Instead, a clue as to why a certain threshold value is adopted should be given somewhere.
• Line 259-260. Regarding the sentence, “An ellipse ratio with a threshold of 0.90 provided the best results across all powders, as shown in Figure 12.” It is not clear what has been evaluated to draw this best threshold value. Please clarify.
• Lines 262-164. Concerning the sentence, “This consistency underscores the utility of the ellipse ratio as a reliable metric for distinguishing elongated particle shapes across various types of metallic powders, regardless of the imaging system employed.” Such a conclusion would have required a comparison with a method other than image analysis to evaluate particle elongation, which is not reported here.
• Line 276. The formula sR/rE shows symbols that were not defined previously. I suggest introducing this ratio in section 2.4. I understand that rE is Er in eq 2, so the symbols should be unified. Instead, rS has never been defined as it should.
• Line 305. A facetted surface could hardly be described as rough. Please verify the meaning.
• Lines 308-309. An unexpected carriage return appears in the line.
• Line 417, Table 3. I suggest adding a column in the table repeating the characteristics that the threshold values reported in the other line should highlight: The presence of satellites, elongated particles, Poligonal particles, and other non-spherical particles. If I understand well, the criterion with Rn&IR has been used to select spherical particles. In that case, the threshold definition should be >0.90, 0.89, 0.93, respectively.

Author Response

We sincerely thank the reviewers for their insightful comments and valuable feedback, which have greatly contributed to enhancing the quality of our manuscript.

Comment 1:
Line 75. “…commercially…” -> “…commercial...”

Response 1:
We have corrected the text by changing "commercially" to "commercial" on line 75 as per your suggestion.

Comment 2:
Lines 90-91. Figure 1 and caption. This figure and caption are confusing. Commonly, sphericity (i.e. 10.1016/j.partic.2020.09.006) is the ratio between the area of the surface of a sphere having the same volume of the particle related and the surface of the particle. Similarly, circularity is the perimeter of a circle having the same projected area of a particle and the perimeter of the particle’s projected area. A “perfect sphere”, indicated in the caption, would produce a value of 1 for both definitions. The equation provided for both indices is the one of circularity. So, the caption should explain that the numerical results provided correspond to the indexes named by the two software and that the deviation from unity is due to differences arising in the application of the automated image analysis procedures, which have different methods to determine the variables appearing in Figure 1. The meaning of A and P, non-used elsewhere, should be given in the caption.

Response 2:
Thank you for your detailed feedback. We have removed Figure 1 from the manuscript to avoid confusion. Instead, we have added more detail in the text to clarify that the same equation is defined as sphericity in Clemex and circularity in ImageJ.

Comment 3:
Section 2.1 and 2.2. Where they are given, Sections 2.1 and 2.2 can hardly be followed. The understanding would be immediate if these sections were placed after section 2.4.

Response 3:
To enhance the clarity and flow of the manuscript, we have transferred the list of different descriptors from Section 2.4 to Section 2.1. This change ensures that readers have a clear understanding of the descriptors before we discuss the thresholds.

Comment 4:
Line 148. A referencing error appears.

Response 4:
We have eliminated the hyperlink to resolve this issue.

Comment 5:
Section 2.4. The use of different equations to define the same symbol is confusing. I suggest adding the context (Clemex, ImagJ or Morfologi) in which that equation is valid to the same line of the equation.

Response 5:
We have restructured the text in Section 2.4 (now Section 2.1) to clearly indicate which formulas correspond to each software (Clemex, ImageJ, or Morphologi) before presenting the equations. Additionally, we have reformatted the equations for clarity.

Comment 6:
Line 176. It should be specified for which parameter (projected area, perimeter or else) the equivalent diameter is calculated.

Response 6:
We have clarified the definition of the equivalent diameter in the text. It is now specified as follows: "where Deq is the equivalent diameter, defined as the diameter of a circle having the same projected area as the image of the particle."

Comment 7:
Line 223-224, Figure 8 and its caption. The name “Extent C” has never been used before. Please use “Extent ratio (EC)” in the caption and Extent ratio in the axis label.

Response 7:
We have updated the term to "Extent ratio (E_C)" in both the caption and the axis label of Figure 7 (initially Figure 8).

Comment 8:
Line 230, caption of Figure 9. “…Ellipse ratio …”-> “Ellipse ratio (Er)”

Response 8:
We have updated the term to "Ellipse ratio (E_r)" in both the caption and the axis label of Figure 8 (initially Figure 9).

Comment 9:
Line 232, caption of Figure 19. “…Roundness index …”-> “Roundness index (Rn)”

Response 9:
We have updated the term to "Roundness index (R_n)" in both the caption and the axis label of Figure 9 (initially Figure 10).

Comment 10:
Lines 244-247. This paragraph is unclear and should be rewritten.

Response 10:
We have rewritten the paragraph to provide more details on the role of the convexity index and extent ratio, enhancing its clarity and coherence.

Comment 11:
Lines 249. The term “original methodology” is confusing. I suggest referring to a “Convexity threshold approach” and an “Extent ratio threshold approach”.

Response 11:
We have replaced the term "original methodology" with "Convexity threshold approach" to improve clarity.

Comment 12:
Line 250 and following. The sentence “captured without using the convexity index” is very confusing. Perhaps it is not necessary to understand Figure 11. Instead, a clue as to why a certain threshold value is adopted should be given somewhere.

Response 12:
We have revised the paragraph to enhance its clarity. The modifications include a detailed explanation of the "corrected threshold" values of 0.955 and 0.95. These values were determined through an iterative trial-and-error process, aiming to maximize the consistency in particle categorization between Clemex and ImageJ.

Comment 13:
Line 259-260. Regarding the sentence, “An ellipse ratio with a threshold of 0.90 provided the best results across all powders, as shown in Figure 12.” It is not clear what has been evaluated to draw this best threshold value. Please clarify.

Response 13:
To address your comment, we have added a sentence to clarify how the threshold value of 0.90 for the ellipse ratio was determined. We explained that this threshold was established using an iterative approach, similar to the method used for the extent ratio

Comment 14:
Lines 262-164. Concerning the sentence, “This consistency underscores the utility of the ellipse ratio as a reliable metric for distinguishing elongated particle shapes across various types of metallic powders, regardless of the imaging system employed.” Such a conclusion would have required a comparison with a method other than image analysis to evaluate particle elongation, which is not reported here.

Response 14:
We have revised the paragraph to address your concerns and provide additional context. The updated text now references ISO 9276, highlighting the ellipse ratio as a robust parameter within the framework of image analysis. We have also refined the language to acknowledge potential variability across different imaging systems and the need for further validation in diverse analytical contexts.

Comment 15:
Line 276. The formula sR/rE shows symbols that were not defined previously. I suggest introducing this ratio in section 2.4. I understand that rE is Er in eq 2, so the symbols should be unified. Instead, rS has never been defined as it should.

Response 15:
Thank you for pointing out the inconsistency. You are correct that ( rS ) was not previously defined. We have aligned the text with the definition provided in Section 2.1 (initially Section 2.4) to ensure clarity and consistency.

Comment 16:
Line 305. A facetted surface could hardly be described as rough. Please verify the meaning.

Response 16:
We have revised the manuscript to clarify that the separation process using the automatic bridge removal tool in Clemex can create an artificial aspect on the surface, which influences the irregularity index.

Comment 17:
Lines 308-309. An unexpected carriage return appears in the line.

Response 17:
The line break was removed to adjust the formatting.

Comment 18:
Line 417, Table 3. I suggest adding a column in the table repeating the characteristics that the threshold values reported in the other line should highlight: The presence of satellites, elongated particles, Poligonal particles, and other non-spherical particles. If I understand well, the criterion with Rn&IR has been used to select spherical particles. In that case, the threshold definition should be >0.90, 0.89, 0.93, respectively.

Response 18:
We have decided to remove Table 3, as all relevant threshold values and equations are already detailed within the text. To enhance clarity, we have updated Figure 1 (previously Figure 2) to include the thresholds from the sequential methodology outlined in our previous study.

Comment 19:
I think, however, that, at least in the future direction, a mention of the potential of using AI tools to classify powder shapes would be advisable.

Response 19:
Thank you for your suggestion regarding the potential of using AI tools for classifying powder shapes. In response, we have expanded the "Future Directions" section of the manuscript to include a discussion on the potential integration of AI-based systems.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

We have a few more suggestions for clarification & improvement:

• Line 137: Be sure to keep subscripts for variables consistent (Rn)
• Repeated from first version – Section 2.2: Will all particles be identified with one of these descriptions or will some be left over and labelled as "unknown"? What if a particle has more than nine satellites or more than nine facets, etc.?
• Line 153-155: SS316L (15-45 µm) is mentioned twice. Is there a difference between them? Is this a typo? It is later mentioned as “recycled” and “as-received” powder. Maybe mention this when first introducing the tests materials.
• Table 1 and Table 2: Table 1 figures are clearer than Table 2 figures. The faceted figures are in a different order in Table 2 than in Table 1…it is recommended that they remain consistent.
• Figure 14: Figure and caption should be centered.

If the authors address these points adequately, it should not be necessary to re-review.

Author Response

Comment 1:
Line 137: Be sure to keep subscripts for variables consistent (Rn)

Response 1:
We have reviewed all instances of variables with subscripts throughout the manuscript.

Comment 2:
Repeated from first version – Section 2.2: Will all particles be identified with one of these descriptions or will some be left over and labelled as "unknown"? What if a particle has more than nine satellites or more than nine facets, etc.?

Response 2:

Thank you for your insightful comment. Section 2.2 focuses on the measurements and descriptors used to assess discrepancies rather than the sequential methodology itself. To directly address your concern, we have added the following sentence to the end of the last paragraph in Section 2.1: “Through this comprehensive approach, all particles are accurately labeled according to their morphology, ensuring that any remaining powder after applying the sequential methodology is classified as spherical.” This addition clarifies that our methodology accounts for all particles and categorizes them appropriately, with no particles left as "unknown."

Comment 3:
Line 153-155: SS316L (15-45 µm) is mentioned twice. Is there a difference between them? Is this a typo? It is later mentioned as “recycled” and “as-received” powder. Maybe mention this when first introducing the tests materials.

Response 3:
We have revised the text in lines 153-155 to clearly differentiate between the "as-received" and "recycled" 316L stainless steel powders.

Comment 4:
Table 1 and Table 2: Table 1 figures are clearer than Table 2 figures. The faceted figures are in a different order in Table 2 than in Table 1…it is recommended that they remain consistent.

Response 4:
We have replaced the images in Table 2 with those used in Table 1 to ensure clarity and consistency across both tables.

Comment 5:
Figure 14: Figure and caption should be centered.

Response 5:
Thank you for your suggestion. The current alignment of figures and captions to the left follows the formatting guidelines provided by the Powders template. If there are any changes to the template requirements, we will be happy to adjust the alignment accordingly.

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have adequately addressed the points raised previously.

Author Response

Comment: The authors have adequately addressed the points raised previously.
Response: Thank you for your positive feedback and for acknowledging that we have adequately addressed the points raised in our previous revision. We appreciate your time and effort in reviewing our manuscript.