Shoulder Implant Manufacturer Detection by Using Deep Learning: Proposed Channel Selection Layer

Round 1

Reviewer 1 Report

It is a very interesting study using AI methods to clarify shoulder implants.

However, it is wondering that the number of dataset are sufficient for making deep learning model and training / testing even though data augmentation were used.

Please compare the deep learning method and machine learning algorithms in detail in the section of the discussion.

Please describe how the proposed model performed much better than the other studies (not in the same research environments).

I would like to thank you for your valuable comments and contributions that helped to significantly increase the quality, readability and organization of the study.

I have completed the revisions which you have reported as follows.

REVISIONS

• It is a very interesting study using AI methods to clarify shoulder implants.
• However, it is wondering that the number of dataset are sufficient for making deep learning model and training / testing even though data augmentation were used.

The following answer has been added to Section 2.1.:

In the study, manufacturer and model information of implants have been obtained from the University of Washington Shoulder website [19]. Commonly, shoulder implants are classified into different generations. First-generation implants like Cofield are mon-oblock and non-modular. Depuy is classified in the second-generation implant class with its modular. Third-generation implants like Tornier and Zimmer are inclination and modular with adaptable variable. Three implant models by manufactured Depuy were used in the study. There is no space between the head and body of the HRP model implant which is monoblock design with a dorsal three holed fin. The Global model is modular. In this implant model a space between upper and rim of stem and head is designed. The Global Advantage implant model is also modular. The difference of the global model from Global Advantage implant model is that it has a more pronounced front edge with three holes. The Tornier manufacturer's implants are distinguished by two holes and a distinctive triangular dorsal fin with a medialized eccentric head. Co-field implants have a pointed distal stem and more space under the head. It differs from other manufacturers with these features. The Zimmer implant model is designed as monoblock with dorsal and ventral fins. Zimmer implants’ stem tip is rounded. And also, they lack a ridge under the head [19]. The number of datasets are sufficient for modelling deep learning architectures owing to model label support and also data augmentation applied.  

• Please compare the deep learning method and machine learning algorithms in detail in the section of the discussion.

The omparison about the deep learning method and machine learning algorithms has been added to section of the discussion:

The success of machine learning algorithm depends on the choice of appropriate fea-tures. Various preprocesses, size reduction, feature selection etc. are performed to reveal these features. It is necessary to get rid of the dependence on features to increase the accuracy and reduce the transaction cost in these processes. In the study, deep learning architectures have a higher accuracy rate and a more efficient system owing to more successful, easier and lower cost extraction of valuable features from the data.

• Please describe how the proposed model performed much better than the other studies (not in the same research environments).

The following sentences have been added to section of the conclusions:

The average score and standard deviation value of the channels are important for the channel selection formula. The current score for each feature is evaluated again with these values. In this way, it can be calculated which feature will be the most prominent and dominant among the features. This is the reason why the accuracy rate increases with a multi-channel network model. The accuracy rate has been increased significantly by using the most effective feature selection which was applied among the channels for each image.

Reviewer 2 Report

This manuscript provides results of a systematic study investigatingUsing Deep Learningon application  of Shoulder Implant  Manufacturer  Detection. The  experimental  results  can  be used as a reference for the combination of computer science and medicine.While the research quality is high, a number of improvements are needed to improve the manuscript:

1.What  is  the Specific Difference  between  the  products  of  different  manufacturers, andit should be clearly mentioned in the text. That's one of yourkey innovations.

The following sentences have been added to Section 2.1:

In the study, manufacturer and model information of implants have been obtained from the University of Washington Shoulder website [19]. Commonly, shoulder implants are classified into different generations. First-generation implants like Cofield are mon-oblock and non-modular. Depuy is classified in the second-generation implant class with its modular. Third-generation implants like Tornier and Zimmer are inclination and modular with adaptable variable. Three implant models by manufactured Depuy were used in the study. There is no space between the head and body of the HRP model implant which is monoblock design with a dorsal three holed fin. The Global model is modular. In this implant model a space between upper and rim of stem and head is designed. The Global Advantage implant model is also modular. The difference of the global model from Global Advantage implant model is that it has a more pronounced front edge with three holes. The Tornier manufacturer's implants are distinguished by two holes and a distinctive triangular dorsal fin with a medialized eccentric head. Co-field implants have a pointed distal stem and more space under the head. It differs from other manufacturers with these features. The Zimmer implant model is designed as monoblock with dorsal and ventral fins. Zimmer implants’ stem tip is rounded. And also, they lack a ridge under the head [19]. The number of datasets are sufficient for modelling deep learning architectures owing to model label support and also data augmentation applied.  

2. What are the advantages of your job compared to Urban's? Is it just accuracy? How does it work?

The following part have been added to Introduction section:

Urban et al. [6] have classified shoulder implants according to manufacturer using the CNN-based deep learning method. The study was aimed to classify implants by manufacturers only. In addition, the pre-trained models were also used for modelling. The classic CNN architecture was applied in the model. Shoulder Implant X-Ray dataset [7] which is available in the UCI open access repository was utilized in their study. In this study, the dataset which was used by Urban et al. [6] has been applied. It has been considered that a better classification result would be obtained by proposing a novel deep learning architecture. In addition, it has been studied to classify the shoulder im-plant as a manufacturer and model in the study. In our study, shoulder implants were classified by both manufacturer and model for extend to extend Urban et al. [6]'s study.

3. It  isstrongly  suggestedthat, on  the  basis  of  summarizing  the  previous  work,the advancement of currentresearchput forwardin the first part.

The summarizing of the previous work was added in introduction section.

4. In order to enhance the readability of the article, specific examples should be attached in Section 2.3, instead of simply displaying the formulas

Figure 1. has been added to section 2.3 in order to enhance the readability of the article.

5. The conclusion part, "In this study, proposed model is presented with the novel channel selection formula which is for selecting the most prominent feature filters unlike in the literature. " Please add further evidence to support this conclusion.

The following part have been added to Conclusion section:

The average score and standard deviation value of the channels are important for the channel selection formula. The current score for each feature is evaluated again with these values. In this way, it can be calculated which feature will be the most prominent and dominant among the features. This is the reason why the accuracy rate increases with a multi-channel network model. The accuracy rate has been increased significantly by using the most effective feature selection which was applied among the channels for each image.

6. References 6~7 are in contradiction with what is written in the paper

References 6~7 are revised in the related works section. (L86-L97)

7. L69: mo-del.

It is revised

8. Figure 1 does not show the frequency, more like a specific quantity, please correct.

Figure 1 has been corrected.

9. References 12~17, the author's first and last name is wrong, and does not correspond to the manuscript.

The references were revised.

Round 2

Reviewer 2 Report

This manuscript is recommended to be accepted.