Evaluation of Integrated CNN, Transfer Learning, and BN with Thermography for Breast Cancer Detection

Round 1

Reviewer 1 Report

The work is interesting. The research significance of this work lies in the implementation and integration of a number of different research techniques for obtaining a diagnosis from breast thermograms from several data sources.I believe that the article sent to me for review is a good one, it is clearly written and does not require any changes or additions. I believe it will find wide interest among readers and will be widely cited as it raises very important medical issues. Few publications exist on this topic. The conclusions are correct. The bibliography list is also correct. I see no need to make any changes or additions to the content of the article. The only thing I have not done is to check the "quality" of the English language.

Author Response

To respected Reviewer # 1: Thank you for the comment. The paper has been proofread for a few rounds. We would like to thank you for your thoughtful comments and efforts towards improving our manuscript. Your comments provided valuable insights to refine its contents and analysis. Track changes are used in the revised manuscript to facilitate your review process.

Reviewer 2 Report

Journal Applied Sciences (ISSN 2076-3417)

Manuscript ID applsci-2094454

Type Article

Title Evaluation of Integrated CNN, Transfer Learning and BN with Thermography for Breast Cancer Detection

 

The idea of using thermography as a mass screening at short time intervals and self- screening method is interesting.

 

1.       Do the authors think that 306 thermal images were enough for a conclusive study? How many images would be statistically recommended? Do they consult a statistical?

 

2.       Add one scientific recognized reference to endorse this affirmative of the authors: “One adverse effect of this technique is that it is a harmful radiation which may trigger tumor growth”

3.       Add one scientifically recognized reference to endorse this affirmative of the authors “However, the expertise of the medical professional doing the test determines whether ultrasonography diagnosis will be successful (note: its accuracy is a function of volume to mass ratio).”

4.       Add one scientifically recognized reference to endorse this affirmative of the authors: “The most reliable and accurate way to diagnose a tumor or breast cancer is via MRI scans.” The most reliable means and accurate is yet mammography, considered the gold standard. The MRI still generates many false positive results.

5.       Add one scientifically recognized reference to endorse this affirmative of the authors: “For the overall evaluation of breast lesions, SPECT imaging has a higher diagnostic value than mammography, has been extensively validated, and has a high sensitivity.”

6.       Are the authors claiming that mammography is ineffective in this sentence? Add an adequate reference that confirms this statement by the authors: “For routine mass screening at frequent intervals, the current diagnostic techniques are ineffective.”

7.       This phrase is confusing, review: “Additionally, they are not suitable for regular breast self-examination (BSE), which is recommended by the WHO to minimize the number of fatalities from breast cancer, because people lack the necessary equipment at home”. BSE is a clinical examination in which the patient herself performs palpation. There is no consensus that BSE minimizes the number of fatalities, as a palpable tumor is not an early tumor. Check which bibliographic reference the WHO used for this statement and revise the sentence.

8.       This phrase is not true, review: “Thermography could detect any abnormality in the body, such as a tumor [7].” There is no diagnostic test that can detect any abnormality in the body. Otherwise, thermography would be an exam that would replace all exams in the world.

9.       Review the word “appropriate” in this phrase: “This method is appropriate for detecting cancers in their preliminary stages since an indication of tumor development is an increase in the temperature of the breast tissue.” It is stating that all tumor development is an increase in the temperature of the breast tissue. All, are you sure? Studies are showing false negative results with thermography and cold tumors. Tumors can be malignant or benign, including cysts, so this sentence should be revised.

10.   This sentence is part of the method and not in the introduction, withdrawal, or rewrite: “The current work focuses on the development of CNN based on a multi-source database without preprocessing for binary classification. Also, the study compares Transfer learning methods with the baseline CNN model to develop an intelligent tool for breast cancer detection.”

11.   In the paragraph of lines 134-145, the authors are affirming the contributions of the study and the novelty without having presented the results beforehand. Therefore, revise and rewrite or replace this text in the discussion part. Put it as a hypothesis of the study without anticipating the results.

12.   Do not write the text in the third person “We”, be impersonal in all text.

13.   Inform the source (reference or website): “We use ImageNet, a massive dataset with 1.2 million…”

14.   Inform the source (reference or website): Very Deep Convolutional Networks (VGG16)

15.   This sentence is without scientific support, what was the objective criterion of choice? Leave most clear this sentence: “This model was amongst the primary choices because it was well-developed and produced outstanding results in various classification cases for medical imagery”

16.   The images in Figure 1. They are at low resolution. Improve.

17.   How many cases of women with prostheses were evaluated in your study? Radiotherapy? Plastic surgery? Biopsy, hormone replacement?

18.   How many times was the algorithm tested within the same sample until the results were achieved?

19.   Was the algorithm tested on a separate sample that was not included in its construction? I didn´t see it. If yes, what was the result of sensitivity, specificity, and accuracy?

20.   Regarding the 266 thermal images from the Database for Mastology Research answer:

a.       Under what conditions of ambient temperature and humidity were they obtained?

b.      How long did thermalization take?

c.       What was the thermal sensor model (inform the technical specifications)?

d.      What was the average age of the patients?

e.      What were the mean size and standard deviation of the nodules? How many were confirmed as cancer? What types of cancers were in this group?

21.   Regarding the 40 thermal images from the Multifunctional Medical Center of Astana, answer:

a.       In what conditions of ambient temperature and humidity they were obtained?

b.      How long did thermalization take?

c.       What was the average age of the patients (18 to 80 is not enough information)?

d.      What were the mean size and standard deviation of the nodules? How many were confirmed as cancer? What types of cancers were in this group?

22.   The word “sick ” does not seem appropriate, as it encompasses any breast alteration, not just malignant tumors, which is the main scope of this work. Also, I don´t know if are palpable nodules, or early tumors (non-palpable), no information about it. The "sick" term encompasses several benign diseases too. Therefore, the study is not specific for early diagnosis of cancer if these items are not specified. How were patients with benign fibrocystic breast disease classified in the healthy or "sick" group? How many cases did you have with this condition so frequently in young women?

23.   The FLUKE TiS60 + is an IR camera with technical specifications aimed at industrial use and with sensors calibrated to measure up to 400 °C. Do you have any certificates of its calibration specifying its application for skin temperature in humans? Are image noise and accuracy of ± 2 °C or 2 % factors that could interfere with the final image quality and the analysis of patterns used? Do you have any literature references about this? Is it considered a medical device certified by any health regulatory institution?

24.   Did the thermal imaging from the Multifunctional Medical Center of Astana collect using the same method and technique/apparatus as those from the Database for Mastology Research?

25.   Due to the small number of images, would it not be better to analyze only the images from the Database for Mastology Research? Did the authors verify whether the results of the algorithms were the same regardless of the source of the images?

26.   Figure 4. The images did not follow the same framing. Some show the patient's mouth, other parts of the clothing, almost entire parts of the abdomen, and others with a reference marker on the abdomen. This can be mentioned in the discussion as a limitation of the study and that can be improved in the next image collections. It is expected that the greater the framing variation and artifacts of the images, the lower the accuracy of the algorithm. The authors cited that the images were resized. Explain how this was done and if the aforementioned artifacts were corrected.

27.   Don't worry. Any of the above items that cannot be answered, be put in the discussion as limitations of the study.

28.   It would be interesting to explain to the reader what these parameters mean, in the same way, that was explained the AI methods used in the text: “Batch size = 32; Optimizer =" adam "; epochs = 25.”

29.   Explain in the text the importance of the Characteristics of the machine (Table 3) used in the study and which of these parameters is most important.

30.   Improving image quality Figure 6 to 12. Increasing resolution.

31.   The authors mentioned that Biopsy is among the most crucial factors for the diagnosis decision. Then report the results of the biopsy, I didn´t have any idea of which kind of tumors the authors studied. Any in situ tumor? More or less than 1 cm?

32.   Specify better what diagnosis the authors are referring to, healthy and "sick" only. “This means that we can safely use the BN constructed expert models to understand which factors play a key role to the final successful diagnosis (something that is not possible with the pure CNN system).”

33.   The paragraph conclusion begins with the objective of the study. Correct or withdraw, this is the wrong part that is just the conclusion.

34.   It is not possible to state: “… assist doctors in reliably and quickly distinguishing cancerous breasts from non-cancerous ones”. It was only healthy and "sick". And it did not specify the sick group in detail.

35.   This sentence is not a conclusion but a quotation from literature, it may be in the introduction or discussion, but it appears to be an author´s conclusion. Withdraw or review: “Thus, we can claim and conclude that the crucial which drive the decision in the BN models comprise a trusted interpretable diagnosis model which is useful both for a patient and a physician [33],[34].” Again the authors are using the phrase “us” in the third person, which confuses the reader into thinking that this is the conclusion of this study.

36.   This sentence is an assumption, the study did not later validate the algorithm with a group of medical doctors and also did not carry out an economic feasibility analysis comparing the values of the entire process of traditional methods: “The developed integrated system can be easily implemented on low-cost portable devices for automatic mass screening and breast self-examination (BSE) as envisaged by WHO to eradicate breast cancer ”

 

Congratulations on the study, it is very important and can clarify many doubts and be a source of inspiration for new researchers. Breast cancer is a disease that is still poorly understood and needs better methods of early diagnosis.

Comments for author File: Comments.pdf

Author Response

To respected Reviewer # 2:

please see separate file attached.

In all, we would like to thank you for your thoughtful comments and efforts towards improving our manuscript. Your comments provided valuable insights to refine its contents and analysis. Track changes are used in the revised manuscript to facilitate your review process.

Author Response File: Author Response.pdf