The Fusion of Focused Spectral and Image Texture Features: A New Exploration of the Nondestructive Detection of Degeneration Degree in Pleurotus geesteranus

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study is interesting. However, please confirm the different of this article with the previous published work from the same authors (https://doi.org/10.3389/fpls.2023.1260625). The two articles seems similiar. Please state the original contribution of this article clearly.

1. Abstract. Please report also other figures of merit: sensitivity, specificity, and error rate of the classification results.
2. In the introduction, please state the different of this article with previous published work from the same authors (https://doi.org/10.3389/fpls.2023.1260625). Please cite that published work in this article. It is very important to show us your original contribution in this article.
3. Sample preparation is not clear especially, there is no any image for class 0, class 1, etc. It is important to show us the visual RGB image of the sample with different type of degradation.
4. Spectral acquisition is not clear. Please show one additional image showing the spectral acquisition using hyperspectral imaging system.
5. How the samples separated into training and testing samples is not clear. Please state it and how did the authors divide the samples into training and testing? What type of validation method used? 
6. Evaluation indicators should be expanded: sensitivity, specificity, and error rate.
7. Figure 5 is missing label x and y. The label y is accuracy (%) and the label x is feature selection.
8. Table 2 should be revised and expanded. For each data type (spectrum, image, and combination) should be subjected to same classification model (None-CNN, CARS-CNN, LBP-CNN, and combination). Show all the results in Table 2.
9. The explanation of Table 2 result should be expanded. It make sense that the combination data type with the combination classification model resulted the best accuracy. I did not see any interesting result based on Table 2.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The article “Fusion of Focused Spectrum and Image Texture: A New Exploration of Nondestructive Detection of Degeneration Degree of Pleurotus geesteranus” describes a method that uses hyperspectral data and machine learning to detect strain degradation in Pleurotus fungi.

Overall, the paper is very interesting and presents a modern approach. However, there are several points that need to be examined. More specifically:

• In the introduction, consider adding a paragraph about simpler non-destructive methods in the agrifood industry, such as conventional machine vision with RGB image data or other statistical techniques. It would also be helpful to highlight the motivation for exploring more complex methods for precise detection, before analyzing the related work using hyperspectral data. The following references might be helpful:
  • Ismail, N., & Malik, O. A. (2022). Real-time visual inspection system for grading fruits using computer vision and deep learning techniques. Information Processing in Agriculture, 9(1), 24-37.
  • Kondoyanni, M.; Loukatos, D.; Templalexis, C.; Lentzou, D.; Xanthopoulos, G.; Arvanitis, K.G. Computer Vision in Monitoring Fruit Browning: Neural Networks vs. Stochastic Modelling. Sensors2025, 25, 2482. https://doi.org/10.3390/s25082482
  • Palumbo, M., Cefola, M., Pace, B., Attolico, G., & Colelli, G. (2023). Computer vision system based on conventional imaging for non-destructively evaluating quality attributes in fresh and packaged fruit and vegetables. Postharvest Biology and Technology, 200, 112332.
• Apart from the references that refer to applications using the pre-processing techniques (SG, MSC, SNV), the spectral analysis techniques (SPA, CARS, PCA) and the classification techniques (SVM, KNN, CNN), it is preferred to also provide references that refer exclusively to the methodologies themselves and the relevant available algorithms, either in the References section or as a footnote.
• In the methodology section, consider describing the rationale behind the specific settings used for data acquisition (i.e., whether the settings were the default configurations of the instruments, selected empirically, or based on a standard protocol for acquiring this type of data).
• In addition to the flowchart and final table, consider providing a consolidated table that presents all the combinations of preprocessing techniques, analyses and classification methods – for example, which techniques were combined with others and their results. Although they are described in the text, the sequence of analyses appears somewhat vague, and such a table would improve clarity for the reader.
• Provide photos of the set up/ the experiment/ the samples/ a minimal screenshot of part of the code to make the overall process more understandable for the readers.
• When describing the processes, such as the pre-processing techniques, it would be helpful to show the products or samples before and after each technique, when that is possible, such as the reflectance curves. Another example would be to present a sample of the raw data and perhaps create a flowchart of that sample showing the changes after each process. The key takeaway from this note is to present representative depictions of your work, both as evidence and as helpful visual examples for peers who want to utilize and quote your work in their research.
• Include some details regarding the analysis environment, such as where the analysis was performed (e.g., type of computer, computer power required, processing time etc.)
• In page 7 where there is the explanation of the formula of accuracy, there is a confusion of the use of “false” and “negative” terms. Better replace the paragraph with the following text:
  “Where TP means that the positive sample is correctly marked as a positive sample; TN indicates that the negative sample is correctly marked as a negative sample; FP means that the negative sample is falsely marked as a positive sample; FN indicates that a positive sample is falsely marked as a negative sample.”
• In page 9 in the following part, it seems like something is missing “The SNV processing then Through the mean and standard deviation correction, the curve is dispersed and the peaks and valleys are obvious,…”
• In the discussion section, address whether this technique could be used in real-time and if it can be run on an edge device, which is important for potential commercial applications (e.g., in edible fungi cultivation).

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have responded very fluently to the suggested changes, thus delivering a manuscript that meets the publication standards of the Agriculture Journal.  

Author Response

We sincerely appreciate your time and expertise in reviewing our manuscript. Your previous comments were invaluable in improving the quality of this work. We are grateful for your continued support throughout the review process.