Evaluating the Classification of Freeze-Dried Slices and Cubes of Red-Fleshed Apple Genotypes Using Image Textures, Color Parameters, and Machine Learning

Round 1

Reviewer 1 Report

see the attachment 

Comments for author File: Comments.pdf

Author Response

The author has proposed “Classification of freeze-dried slices and cubes of red-fleshed apple genotypes evaluated using image textures, color parameters and machine learning”, it is an interesting topic, however, I have following comments.

Response: Thank you very much for your careful reading of the manuscript and all the comments. Your comments have been considered and the manuscript has been significantly improved.

 

Comment: I can say that author has just focused on data rather than machine learning models.

Response: The main objective of the study was to reveal the usefulness of image textures and color parameters for the classification of freeze-dried slices and cubes of red-fleshed apples belonging to different genotypes. Therefore, most attention has been paid to the parameters. However, the machine learning models have been described in more detail in the revised version of the manuscript as follows:

“Firstly, the classification was carried out using texture parameters selected from all color channels of freeze-dried slice images of ‘Alex Red’, ‘Trinity’, ‘314’, and ‘602’ red-fleshed apples. Then, also color parameters were added to a set of selected image textures, and the analysis was performed for a set of selected image textures and color features of apple slices. Afterward, freeze-dried cubes of ‘Alex Red’, ‘Trinity’, ‘314’, and ‘602’ apples were classified using models involving attributes selected from a set of textures extracted from images in all color channels. In the next step of the analysis, classification models were built using a combined set of selected image textures and color parameters of apple cubes. The attribute selection was performed using the Best First and CFS (Correlation-based Feature Selection) subset evaluator. This step was performed separately for a set of image textures of apple slices, a combined set of image textures and color parameters of apple slices, a set of image textures of apple cubes, and a combined set of image textures and color parameters of apple cubes.”

“Selected features were used for building classification models using machine learning algorithms from the groups of Functions, Lazy, Meta, Trees, and Bayes. In the case of each group, the most successful algorithm was chosen. The LDA (Linear Discriminant Analysis) from the group of Functions), IBk (Instance-Based k) from Lazy, LogitBoost from Meta, LMT (Logistic Model Tree) from Trees, and Bayes Net from Bayes were chosen. IBk is a k-nearest-neighbors classifier. LogitBoost performs additive logistic regression. LMT builds logistic model trees. The function of Bayes Net is learning Bayesian nets [33-35]. The criterion for the selection of the algorithms was the highest overall accuracy. Besides overall accuracies, the number of correctly and incorrectly classified cases, and the values of the Kappa statistic, Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), TP (True Positive) Rate, FP (False Positive) Rate, Precision, F-Measure, MCC (Matthews Correlation Coefficient), ROC (Receiver Operating Characteristic) Area, and PRC (Precision-Recall) Area were determined [21, 36-37].”

 

Comment: The dataset is relatively very small, there are 7,500 apple types and author has decided to just choose 4? Author must increase the dataset class size.

Response: We agree with the Reviewer that there are about 7,500 varieties of apples in the world. However, we considered red-fleshed apples, of which there are few. The use of red-fleshed apples is one of the innovative parts of the study. 4 varieties were selected from several grown at the Experimental Orchard of the National Institute of Horticultural Research in Dabrowice near Skierniewice. These varieties yielded the best fruit. In addition, they were indicated as the most useful for drying. For other varieties, sufficient samples were not available to perform the experiments. However, further studies are planned that may include more apple types, other drying techniques, deep learning models in addition to the traditional machine learning algorithms, and also the estimation of chemical properties of dried red-fleshed apple samples based on image textures and color parameters. The tests carried out on four varieties showed the usefulness of the developed procedures and indicated the possibility of performing more in-depth research.

 

Comment: Using machine learning algorithms is old trend and not going to be helpful at all in terms of usage, better to use deep learning.

Response: In the present study traditional machine learning algorithms proved to be useful to distinguish samples with very satisfactory accuracies and low errors. Their usefulness has been confirmed. However, future studies are planned as indicated in the manuscript:

“In further studies, besides the traditional machine learning algorithms, the models could be built using deep learning. Additionally, in future research, image textures and color parameters may be used for the prediction of the changes in the flesh structure of red-fleshed apples caused by various drying techniques and for the non-destructive and objective estimation of the chemical properties of dried samples.” 

 

Comment: There is no practical usage mentioned in manuscript about this work.

Response: The practical usage has been indicated as follows:

“The developed models can be used in practice to distinguish freeze-dried red-fleshed apples in a non-destructive and objective manner. It can avoid mixing samples belonging to different genotypes with different chemical properties.”

“The approach combining image textures, color parameters, and machine learning is a great novelty in classifying dried red-fleshed apples and can be used in practice to avoid mixing different genotypes.”

 

Comment: In abstract Author mentioned they developed classification models but fails to mention which classification models? Did they use deep learning/ machine learning?

Response: It has been indicated as follows:

“The classification models were developed based on a set of selected image textures and a set of combined selected image textures and color parameters of freeze-dried apple slices or cubes using various traditional machine learning algorithms.”

 

Comment: They have focused more on parameters than on data and models.

Response: As mentioned in one of the previous comments, the main objective of the study was to reveal the usefulness of image textures and color parameters for the classification of freeze-dried slices and cubes of red-fleshed apples belonging to different genotypes. Therefore, most attention has been paid to the parameters. However, the dataset and machine learning models have been described in more detail in the revised version of the manuscript as follows:

“the dataset included:

- one hundred freeze-dried slices of the ‘Alex Red’ red-fleshed apple genotype,

- one hundred freeze-dried slices of the ‘Trinity’ red-fleshed apple genotype,

- one hundred freeze-dried slices of the ‘314’ red-fleshed apple genotype,

- one hundred freeze-dried slices of the ‘602’ red-fleshed apple genotype,

and

- one hundred freeze-dried cubes of the ‘Alex Red’ red-fleshed apple genotype,

- one hundred freeze-dried cubes of the ‘Trinity’ red-fleshed apple genotype,

- one hundred freeze-dried cubes of the ‘314’ red-fleshed apple genotype,

- one hundred freeze-dried cubes of the ‘602’ red-fleshed apple genotype.”

 “Firstly, the classification was carried out using texture parameters selected from all color channels of freeze-dried slice images of ‘Alex Red’, ‘Trinity’, ‘314’, and ‘602’ red-fleshed apples. Then, also color parameters were added to a set of selected image textures, and the analysis was performed for a set of selected image textures and color features of apple slices. Afterward, freeze-dried cubes of ‘Alex Red’, ‘Trinity’, ‘314’, and ‘602’ apples were classified using models involving attributes selected from a set of textures extracted from images in all color channels. In the next step of the analysis, classification models were built using a combined set of selected image textures and color parameters of apple cubes. The attribute selection was performed using the Best First and CFS (Correlation-based Feature Selection) subset evaluator. This step was performed separately for a set of image textures of apple slices, a combined set of image textures and color parameters of apple slices, a set of image textures of apple cubes, and a combined set of image textures and color parameters of apple cubes.”

“Selected features were used for building classification models using machine learning algorithms from the groups of Functions, Lazy, Meta, Trees, and Bayes. In the case of each group, the most successful algorithm was chosen. The LDA (Linear Discriminant Analysis) from the group of Functions), IBk (Instance-Based k) from Lazy, LogitBoost from Meta, LMT (Logistic Model Tree) from Trees, and Bayes Net from Bayes were chosen. IBk is a k-nearest-neighbors classifier. LogitBoost performs additive logistic regression. LMT builds logistic model trees. The function of Bayes Net is learning Bayesian nets [33-35]. The criterion for the selection of the algorithms was the highest overall accuracy. Besides overall accuracies, the number of correctly and incorrectly classified cases, and the values of the Kappa statistic, Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), TP (True Positive) Rate, FP (False Positive) Rate, Precision, F-Measure, MCC (Matthews Correlation Coefficient), ROC (Receiver Operating Characteristic) Area, and PRC (Preci-sion-Recall) Area were determined [21, 36-37].”

 

Comment: Mention the dataset size and other details briefly in abstract.

Response: It has been added as follows:

“One hundred apple slices with a thickness of 4 mm and one hundred cubes with dimensions of 1.5 cm x 1.5 cm x 1.5 cm of each genotype were subjected to freeze-drying.”

“Apple samples were at the stage of harvest maturity.”

“The average fruit weight, starch index, internal ethylene concentration, flesh firmness, total soluble sugar content, and titratable acidity were determined.”

“For each apple sample (slice or cube), 2172 image texture parameters were extracted from images in 12 color channels and color parameters L*, a*, and b* were determined.”

 

Comment: The abbreviation such as (R, G, B, L, a, b, X, Y, Z, U, V, and S) are decreasing the readability, either remove such abbreviations or write something in general reflecting the same.

Response: It has been replaced by:

“For each apple sample (slice or cube), 2172 image texture parameters were extracted from images in 12 color channels.”

“Models built based on selected textures of slice images in 11 selected color channels correctly classified freeze-dried red-fleshed apple genotypes”

“The classification of apple cube images using models including selected texture parameters from images in 11 selected color channels was characterized by an overall accuracy of up to 74.74%”

“Firstly, the classification was carried out using texture parameters selected from all color channels of freeze-dried slice images of ‘Alex Red’, ‘Trinity’, ‘314’, and ‘602’ red-fleshed apples.”

“Afterward, freeze-dried cubes of ‘Alex Red’, ‘Trinity’, ‘314’, and ‘602’ apples were classified using models involving attributes selected from a set of textures extracted from images in all color channels.”

“Models developed based on selected textures from cubes’ images in selected color channels and color parameters (L*(D65), a*(D65), b*(D65)) were characterized by an overall accuracy of 71.75% (IBk) to 80.50% (LogitBoost)”

 

Comment: mention the equal error rate in abstract.

Response: It has been mentioned as follows:

“Models built based on selected textures of slice images in 11 selected color channels correctly classified freeze-dried red-fleshed apple genotypes with an overall accuracy reaching 90.25% and the Mean Absolute Error of 0.0545 and by adding selected color parameters (L*, b*) to models, an increase in the overall accuracy to 91.25% and a decrease in the Mean Absolute Error to 0.0486 were observed.”

 

Comment: The literature is not sufficient to cover the said area. Many latest references have not been mentioned in the literature. The author should include the latest literature in the manuscript and highlight their contribution. Some of the studies are as follows:

-Thermal and non-thermal processing of red-fleshed apple: how are (poly) phenol composition and bioavailability affected?

-An Artificial Intelligence-Based Stacked Ensemble Approach for Prediction of Protein Subcellular Localization in Confocal Microscopy Images

-Fruit Image Classification Model Based on MobileNetV2 with Deep Transfer Learning Technique

-Enhancing Image Annotation Technique of Fruit Classification Using a Deep Learning Approach

Response: The Introduction has been corrected as follows:

“Red-fleshed apples with high content of anthocyanins and other phenolic compounds are characterized by a strong antioxidant capacity [5]. They are attractive to consumers due to their red color and positive effect on health [6]. Red-fleshed apples can be consumed in the form of snacks. Drying is a promising process for producing healthy snacks with the retention of bioactive compounds and nutrients. The presence of anthocyanins in the flesh of red-fleshed apples results in the obtaining of value-added snack products [7]. It was reported that ready-to-eat dried snacks can be interesting and attractive food products [8]. Drying is used to preserve food by reducing water content and thus inhibiting microbial growth. Drying at a low temperature and reduced pressure minimizes changes in the dried material and the loss of heat-labile compounds. Freeze-drying is one of the least destructive drying methods in terms of phenolic compounds due to low temperature and scant contact with air [9]. Freeze-drying is non-thermal processing technology. It was found that thermal processing caused greater changes in red-fleshed apple snacks. For example, infrared-drying resulted in great losses in the apple (poly)phenolics. Hot air-drying maintained approximately 83% of the total (poly)phenols compared with the freeze-dried samples, and purée pasteurization only 65%. Whereas the degradation of anthocyanins was higher and hot air-dried apple snacks maintained 26% and pasteurized purée samples only 9% compared with freeze-dried apple snacks [10]. Wojdyło et al. [11] reported that novel red-fleshed apple snacks are a promising high-quality dehydrated product belonging to the functional foods category. The authors [11] revealed that freeze-drying allowed for the retaining of the highest content of bioactive compounds in dried red-fleshed apple snacks, followed by hybrid (convective pre-drying + vacuum-microwave drying), vacuum-microwave, and convection drying.”

“Machine Learning (ML) algorithms aim to automate the learning process through training and then the optimized regression and/or classification algorithms can be utilized to predict accurate estimation of the unknown samples/objects [15]. ML is a key tool for Industry 4.0 where all elements of the manufacturing process can be interconnected through intelligent data analysis [16].”

“Image analysis and machine learning were successfully applied in previous studies to classify fruit and vegetables [17-19] and detect the changes in the product quality as a result of different processing, such as drying and fermentation [20-23]. Furthermore, image features were used to estimate and predict the chemical properties of food products [24,25]. Machine learning techniques can be very effective in predicting chemical substances [26]. In the case of apples, the usefulness of image processing to determine the effect of drying on shrinkage, color, and image texture of slices was proved. The correlation was feasible between drying time and the slice area, perimeter, and other morphological features of the image. Moreover, a*, and b* values showed a stable increase with the drying time [27]. While previous studies presented an understanding of how to correlate image texture with the quality of the dried apple, no study investigated the possibility of using digital imaging for evaluating the quality of freeze-dried red-fleshed apples.”

 

5. Wang, X.; Li, C.; Liang, D.; Zou, Y.; Li, P.; Ma, F. Phenolic compounds and antioxidant activity in red-fleshed apples. Journal of Functional Foods 2015, 18, 1086-1094.
6. Juhart, J.; Medic, A.; Veberic, R.; Hudina, M.; Jakopic, J.; Stampar, F. Phytochemical Composition of Red-Fleshed Apple Cul-tivar ‘Baya Marisa’ Compared to Traditional, White-Fleshed Apple Cultivar ‘Golden Delicious’. Horticulturae 2022, 8, 811.
7. Joshi, A.P.K.; Rupasinghe, H.P.V.; Khanizadeh, S. Impact of drying processes on bioactive phenolics, vitamin c and antioxidant capacity of red-fleshed apple slices. J. Food Process. Pres. 2011, 35, 453-457.
8. Konopacka, D.; Seroczynska, A.; Korzeniewska, A.; Jesionkowska, K.; Niemirowicz-Szczytt, K.; Płocharski, W. Studies on the usefulness of Cucurbita maxima for the production of ready-to-eat dried vegetable snacks with a high carotenoid content. LWT - Food Science and Technology 2010, 43(2), 302–309.
9. Kidoń, M.; Grabowska, J.; Bioactive compounds, antioxidant activity, and sensory qualities of red-fleshed apples dried by different methods. LWT - Food Science and Technology 2021, 136, 2, 110302.
10. Yuste, S.; Macià, A.; Motilva, M.J.; Prieto-Diez, N.; Rubió, L. Thermal and non-thermal processing of red-fleshed apple: How are (poly)phenol composition and bioavailability affected? Food Funct. 2020, 11, 10436.
11. Wojdyło, A.; Lech, K.; Nowicka, P. Effects of Different Drying Methods on the Retention of Bioactive Compounds, On-Line Antioxidant Capacity and Color of the Novel Snack from Red-Fleshed Apples. Molecules 2020, 25, 5521.
12. Alpaydin, E. Introduction to machine learning: MIT press 2020.
13. Sarker, I.H. Machine learning: Algorithms, real-world applications and research directions. SN Computer Science 2021, 2(3), 1-21.
14. Gulzar, Y. Fruit Image Classification Model Based on MobileNetV2 with Deep Transfer Learning Technique. Sustainability 2023, 15, 1906.
15. Mamat, N.; Othman, M.F.; Abdulghafor, R.; Alwan, A.A.; Gulzar, Y. Enhancing Image Annotation Technique of Fruit Classification Using a Deep Learning Approach. Sustainability 2023, 15, 901.
16. Suryawanshi, Y., Patil, K.; Chumchu, P. VegNet: Dataset of vegetable quality images for machine learning applications. Data in Brief 2022, 45, 108657.
17. Ropelewska, E.; Sabanci, K.; Aslan, M.F. The Use of Digital Color Imaging and Machine Learning for the Evaluation of the Effects of Shade Drying and Open-Air Sun Drying on Mint Leaf Quality. Applied Sciences 2023, 13, 206.
18. Ropelewska, E. Distinguishing lacto-fermented and fresh carrot slice images using the Multilayer Perceptron neural network and other machine learning algorithms from the groups of Functions, Meta, Trees, Lazy, Bayes and Rules. Eur. Food Res. Technol. 2022, 248, 2421–2429.
19. Ropelewska, E.; Sabanci, K.; Aslan, M.F. The Changes in Bell Pepper Flesh as a Result of Lacto-Fermentation Evaluated Using Image Features and Machine Learning. Foods 2022, 11, 2956.
20. Ropelewska, E.; Wrzodak, A. The Use of Image Analysis and Sensory Analysis for the Evaluation of Cultivar Differentiation of Freeze-Dried and Lacto-Fermented Beetroot (Beta vulgaris L.). Food Analytical Methods 2022, 15, 1026-1041.
21. Nazari, L.; Khazaei, A.; Ropelewska, E. Prediction of tannin, protein, and total phenolic content of grain sorghum using image analysis and machine learning. Cereal Chemistry 2022, 99(4), 843-849.
22. Ropelewska, E.; Szwejda-Grzybowska, J. Relationship of Textures from Tomato Fruit Images Acquired Using a Digital Camera and Lycopene Content Determined by High-Performance Liquid Chromatography. Agriculture-Basel 2022, 12, 1495.
23. Aggarwal, S.; Gupta, S.; Gupta, D.; Gulzar, Y.; Juneja, S.; Alwan, A.A.; Nauman, A. An Artificial Intelligence-Based Stacked Ensemble Approach for Prediction of Protein Subcellular Localization in Confocal Microscopy Images. Sustainability 2023, 15, 1695.
24. Fernandez, L.; Castillero, C.; Aguilera, J. An application of image analysis to dehydration of apple discs. Journal of Food Engi-neering 2005, 67(1-2), 185-193

 

Comment: 2.2 how many apples were chosen for dataset? mention in manuscript

Response: It has been mentioned as follows:

“5 kg of apples were used in each drying experiment.”

 

Comment: 2.3.1: mentioned the dataset size per class per apple in a tabular form to increase the readability

Response: It has been corrected as follows:

“the dataset included:

- one hundred freeze-dried slices of the ‘Alex Red’ red-fleshed apple genotype,

- one hundred freeze-dried slices of the ‘Trinity’ red-fleshed apple genotype,

- one hundred freeze-dried slices of the ‘314’ red-fleshed apple genotype,

- one hundred freeze-dried slices of the ‘602’ red-fleshed apple genotype,

and

- one hundred freeze-dried cubes of the ‘Alex Red’ red-fleshed apple genotype,

- one hundred freeze-dried cubes of the ‘Trinity’ red-fleshed apple genotype,

- one hundred freeze-dried cubes of the ‘314’ red-fleshed apple genotype,

- one hundred freeze-dried cubes of the ‘602’ red-fleshed apple genotype.”

 

Comment: Table 1: fruit weight is in average?

Response: It has been corrected as “Average fruit weight (g)”.

 

Comment: 3.2: as mentioned above authors have focused on data while totally ignored the machine learning part. No information whatsoever is given about models.

Response: Information about machine learning models had been expanded in sections 2. Materials and Methods and 3. Results and Discussion as follows:

“The traditional machine learning models for distinguishing apple samples were developed based on selected image textures and image textures combined with color parameters using traditional machine learning algorithms. Four different approaches to the classification of freeze-dried apple samples were applied. Firstly, the classification was carried out using texture parameters selected from all color channels of freeze-dried slice images of ‘Alex Red’, ‘Trinity’, ‘314’, and ‘602’ red-fleshed apples. Then, also color parameters were added to a set of selected image textures, and the analysis was performed for a set of selected image textures and color features of apple slices. Afterward, freeze-dried cubes of ‘Alex Red’, ‘Trinity’, ‘314’, and ‘602’ apples were classified using models involving attributes selected from a set of textures extracted from images in all color channels. In the next step of the analysis, classification models were built using a combined set of selected image textures and color parameters of apple cubes. The attribute selection was performed using the Best First and CFS (Correlation-based Feature Selection) subset evaluator. This step was performed separately for a set of image textures of apple slices, a combined set of image textures and color parameters of apple slices, a set of image textures of apple cubes, and a combined set of image textures and color parameters of apple cubes. In the case of apple slices, the selected image textures belonged to color channels R, G, B, L, a, b, X, Z, U, V, and S. Among the color parameters, L*(D65) and b*(D65) were characterized by the highest power to distinguish freeze-dried slices. For apple cubes, the textures from images in color channels R, G, B, L, a, b, X, Y, Z, U, and S were selected. Additionally, color parameters L*(D65), a*(D65), b*(D65) were added to models. The classifications were performed using a test mode of 10-fold cross-validation. The dataset of freeze-dried red-fleshed apple slices was randomly divided into 10 parts. Each part was treated in turn as the test set and the 9 parts as the training sets. The learning procedure was performed 10 times on different training sets. The overall error was determined as the average of 10 error estimates. Selected features were used for building classification models using machine learning algorithms from the groups of Functions, Lazy, Meta, Trees, and Bayes. In the case of each group, the most successful algorithm was chosen. The LDA (Linear Discriminant Analysis) from the group of Functions), IBk (Instance-Based k) from Lazy, LogitBoost from Meta, LMT (Logistic Model Tree) from Trees, and Bayes Net from Bayes were chosen. IBk is a k-nearest-neighbors classifier. LogitBoost performs additive logistic regression. LMT builds logistic model trees. The function of Bayes Net is learning Bayesian nets [33-35]. The criterion for the selection of the algorithms was the highest overall accuracy.”

“Traditional machine learning models developed using selected image texture parameters extracted from the images of freeze-dried sliced samples in color channels yielded an overall classification accuracy of 90.25% for the studied genotypes and the LDA algorithm from the group of Functions (Table 2). ”

“The optimal model was characterized by 361 correctly classified cases and 39 incorrectly classified cases with values of Kappa, MAE, and RMSE of 0.8700, 0.0545, and 0.2010, respectively.”

“Slightly lower overall classification accuracies were obtained using the model built by LMT from the group of Trees (89.50%), the LogitBoost from Meta (85.50%), and the IBk from Lazy (85.25%).”

“Whereas the lowest overall accuracy of 84.75% was observed for the model built using Bayes Net from the group of Bayes. In this model, 339 cases were correctly classified, whereas 61 cases were misclassified. Kappa statistic reached 0.7967, MAE was equal to 0.0768, RMSE was 0.2669, and the accuracies for the individual classes were equal to 83% for ‘Alex Red’, 86% for ‘Trinity’, 84% for ‘314’, and 86% for ‘602’.”

“Including selected color parameters (L*(D65), b*(D65)) in classification models increased the overall accuracies of the classification of freeze-dried red-fleshed apple slices. Models combining selected image textures and color parameters were characterized by overall accuracy ranging from 85.50% (Bayes Net) to 91.25% (LDA) as shown in Table 4. For the model developed using the LDA algorithm, the Kappa statistic of 0.8833 was the highest.”

“The classification models developed based on selected image textures and color parameters of apple slices were characterized by values of the TP Rate, Precision, F-Measure, MCC, ROC Area, PRC Area less than 1.000 and FP Rate higher than 0.000 for each genotype (Table 5). The models developed for selected image textures and color parameters showed better classification than those including only selected image texture parameters”

“The models built using LMT produced also the highest Precision and F-Measure of 0.940, and MCC equal to 0.920 for ‘Trinity’. Whereas the values of ROC Area (0.990) and PRC Area (0.973) were the highest for ‘Trinity’ and the model developed using Bayes Net.”

“For the classification of images of apple cubes, models including selected image textures produced overall accuracies ranging from 68.50% (IBk) to 74.74% (LMT)”

“Models developed based on selected textures from cubes’ images in selected color channels and color parameters (L*(D65), a*(D65), b*(D65)) were characterized by an overall accuracy of 71.75% (IBk) to 80.50% (LogitBoost)”

 

Comment: How did you build model/models? what models?

Response: It has been corrected as follows:

“The traditional machine learning models for distinguishing apple samples were developed based on selected image textures and image textures combined with color parameters using traditional machine learning algorithms. Four different approaches to the classification of freeze-dried apple samples were applied. Firstly, the classification was carried out using texture parameters selected from all color channels of freeze-dried slice images of ‘Alex Red’, ‘Trinity’, ‘314’, and ‘602’ red-fleshed apples. Then, also color parameters were added to a set of selected image textures, and the analysis was performed for a set of selected image textures and color features of apple slices. Afterward, freeze-dried cubes of ‘Alex Red’, ‘Trinity’, ‘314’, and ‘602’ apples were classified using models involving attributes selected from a set of textures extracted from images in all color channels. In the next step of the analysis, classification models were built using a combined set of selected image textures and color parameters of apple cubes.”

 

Comment: How did you extract the parameters from dataset?

Response: It has been explained as follows:

“freeze-dried cubes of ‘Alex Red’, ‘Trinity’, ‘314’, and ‘602’ apples were classified using models involving attributes selected from a set of textures extracted from images in all color channels. In the next step of the analysis, classification models were built using a combined set of selected image textures and color parameters of apple cubes. The attribute selection was performed using the Best First and CFS (Correlation-based Feature Selection) subset evaluator. This step was performed separately for a set of image textures of apple slices, a combined set of image textures and color parameters of apple slices, a set of image textures of apple cubes, and a combined set of image textures and color parameters of apple cubes.”

 

Comment: How did you get 90.25% accuracy (line 239) what did you do?

Response: It has been explained as follows:

 “The overall accuracy of 90.25% referred to 92 correctly classified cases from class ‘Alex Red’, 90 cases from ‘Trinity’, 88 cases from ‘314’, and 91 cases from ‘602’.”

 

Comment: What is LDA, IBk , LogitBoost, LMT, Bayes Net algorithm? Why did you choose that to use it?

Response: It has been explained as follows:

“Selected features were used for building classification models using machine learning algorithms from the groups of Functions, Lazy, Meta, Trees, and Bayes. In the case of each group, the most successful algorithm was chosen. The LDA (Linear Discriminant Analysis) from the group of Functions), IBk (Instance-Based k) from Lazy, LogitBoost from Meta, LMT (Logistic Model Tree) from Trees, and Bayes Net from Bayes were chosen. IBk is a k-nearest-neighbors classifier. LogitBoost performs additive logistic regression. LMT builds logistic model trees. The function of Bayes Net is learning Bayesian nets [33-35]. The criterion for the selection of the algorithms was the highest overall accuracy.”

 

Comment: Table 2 confusion matrix? Is it really confusion matrix? Use a confusion matrix obtained from model not like this

Response: Thank you for this valuable comment. It has been corrected as “the number of correctly and incorrectly classified cases” instead of a confusion matrix.

 

Comment: Line 265: “Other performance metrics for classification models are presented in Table 3. None 265 of the TP Rate, Precision, F-Measure, MCC, ROC Area, and PRC Area reached 1.000 which 266 confirmed that none of the samples were classified with 100% accuracy.”

What is TP, MCC, ROC, PRC? Without providing full details you just wrote them like that?

Response: It was explained in subsection 2.4. Statistical analysis as follows:

“TP (True Positive) Rate, FP (False Positive) Rate, Precision, F-Measure, MCC (Matthews Correlation Coefficient), ROC (Receiver Operating Characteristic) Area, and PRC (Precision-Recall) Area”

and as footnotes below under each table as follows:

“TP Rate - True Positive Rate, FP Rate - False Positive Rate, MCC - Matthews Correlation Coefficient, ROC Area - Receiver Operating Characteristic Area, PRC Area - Precision-Recall Area”

 

Comment: What is the purpose of this study? Just to classify freeze-dried apples for what? (conclusion)

Response: The results can be of great practical application in food processing. It was explained as follows:

“The approach involving image textures and color parameters proved to be useful for distinguishing freeze-dried samples of red-fleshed apples belonging to the ‘Alex Red’, ‘Trinity’, ‘314’, and ‘602’ genotypes using machine learning algorithms. The developed procedure can be used in practice to avoid mixing different genotypes of apples with different chemical properties.”

Author Response File: Author Response.docx

Reviewer 2 Report

Classification of freeze-dried slices and cubes of red-fleshed  apple genotypes evaluated using image textures, color parameters and machine learning

The works proposes the application of image analysis using texture features (GLCM) and color to classify apples from different varieties after drying.

The novelty resides on the application of the method, since image analysis/CV has several reported applications in agricultural products.

The text is well written, I could suggest a short topic sentence in the abstract to introduce the relevance of the work. In addition, it is necessary to improve the background for the work, highlighting the need for such method. The authors mention in lines 43-59 that it is important to preserve functional compounds in fruits during drying, but the authors did not measure these compounds in the samples. Hence, there is a major gap that must be fulfilled, justifying the application of such computer vision approach, since it is not clear why it is important to identify apple varieties after drying. I suggest the authors to reorganize and rewrite the whole introduction to justify the investigation.

Material and methods

Please clearly describe the number of samples analyzed, for fruits and images. It is mandatory to have an external set of samples for validation

Introduction/Results and discussion

The authors report that samples were frozen at -28oC for one day, and dried for 48h. What was the final moisture content? This should be determined, or at least samples should be dried until constant weight, as this may highly affect the samples. Please see some works that have determined the influence of drying in agricultural samples such as melon, beet, apple, etc, and address these works for discussion, and how it could have affected the current results.

In addition, some works have previously discussed the effect of drying temperature in functional properties of food, and detected in comparison with color. I suggest the authors to compare the results obtained with these works as it could be relevant to the scientific community how the components of apple were affected, since the authors did not perform chemical measurements:

I suggest the authors to present an analysis of importance for the parameters used and discuss which parameter had the most influence on the results, as it may provide deeper understanding of the results and great support for future works.

Author Response

Classification of freeze-dried slices and cubes of red-fleshed  apple genotypes evaluated using image textures, color parameters and machine learning

The works proposes the application of image analysis using texture features (GLCM) and color to classify apples from different varieties after drying.

The novelty resides on the application of the method, since image analysis/CV has several reported applications in agricultural products.

Response: Thank you very much for your careful reading of the manuscript and your valuable comments.

 

Comment: The text is well written, I could suggest a short topic sentence in the abstract to introduce the relevance of the work. In addition, it is necessary to improve the background for the work, highlighting the need for such method. The authors mention in lines 43-59 that it is important to preserve functional compounds in fruits during drying, but the authors did not measure these compounds in the samples. Hence, there is a major gap that must be fulfilled, justifying the application of such computer vision approach, since it is not clear why it is important to identify apple varieties after drying. I suggest the authors to reorganize and rewrite the whole introduction to justify the investigation.

Response: The abstract has been supplemented with the following sentence:

“Dried red-fleshed apples are considered a promising high-quality product from the functional foods category.”

The Introduction has been corrected as follows:

“Red-fleshed apples with high content of anthocyanins and other phenolic compounds are characterized by a strong antioxidant capacity [5]. They are attractive to consumers due to their red color and positive effect on health [6]. Red-fleshed apples can be consumed in the form of snacks. Drying is a promising process for producing healthy snacks with the retention of bioactive compounds and nutrients. The presence of anthocyanins in the flesh of red-fleshed apples results in the obtaining of value-added snack products [7]. It was reported that ready-to-eat dried snacks can be interesting and attractive food products [8]. Drying is used to preserve food by reducing water content and thus inhibiting microbial growth. Drying at a low temperature and reduced pressure minimizes changes in the dried material and the loss of heat-labile compounds. Freeze-drying is one of the least destructive drying methods in terms of phenolic compounds due to low temperature and scant contact with air [9]. Freeze-drying is non-thermal processing technology. It was found that thermal processing caused greater changes in red-fleshed apple snacks. For example, infrared-drying resulted in great losses in the apple (poly)phenolics. Hot air-drying maintained approximately 83% of the total (poly)phenols compared with the freeze-dried samples, and purée pasteurization only 65%. Whereas the degradation of anthocyanins was higher and hot air-dried apple snacks maintained 26% and pasteurized purée samples only 9% compared with freeze-dried apple snacks [10]. Wojdyło et al. [11] reported that novel red-fleshed apple snacks are a promising high-quality dehydrated product belonging to the functional foods category. The authors [11] revealed that freeze-drying allowed for the retaining of the highest content of bioactive compounds in dried red-fleshed apple snacks, followed by hybrid (convective pre-drying + vacuum-microwave drying), vacuum-microwave, and convection drying.”

“Machine Learning (ML) algorithms aim to automate the learning process through training and then the optimized regression and/or classification algorithms can be utilized to predict accurate estimation of the unknown samples/objects [15]. ML is a key tool for Industry 4.0 where all elements of the manufacturing process can be interconnected through intelligent data analysis [16].”

“Image analysis and machine learning were successfully applied in previous studies to classify fruit and vegetables [17-19] and detect the changes in the product quality as a result of different processing, such as drying and fermentation [20-23]. Furthermore, image features were used to estimate and predict the chemical properties of food products [24,25]. Machine learning techniques can be very effective in predicting chemical substances [26]. In the case of apples, the usefulness of image processing to determine the effect of drying on shrinkage, color, and image texture of slices was proved. The correlation was feasible between drying time and the slice area, perimeter, and other morphological features of the image. Moreover, a*, and b* values showed a stable increase with the drying time [27]. While previous studies presented an understanding of how to correlate image texture with the quality of the dried apple, no study investigated the possibility of using digital imaging for evaluating the quality of freeze-dried red-fleshed apples.”

 

Comment: Material and methods

Please clearly describe the number of samples analyzed, for fruits and images. It is mandatory to have an external set of samples for validation

Response: It has been indicated as follows:

The number of samples analyzed for fruits:

“Qualitative characteristics were assessed using standard methods… The measurements were carried out in four replicates.”

 

The number of samples analyzed for images:

“In total, digital color images of one hundred freeze-dried slices and one hundred freeze-dried cubes of each of the ‘Alex Red’, ‘Trinity’, ‘314’, and ‘602’ red-fleshed apple genotypes were obtained.

Thus, the dataset included:

- one hundred freeze-dried slices of the ‘Alex Red’ red-fleshed apple genotype,

- one hundred freeze-dried slices of the ‘Trinity’ red-fleshed apple genotype,

- one hundred freeze-dried slices of the ‘314’ red-fleshed apple genotype,

- one hundred freeze-dried slices of the ‘602’ red-fleshed apple genotype,

and

- one hundred freeze-dried cubes of the ‘Alex Red’ red-fleshed apple genotype,

- one hundred freeze-dried cubes of the ‘Trinity’ red-fleshed apple genotype,

- one hundred freeze-dried cubes of the ‘314’ red-fleshed apple genotype,

- one hundred freeze-dried cubes of the ‘602’ red-fleshed apple genotype.”

 

The validation has been described as follows:

“The classifications were performed using a test mode of 10-fold cross-validation. The dataset of freeze-dried red-fleshed apple slices was randomly divided into 10 parts. Each part was treated in turn as the test set and the 9 parts as the training sets. The learning procedure was performed 10 times on different training sets. The overall error was determined as the average of 10 error estimates.”

The applied procedure ensured external sets of samples for validation. The training sets were not treated as the test set in each estimation.

 

Comment: Introduction/Results and discussion

The authors report that samples were frozen at -28oC for one day, and dried for 48h. What was the final moisture content? This should be determined, or at least samples should be dried until constant weight, as this may highly affect the samples. Please see some works that have determined the influence of drying in agricultural samples such as melon, beet, apple, etc, and address these works for discussion, and how it could have affected the current results.

Response: The water activity of dried apple samples was measured. The differences between sample forms and genotypes were observed. The water activity of slices was in the range of about 0.20 for ‘314’ and ‘602’ and 0.15 for ‘Alex Red’ and ‘Trinity’ and cubes of 0.22 for ‘314’ and ‘602’ and 0.17 for ‘Alex Red’ and ‘Trinity’. As you mentioned, the final moisture content can affected the results. This relationship between  moisture content and image features has been discussed as follows:

“The application of optical sensors in general and CIS in particular for monitoring the quality of apple slices during drying was feasibly implemented in several studies. Digital images were studied to yield a high correlation between CIELAB color parameters (L*, a*, b*, and ΔE) and either the moisture content or the drying time with values of coefficient of determination, or R², of 92.0-96.0% for moisture content and 68.0-99.0% for the drying time [38].”

“The results obtained in this study also agree with those deduced by Sampson et al. [39]. Authors [39] investigated the utilization of a dual-camera system to evaluate the drying characteristics of apple slices. Texture features of images resulted in a coefficient of correlation, or r, of as high as 55.9% for the peak force. Whereas image texture parameters presented R² values higher than 90% for predicting the change of moisture content of hot air-dried organic apple slices (‘Gala’). Another study by Raponi et al. [40], used a CMOS color camera-vision system to evaluate the quality of apple cylinders during and after drying. The results indicated R² values of 99.8% for predicting moisture content as a function of the shrinkage in the surface area of the disc sample based on image processing.”

 

Comment: In addition, some works have previously discussed the effect of drying temperature in functional properties of food, and detected in comparison with color. I suggest the authors to compare the results obtained with these works as it could be relevant to the scientific community how the components of apple were affected, since the authors did not perform chemical measurements:

Response: The Results and discussion section has been enriched as follows:

“Previous literature data revealed using image textures in studies on apple drying. The application of optical sensors in general and CIS in particular for monitoring the quality of apple slices during drying was feasibly implemented in several studies. Digital images were studied to yield a high correlation between CIELAB color parameters (L*, a*, b*, and ΔE) and either the moisture content or the drying time with values of coefficient of determination, or R², of 92.0-96.0% for moisture content and 68.0-99.0% for the drying time [38]. Among others, it is important to state that image texture parameters such as energy, contrast, entropy, and inverse different moment were reported to achieve a positive correlation with different drying times for apple slices [27]. The results obtained in this study also agree with those deduced by Sampson et al. [39]. Authors [39] investigated the utilization of a dual-camera system to evaluate the drying characteristics of apple slices. Texture features of images resulted in a coefficient of correlation, or r, of as high as 55.9% for the peak force. Whereas image texture parameters presented R² values higher than 90% for predicting the change of moisture content of hot air-dried organic apple slices (‘Gala’). Another study by Raponi et al. [40], used a CMOS color camera-vision system to evaluate the quality of apple cylinders during and after drying. The results indicated R² values of 99.8% for predicting moisture content as a function of the shrinkage in the surface area of the disc sample based on image processing. However, in the case of freeze-dried red-fleshed apples, the effectiveness of the models built using selected image textures and color parameters to distinguish genotypes was not confirmed by previous literature. Therefore, the present results are very promising. Another note is the LDA and LMT classifiers generally showed a consistently higher classification performance than other algorithms. The results illustrated the advantage of LDA and LMT algorithms. LDA can be effectively used for dimensional reduction for data sets with a relatively large number of features compared with the number of samples in the training set [41]. LMT benefits from employing logistic regression which leads to significantly reducing the number of features used to build the classification model and consequently reduces the likelihood of overfitting [42].”

 

Comment: I suggest the authors to present an analysis of importance for the parameters used and discuss which parameter had the most influence on the results, as it may provide deeper understanding of the results and great support for future works.

Response: The selection of parameters has been performed. It has been described as follows:

“The attribute selection was performed using the Best First and CFS (Correlation-based Feature Selection) subset evaluator. This step was performed separately for a set of image textures of apple slices, a combined set of image textures and color parameters of apple slices, a set of image textures of apple cubes, and a combined set of image textures and color parameters of apple cubes. In the case of apple slices, the selected image textures belonged to color channels R, G, B, L, a, b, X, Z, U, V, and S. Among the color parameters, L*(D65) and b*(D65) were characterized by the highest power to distinguish freeze-dried slices. For apple cubes, the textures from images in color channels R, G, B, L, a, b, X, Y, Z, U, and S were selected. Additionally, color parameters L*(D65), a*(D65), b*(D65) were added to models.”

 

 

Author Response File: Author Response.docx

Reviewer 3 Report

The study is an important methodical investigation describing a new approach in quality control for red apples, dried slices and cubes of its flesh. It combines image analysis and machine learning. The manuscript is recommended for publication in “Agriculture” with minor revisions. Please consider the following points when revising the manuscript: 

(1)       General: In fruit science the term “texture” is used to describe the mechanical properties, physical food structure and design. Please define the term as it was used here as “image texture” and thus avoid misunderstanding.

 

(2)       General: The fruit material examined here primarily comprise different fruit samples/batches. Their characteristics are not exclusively dependent on genotype. They also reflect the variability due to developmental stages and environmental responses (e.g.,  '314' showed increased starch breakdown, decreased firmness, increased ethylene concentration compared to the other batches, indicating progressive maturation).

 

(3)       Line 98-99: "...used in practice to avoid mixing of different genotypes". I think the problem is not the mixing of genotypes, but of fruit batches of different quality, i.e. with different concentrations of valuable ingredients. Unfortunately, no classification of anthocyanin and/or flavonoids content, antioxidant activity, etc. was made in the present study. In my opinion, these or similar parameters are interesting for quality control. If they are accurately estimated using image analysis combined with machine learning, a new, highly effective and advanced technology will become available for the control of red apple products. However, in the present paper, the authors should emphasize how their research helps to achieve this goal.

 

(4)       Lines 119-131: The description of the analytical methods is incomplete and not reproducible in the present form. It can be improved by inserting literature which explain the missing details. Here are some examples of critical points: The specification of the GC column und solvents is missing. The unit of firmness (kg) sounds strange; firmness was quantified in a penetration test, the unit should be force per square unit (e.g. N mm-2), not mass. What do the TTS and TA percentages refer to?

 

(5)       Line 137: The cube has three dimensions: 1.5 x 1.5 x ? cm.

 

(6)       Line 154; Figs. 1, 2: The images shown in Figs. are not the original scans. They are processed (background = black). Please indicate the processing steps in the caption. The edges of specimens can cause some trouble during image processing (e.g. shadows: slices in Fig. 1 appears “peeled”, side face of the cubes are partly visible, see Fig. 2). How did you deal with these problems?

 

(7)       Lines 176-183: L*a*b were extracted from pixel in scans and recorded by the spectrometer. What are the differences? Why different measurements? Describe when L*a*b from scans and the spectrometer were used?

 

(8)       Line 223-231: How is "harvest maturity" to be defined? There are significant differences between the genotypes/samples studied, see the example discussed in (2). Ethylene acts in the flesh of the fruit. What role does it play in the seed chamber?

 

(9)       Tab. 2-9, lines 188-216, 384-409: In my opinion, it would be helpful to add some basic information about the algorithms and mathematical tools. References can point to more detailed information. What were the criteria for their selection? What are the distinguishing features, advantages, disadvantages of the best models (the discussion in l. 400-409 is fragmentary)? Why did you use L*b and L*a*b in slices (3.3.) and cubes (3.5.), respectively? Is there information about the most and least important parameters that influence the results?

 

(10)   Lines 43-46: Check phrasing at the beginning of the sentences.

 

(11)   Line 51: "Drying", Line 130: “Mettler-Toledo”, Line 185: “analyses”

 

 

Author Response

The study is an important methodical investigation describing a new approach in quality control for red apples, dried slices and cubes of its flesh. It combines image analysis and machine learning. The manuscript is recommended for publication in “Agriculture” with minor revisions. Please consider the following points when revising the manuscript: 

Response: Thank you very much for your careful reading of the manuscript and your valuable comments. All the comments have been considered and the manuscript has been significantly improved.

 

(1)       General: In fruit science the term “texture” is used to describe the mechanical properties, physical food structure and design. Please define the term as it was used here as “image texture” and thus avoid misunderstanding.

Response: The term “image texture” were used in the whole manuscript instead of “texture”.

The explanation has been added as follows: “The image texture may be defined as a function of the spatial variation of the pixel brightness intensity. Image textures carry important information about the physical object structure. The quantitative analyses of image texture parameters provide insights into product quality [31, 32].”

 

 

(2)       General: The fruit material examined here primarily comprise different fruit samples/batches. Their characteristics are not exclusively dependent on genotype. They also reflect the variability due to developmental stages and environmental responses (e.g.,  '314' showed increased starch breakdown, decreased firmness, increased ethylene concentration compared to the other batches, indicating progressive maturation).

Response: The variability was related to the genotype. The genotype affected maturation. It has been explained as follows:

“Fruit ripening depended on the genotype. This means that genotypes in the experiment can be divided into two categories: late summer (‘Alex Red’ and ‘Trinity’) and early autumn (‘314’ and ‘602’). The values of the starch index indicate that the harvested fruits were in the phase of harvest maturity. This phase meant that apples were mature, but not overripe (Table 1). Internal ethylene concentration in apple core also confirms this thesis. The fruits of the ‘314’ clone were more advance mature compare to other evaluated cultivars/clone. The firmness of the flesh was varied. It had significantly higher values for clone ‘602’ whereas the lowest for clone ‘314’; while for the genotypes ‘Alex Red’ and ‘Trinity’ were similar. Sugar content and acidity were related to the date of harvest. Late summer genotypes were characterized by lower sugar content and higher acidity compared to early autumn apples.”

 

 (3)       Line 98-99: "...used in practice to avoid mixing of different genotypes". I think the problem is not the mixing of genotypes, but of fruit batches of different quality, i.e. with different concentrations of valuable ingredients. Unfortunately, no classification of anthocyanin and/or flavonoids content, antioxidant activity, etc. was made in the present study. In my opinion, these or similar parameters are interesting for quality control. If they are accurately estimated using image analysis combined with machine learning, a new, highly effective and advanced technology will become available for the control of red apple products. However, in the present paper, the authors should emphasize how their research helps to achieve this goal.

Response: In the present study traditional machine learning algorithms proved to be useful to distinguish samples with very satisfactory accuracies and low errors. The future studies are planes as indicated in the manuscript:

“In further own studies, the classification accuracy can be improved, for example, by using deep learning.”

“Moreover, in our future studies, image textures and color parameters may also be used in other types of drying experiments for the prediction of flesh changes caused by various drying techniques and the estimation of chemical properties of dried samples by regression equations built based on image textures and color features.”

  “Additionally, in future research, image textures and color parameters may be used for the prediction of the changes in the flesh structure of red-fleshed apples caused by various drying techniques and for the non-destructive and objective estimation of the chemical properties of dried samples. “

 

In additions, the Introduction has been supplemented with the following sentences:

“Freeze-drying is non-thermal processing technology. It was found that thermal processing caused greater changes in red-fleshed apple snacks. For example, infrared-drying resulted in great losses in the apple (poly)phenolics. Hot air-drying maintained approximately 83% of the total (poly)phenols compared with the freeze-dried samples, and purée pasteurization only 65%. Whereas the degradation of anthocyanins was higher and hot air-dried apple snacks maintained 26% and pasteurized purée samples only 9% com-pared with freeze-dried apple snacks [10].”

 

(4)       Lines 119-131: The description of the analytical methods is incomplete and not reproducible in the present form. It can be improved by inserting literature which explain the missing details. Here are some examples of critical points: The specification of the GC column und solvents is missing. The unit of firmness (kg) sounds strange; firmness was quantified in a penetration test, the unit should be force per square unit (e.g. N mm-2), not mass. What do the TTS and TA percentages refer to?

Response: It has been corrected as follows:

“To assess the quality of apples of all genotypes in the experiment, fruits were randomly picked during harvest. Qualitative characteristics were assessed using standard methods. Fruit weight was determined by weighing individual fruits on a WPS2100/C/2 laboratory balance (Radwag, Poland), and expressed in grams (g). The starch index was determined using a ten-point scale in the standard iodine test (1-black, 10-white) using the “Starch conversion chart for apples” Ctifl (France). For internal ethylene concentration (IEC), a 1 ml gas sample was taken from the apple core and injected into an HP 5890 II gas chromatograph equipped with alumina packed glass column (6 mm diameter and 1200mm length, packed with Alumina F-1, 60/80 mesh) and detector FID. Results were expressed in µl L^-1 (ppm). Flesh firmness was measured on two opposite sides of the fruit (after skin removing) using Zwick Roell Z010 (Germany) equipped with a Magness-Taylor 11.1 mm probe. The speed with which the head moved during a single firmness measurement was 100 mm/min. Firmness was defined as the maximum force needed to penetrate the plunger into the flesh to a depth of 8.7 mm. The results were expressed in newtons [N]. Total soluble sugar content (TSS) was measured (in juice collected from individual fruits) in fresh juices with a digital refractometer Atago PR-101 (Atago Co. Ltd., Japan) and expressed as % (°Brix). Titratable acidity (TA) was measured using an automatic titrator DL 21 (Mettler-Toledo, Swiss), standard titration method: titration with 0.1 N NaOH to the end point pH=8.1, and expressed in % as a malic acid [28]. The measurements were carried out in four replicates.”

 

 

 

(5)       Line 137: The cube has three dimensions: 1.5 x 1.5 x ? cm.

 Response: It has been corrected as 1.5 cm x 1.5 cm x 1.5 cm

 

(6)       Line 154; Figs. 1, 2: The images shown in Figs. are not the original scans. They are processed (background = black). Please indicate the processing steps in the caption. The edges of specimens can cause some trouble during image processing (e.g. shadows: slices in Fig. 1 appears “peeled”, side face of the cubes are partly visible, see Fig. 2). How did you deal with these problems?

 Response: It has been explained as follows:

“Firstly, the background of the slice and cube images was changed to black, and the images were saved in the BMP file format. This step allowed for image segmentation and feature extraction.”

“The image segmentation into the black background and lighter apple samples was carried out using the manually determined brightness threshold. Each apple slice or cube was considered as one region of interest (ROI).”

 

 

(7)       Lines 176-183: L*a*b were extracted from pixel in scans and recorded by the spectrometer. What are the differences? Why different measurements? Describe when L*a*b from scans and the spectrometer were used?

 Response: The determined image texture parameters and color parameters were completely different. In the case of scans, L, a, and b were the color channels. In the case of each color channel of images, the texture parameters were computed. Color parameters were not extracted from images, only textures. Therefore, color parameters L*, a, and b were determined additionally in other measurements. Texture parameters indicated the changes in the structure of samples and color parameters L*, a, and b indicated the color changes.

The following sentences have been included in the manuscript.

“The images were converted to color channels R, G, B, L, a, b, X, Y, Z, U, V, and S.”

“For each ROI, 2172 image texture parameters including 181 textures based on the gradient map, histogram, Haar wavelet transform, autoregressive model, co-occurrence matrix, and run-length matrix for each color channel were determined.”

“The image texture may be defined as a function of the spatial variation of the pixel brightness intensity. Image textures carry important information about the physical object structure. The quantitative analyses of image texture parameters provide insights about product quality [31, 32].”

“Color parameters L* (lightness from 0 (dark) to 100 (light)), a* (red (+) – green (-)), and b* (yellow (+) – blue (-)) were measured using the Konica Minolta CM-2500c portable spectrophotometer.”

 

(8)       Line 223-231: How is "harvest maturity" to be defined? There are significant differences between the genotypes/samples studied, see the example discussed in (2). Ethylene acts in the flesh of the fruit. What role does it play in the seed chamber?

 Response: It has been corrected as follows:

“Fruit ripening depended on the genotype. This means that genotypes in the experiment can be divided into two categories: late summer (‘Alex Red’ and ‘Trinity’) and early autumn (‘314’ and ‘602’). The values of the starch index indicate that the harvested fruits were in the phase of harvest maturity. This phase meant that apples were mature, but not overripe (Table 1). Internal ethylene concentration in apple core also confirms this thesis. The fruits of the ‘314’ clone were more advance mature compare to other evaluated cultivars/clone.”

 

(9)       Tab. 2-9, lines 188-216, 384-409: In my opinion, it would be helpful to add some basic information about the algorithms and mathematical tools. References can point to more detailed information. What were the criteria for their selection? What are the distinguishing features, advantages, disadvantages of the best models (the discussion in l. 400-409 is fragmentary)? Why did you use L*b and L*a*b in slices (3.3.) and cubes (3.5.), respectively? Is there information about the most and least important parameters that influence the results?

Response: The information about the algorithms, attribute selection, and the criterion for the selection of the algorithms have been described in more detail as follows:

“The attribute selection was performed using the Best First and CFS (Correlation-based Feature Selection) subset evaluator. This step was performed separately for a set of image textures of apple slices, a combined set of image textures and color parameters of apple slices, a set of image textures of apple cubes, and a combined set of image textures and color parameters of apple cubes. In the case of apple slices, the selected image textures belonged to color channels R, G, B, L, a, b, X, Z, U, V, and S. Among the color parameters, L*(D65) and b*(D65) were characterized by the highest power to distinguish freeze-dried slices. For apple cubes, the textures from images in color channels R, G, B, L, a, b, X, Y, Z, U, and S were selected. Additionally, color parameters L*(D65), a*(D65), b*(D65) were added to models. The classifications were performed using a test mode of 10-fold cross-validation. The dataset of freeze-dried red-fleshed apple slices was randomly divided into 10 parts. Each part was treated in turn as the test set and the 9 parts as the training sets. The learning procedure was performed 10 times on different training sets. The overall error was determined as the average of 10 error estimates. Selected features were used for building classification models using machine learning algorithms from the groups of Functions, Lazy, Meta, Trees, and Bayes. In the case of each group, the most successful algorithm was chosen. The LDA (Linear Discriminant Analysis) from the group of Functions), IBk (Instance-Based k) from Lazy, LogitBoost from Meta, LMT (Logistic Model Tree) from Trees, and Bayes Net from Bayes were chosen. IBk is a k-nearest-neighbors classifier. LogitBoost performs additive logistic regression. LMT builds logistic model trees. The function of Bayes Net is learning Bayesian nets [33-35]. The criterion for the selection of the algorithms was the highest overall accuracy.”

33. Witten, I.H.; Frank, E. Data mining: Practical machine learning tools and techniques (525, 2nd ed.). San Francisco, CA: Elsevier 2005.
34. Bouckaert, R.R.; Frank, E.; Hall, M.; Kirkby, R.; Reutemann, P.; Seewald, A.; Scuse, D. WEKA manual for version 3-9-1. University of Waikato, Hamilton, New Zealand 2016.
35. Frank, E.; Hall, M.A.; Witten, I.H. The WEKA Workbench. Online Appendix for Data Mining: Practical Machine Learning Tools and Techniques, Morgan Kaufmann, Fourth Edition 2016.

 

 

(10)   Lines 43-46: Check phrasing at the beginning of the sentences.

 Response: It has been corrected as follows:

“Red-fleshed apples with high content of anthocyanins and other phenolic compounds are characterized by a strong antioxidant capacity [5]. They are attractive to consumers due to their red color and positive effect on health [6]. Red-fleshed apples can be consumed in the form of snacks.”

 

(11)   Line 51: "Drying", Line 130: “Mettler-Toledo”, Line 185: “analyses”

Response: It has been corrected.

Author Response File: Author Response.docx

Reviewer 4 Report

The paper by Ropelewska et al. entitled Classification of freeze-dried slices and cubes of red-fleshed apple genotypes evaluated using image textures, color parameters and machine learning, is dealing with an emerging topic which is the nondestructive evaluation of fruits and vegetables with a good accuracy.

I have read the paper and I think that it would be accptable for publication upon minor revision.

here below are some general comments

1/The abstract is fine but lacks only 1-2 sentence in my opinion highlighting where future works should focus based on the results of this study

2/Introduction section is good

3/Materials and methods section is fine and well organized. The statistical analysis is relatively long sine too much details was inserted here I suggest to reduce the length while deleting already reported data previously 

4/The results and discussion section should be enriched 

authors should compare their data (the obtained accuracy) with other researchers 

although the topic is relatively new by We can find some close works such as 

Rehman, T. U., Mahmud, M. S., Chang, Y. K., Jin, J., & Shin, J. (2019). Current and future applications of statistical machine learning algorithms for agricultural machine vision systems. Computers and electronics in agriculture, 156, 585-605.

Osako, Y., Yamane, H., Lin, S. Y., Chen, P. A., & Tao, R. (2020). Cultivar discrimination of litchi fruit images using deep learning. Scientia Horticulturae, 269, 109360.

5/ Conclusion: here at the end the authors should insert 1 sentence regarding where future research should focus

 6/Reeferences are fine

7/Tabales: authors should try to reduce the number if possible

 

Comments for author File: Comments.pdf

Author Response

The paper by Ropelewska et al. entitled Classification of freeze-dried slices and cubes of red-fleshed apple genotypes evaluated using image textures, color parameters and machine learning, is dealing with an emerging topic which is the nondestructive evaluation of fruits and vegetables with a good accuracy.

Response: Thank you very much for your careful reading of the manuscript and this opinion.

 

I have read the paper and I think that it would be accptable for publication upon minor revision.

Response: Thank you very much for all your comments. They have been considered and the manuscript has been significantly improved.

 

here below are some general comments

Comment 1/The abstract is fine but lacks only 1-2 sentence in my opinion highlighting where future works should focus based on the results of this study

Response:

It has been added as follows:

“Further studies can focus on using deep learning in addition to traditional machine learning to build models to distinguish dried red-fleshed apple samples. Moreover, other drying techniques can be applied, and image texture parameters and color features can be used to predict the changes in flesh structure and estimate the chemical properties of dried samples.”

 

Comment 2/Introduction section is good

Response: Thank you for your comment.

 

Comment 3/Materials and methods section is fine and well organized. The statistical analysis is relatively long sine too much details was inserted here I suggest to reduce the length while deleting already reported data previously 

Response: Thank you for your opinion. The references have been indicated for applied machine learning software and algorithms. However, the developed models are innovative, and their description has been included.

 

Comment 4/The results and discussion section should be enriched 

authors should compare their data (the obtained accuracy) with other researchers 

although the topic is relatively new by We can find some close works such as 

Rehman, T. U., Mahmud, M. S., Chang, Y. K., Jin, J., & Shin, J. (2019). Current and future applications of statistical machine learning algorithms for agricultural machine vision systems. Computers and electronics in agriculture, 156, 585-605.

Osako, Y., Yamane, H., Lin, S. Y., Chen, P. A., & Tao, R. (2020). Cultivar discrimination of litchi fruit images using deep learning. Scientia Horticulturae, 269, 109360.

Response: The Results and discussion section has been enriched and additional articles have been cited as you recommended as follows:

“Previous literature data revealed using image textures in studies on apple drying. The application of optical sensors in general and CIS in particular for monitoring the quality of apple slices during drying was feasibly implemented in several studies. Digital images were studied to yield a high correlation between CIELAB color parameters (L*, a*, b*, and ΔE) and either the moisture content or the drying time with values of coefficient of determination, or R², of 92.0-96.0% for moisture content and 68.0-99.0% for the drying time [38]. Among others, it is important to state that image texture parameters such as energy, contrast, entropy, and inverse different moment were reported to achieve a positive correlation with different drying times for apple slices [27]. The results obtained in this study also agree with those deduced by Sampson et al. [39]. Authors [39] investigated the utilization of a dual-camera system to evaluate the drying characteristics of apple slices. Texture features of images resulted in a coefficient of correlation, or r, of as high as 55.9% for the peak force. Whereas image texture parameters presented R² values higher than 90% for predicting the change of moisture content of hot air-dried organic apple slices (‘Gala’). Another study by Raponi et al. [40], used a CMOS color camera-vision system to evaluate the quality of apple cylinders during and after drying. The results indicated R² values of 99.8% for predicting moisture content as a function of the shrinkage in the surface area of the disc sample based on image processing. However, in the case of freeze-dried red-fleshed apples, the effectiveness of the models built using selected image textures and color parameters to distinguish genotypes was not confirmed by previous literature. Therefore, the present results are very promising. Another note is the LDA and LMT classifiers generally showed a consistently higher classification performance than other algorithms. The results illustrated the advantage of LDA and LMT algorithms. LDA can be effectively used for dimensional reduction for data sets with a relatively large number of features compared with the number of samples in the training set [41]. LMT benefits from employing logistic regression which leads to significantly reducing the number of features used to build the classification model and consequently reduces the likelihood of overfitting [42].”

“Current research dynamics in the applications of machine learning in machine vision systems indicate a further spread of it in agriculture and can make agricultural technologies robust, accurate, and low-cost [43]. In further own studies, the classification accuracy can be improved, for example, by using deep learning. Deep learning (DL) can al-low for precise image classification. The application of deep learning can increase learning capabilities and correctness of image classification due to a hierarchical data representation by means of various convolutions [44]. Moreover, in our future studies, image textures and color parameters may also be used in other types of drying experiments for the prediction of flesh changes caused by various drying techniques and the estimation of chemical properties of dried samples by regression equations built based on image textures and color features.” 

 

Comment 5/ Conclusion: here at the end the authors should insert 1 sentence regarding where future research should focus

Response: It has been indicated as follows:

“In further studies, besides the traditional machine learning algorithms, the models could be built using deep learning. Additionally, in future research, image textures and color parameters may be used for the prediction of the changes in the flesh structure of red-fleshed apples caused by various drying techniques and for the non-destructive and objective estimation of the chemical properties of dried samples.“

 

 

Comment 6/Reeferences are fine

Response: Thank you for this comment.

 

Comment 7/Tabales: authors should try to reduce the number if possible

Response: Eight tables include many relevant results for comprehensive experiments performed on apple slices and cubes. The results present classification performance metrics for models built for image textures and the combination of image textures and color parameters. Tables are the best form of presenting such many results in a clear manner. The reduction of the number of tables would be very difficult and result in less understanding of the results. However, if the Reviewer insists on reducing the number of tables, we can try it.

 

 

 

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

author incorporated all the comments, accept it in current form