You are currently on the new version of our website. Access the old version .
MDPIMDPI
Search all articles
by
  • Asma Channa1,2,*,
  • Giuseppe Ruggeri2 and
  • Nirvana Popescu1
  • et al.

Reviewer 1: Anonymous Reviewer 2: Weitao Li Reviewer 3: Anonymous

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript is a detailed study on a cloud-connected bracelet designed for continuous monitoring of Parkinson's Disease (PD) patients, integrating Continuous Wavelet Transform and AlexNet Network for severity analysis. It discusses the development and implementation of a smart bracelet, AWEAR, equipped with inertial sensors for motion data collection, which is then securely transmitted to the cloud for storage, processing, and severity estimation using bespoke learning algorithms. The manuscript includes an in-depth review of related literature, system overview, design details of the AWEAR bracelet, cloud service integration, and the application of a deep learning model for PD severity scoring. The results demonstrate the system's effectiveness in PD monitoring with high accuracy in estimating bradykinesia and tremor severity. The research content is interesting and with meaningful implications.

 

Comments:

1.     Given the prior introduction of A-WEAR in Sensors 2021, this manuscript should elaborate on the advancements and their significance, especially within the Introduction, Literature Review, Discussion, and Abstract sections.

2.     Table 1 lacks clarity and does not effectively demonstrate A-WEAR's uniqueness or the necessity for comparing data analysis methods for Parkinson's Disease assessment. A revision to establish a clearer logical connection is recommended.

3.     Figure 4 is not easily interpretable due to unreadable x and y axes. Expanding on the impact of nanoparticle ligand on the device's electro-optic response in the Discussion could enhance understanding.

4.     The manuscript's focus, as suggested by the title and abstract on applying machine learning techniques for severity analysis, is overshadowed by extensive hardware component descriptions. It is advisable to adjust the focus to align with the title or consider revising the title to reflect the content more accurately.

5.     Discussion on the device's battery life and durability is essential to gauge its viability for continuous, reliable monitoring without frequent need for recharging or replacement.

6.     Table 2 lacks clarity and focus. A possible solution may be to incorporate A-WEAR into Table 2, which lists wrist-wearable devices for Parkinson's Disease, would underscore its relevance. Relocating this updated table to the Discussion might highlight A-WEAR's significance in the current context.

7.     Table 3 requires further detail and explanation to effectively convey its intended message and enhance the manuscript's contribution. Adding comprehensive information and a clearer exposition of how the data within Table 3 relates to the study's objectives will strengthen the reader's understanding and appreciation of the research's scope and significance.

8.     Figure 8 suffers from readability issues. Enhancements in clarity and a more informative caption are necessary.

9.     The visuals in Figure 9 are also challenging to interpret. Consideration of alternative representations or improvements to existing figures is advised for clarity.

10.  The manuscript would benefit from a section on optimization procedures, including the exploration of various deep learning models beyond the currently used ones, to potentially enhance performance and efficiency in PD severity estimation. A comparative analysis of model performances could offer valuable insights.

Author Response

Thank you for your valuable comments.

  1. We recognize the importance of clearly delineating the advancements made in the development of the A-WEAR bracelet since its initial introduction in Sensors 2021. In our literature review, we now clarify that the initial version of the A-WEAR bracelet, as documented in our previous work, faced several limitations, including the absence of wireless data transmission, the inability to support continuous monitoring, and constraints related to battery life. These limitations, directly addressed in the current version of A-WEAR, represent significant advancements in the device's functionality and application potential for PD patients.
  2. Thank you for your constructive feedback on Table 1. We understand your concerns regarding the clarity and the demonstration of A-WEAR's uniqueness in the context of comparing data analysis methods. Upon review, we want to clarify the purpose and context of Table 1 within our manuscript. Table 1 is designed to provide a comprehensive overview of the existing work in the domain of diagnosing and determining the severity of PD symptoms. Its primary aim is to highlight the progression and the variety of methodologies employed in PD assessment over time, rather than demonstrating the uniqueness of the A-WEAR device directly. This table sets the stage for the subsequent introduction of our A-WEAR device, contextualizing it within the broader landscape of PD monitoring technologies.

To directly address your feedback, we have revised the manuscript to include a clearer explanation of Table 1's role and purpose. We have also ensured that Tables 2 and 9, which specifically focus on the A-WEAR device, its features, and comparative analysis, are highlighted as the primary sources for demonstrating A-WEAR's unique contributions to PD assessment and monitoring.

  1. In response to the feedback for clearer interpretability of Figure 4, which showcases our device's power consumption under various Wi-Fi configurations, we have revised the figure for enhanced readability. Figure 4 illustrates how different operational modes, specifically Wi-Fi configurations, impact the device's energy usage, which is critical for ensuring its effectiveness in continuous monitoring of Parkinson’s Disease symptoms. Each configuration represents a potential use case, from energy-saving modes without Wi-Fi connectivity to higher power modes required for cloud data synchronization. While the term 'nanoparticle ligand' mentioned in the feedback does not directly apply to our discussion on Wi-Fi configurations and power consumption.
  2. We have carefully reviewed and revised title.
  3. Thank you for highlighting the importance of discussing the device's battery life and durability to assess its viability for continuous, reliable monitoring. In response to your valuable feedback, we have added a detailed paragraph within the Sections on wireless connectivity and power consumption directly linking the device's power consumption analysis (as presented in Table 3) to its implications for battery life and durability. This addition aims to provide a clearer exposition of how our design and technology choices ensure the device's operational efficiency and longevity in real-world monitoring scenarios. We believe these enhancements will strengthen the manuscript by making the practical implications of our research more explicit to the reader.
  4. Table 2 is redefined (now includes the information of this work i.e. A-WEAR device
  5. Table 3 is well explained in Section titled ‘Power consumption’
  6. We have revised Figure 8 to improve its readability.
  7. We have revised Figure 9 to improve its readability.
  8. We acknowledge the importance of understanding how extreme environments and special conditions might affect the wristband's data accuracy. The detail has been added in Section 9 to discuss potential deviations in measured data under such conditions and the limitations of our study in this context.

Reviewer 2 Report

Comments and Suggestions for Authors

Thank you very much for submitting your manuscript. We are honored to have the opportunity to review your paper. This article revolves around energy-saving Wi Fi wearable wristbands, known as AWEAR. AWEAR serves as a data collection channel for Parkinson's disease related motion data, securely transmitting the data to the cloud for storage, processing, and severity assessment through customized learning algorithms. The experimental results have demonstrated the adaptability and effectiveness of this method. Although there are still some limitations in this study, it has made a significant contribution to the overall prevention and control pathways of Parkinson's disease. This article emphasizes the latest developments in this field, covering various applications in the field. And call on more people to participate in research in this field. Your paper is excellent in terms of content, innovation, and scientific value. Although there are some formatting and grammar issues, such as some sentences not reading smoothly, these issues have not affected the core points and findings of the paper. Here are some suggestions for improving the manuscript:

1. In the introduction section of this article, there is no clear explanation of the relationship between levodopa and Parkinson's disease.

2. The word "may" was used in line 43 of this article, which is not rigorous enough.

3. In line 98, the Hidden Markov Model should be explained.

4. Line 113 should explain the method of leaving one crossover.

5. The features extracted using Continuous Wavelet Transform (CWT) can effectively distinguish PD symptoms. Please provide a detailed explanation.

6. Please provide examples of the corresponding symptoms for each waveform in Figure 1.

7. In line 239, MT3620 can meet the needs of various IoT applications, and the term "various" is too absolute.

8. This article is overall good, but lacks analysis of extreme environments and special conditions. Will it cause deviation in the measured data of the developed wristband? If so, please provide a brief analysis and overview. If not resolved, please elaborate on the limitations in detail.

Author Response

Thank you for your insightful feedback.

  1. We appreciate your observation regarding the lack of a clear explanation on the relationship between Levodopa and Parkinson's disease (PD) in the introduction. To address this, we have revised the introduction to include a detailed explanation of Levodopa's role in managing PD symptoms, its mechanism of action, and its significance as a cornerstone in PD treatment.
  2. Thank you for pointing out the need for precision in our language. We have carefully reviewed the context of line 43 and have replaced "may" with a more definitive term to enhance the statement's clarity and assertiveness.
  3. Acknowledging the need for clarity on the Hidden Markov Model mentioned in line 98, we have incorporated a brief yet comprehensive description of the model, its relevance to our study, and how it was utilized in analyzing PD symptoms.
  4. We recognize the ambiguity surrounding the "leaving one crossover" method mentioned in line 113. A revision has been made to elaborate on this technique.
  5. Thank you for your comment requesting a detailed explanation of the features extracted using Continuous Wavelet Transform (CWT) and their efficacy in distinguishing Parkinson's Disease (PD) symptoms. We have addressed this in the subsection (6.3) titled "Continuous Wavelet Transform," where we elaborate on the application of CWT in our study and its prior applications in biomedical contexts. This section discusses the advantages of CWT for precise event localization and feature extraction, surpassing traditional Fourier transform methods. We detail our approach to applying CWT to accelerometer signals from PD patients, highlighting how it enables the effective differentiation of PD symptoms such as tremor and bradykinesia through time-frequency domain analysis. The discussion includes the mathematical foundation of CWT, our choice of the Morlet wavelet, and the generation of scalograms that visually represent symptom severity. This comprehensive explanation underscores the robustness of CWT in capturing the nuanced motor symptoms of PD, thereby justifying its selection for our study.
  6. Thank you for your valuable feedback concerning Figure 1. We understand your request for examples of corresponding symptoms for each waveform depicted. However, Figure 1 is intended to illustrate the raw accelerometer signals obtained from a Parkinson's Disease (PD) patient, showcasing data representation along the XYZ axes. This figure primarily serves to demonstrate the nature of the data collected before processing and analysis, rather than directly correlating these waveforms with specific PD symptoms.
  7. We have reconsidered the use of "various" in line 239 and agree that it could convey a sense of absoluteness. The sentence has been modified to specify the types of IoT applications the MT3620 supports, providing clarity and precision.
  8. We acknowledge the importance of understanding how extreme environments and special conditions might affect the wristband's data accuracy. Details are included in Section 9.

Reviewer 3 Report

Comments and Suggestions for Authors

An innovative approach for real-time assessment of the health status of patients with Parkinson's disease is presented in the article. In the technological part, the conceptual implementation includes a smart bracelet with a built-in 3D accelerometer for ambulatory monitoring of patients, allowing prediction of the severity of tremor and bradykinesia during various activities. The data collected by the sensor is transmitted via Wi-Fi to a ServiceNow platform acting as a cloud server. The cloud-based web application stores and processes the data, transforming it into images via Continuous Wavelet Transform. The resulting 2D images (scalograms) are analyzed applying Deep Learning based Model for Classification (AlexNet deep convolutional neural network) for image classification, to assess severity in scores ranging from 0-4. The results are stored in the cloud database and are authorized accessible for the patient and the clinician conducting the study. The authors note that the results obtained for accuracy of 86.5% for bradykinesia and of 90.9% for tremor classification, exceed those of another similar study using CNN classification model and 2D image representation of inertial data. The presented literature review, the synthesis and description of the experimental set-up and the data analysis methodology testify to deep knowledge in the field and competences in conducting experiments. The abstract is well written and consistent with the presented research and results in the article. Reference sources are relevant to the content and cited at appropriate places in the text. The discussion based on statistical indicators correctly interprets the obtained results.

Some comments and remarks:

1.     A more detailed description of the method and criteria for verification of the primary data from the smart bracelet with those received from the Shimmer device is needed.

2.     Ln 501: “Figure??” – missing number.

3.     Ln 571: The reference is not properly cited as "similar" because the preceding sentence analyzes the same source.

4.     What is the cost to use cloud services?

Author Response

Thank you for your insightful feedback.

  1. Regarding the need for a more detailed description of the method and criteria for verification of the primary data collected from the smart bracelet in comparison with the data received from the Shimmer device. We have revisited our methodology section 6.1 to provide a clearer and more comprehensive explanation of the existing verification process. We believe these enhancements will satisfactorily address your concerns.
  2. Ln 501: “Figure??” – missing number is corrected.
  3. Ln 571: The reference is now properly cited, thank you.
  4. The cloud subscription would cost around 100-200$ per month and might vary depending on the number of end-users.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

the authors address the comments properly, so the manuscript is recommended to be accepted for publication.

Reviewer 2 Report

Comments and Suggestions for Authors

none