Classification of Photoplethysmographic Signal Quality with Fuzzy Neural Network for Improvement of Stroke Volume Measurement

Round 1

Reviewer 1 Report

Authors present a fuzzy neural network based approach for classification of Photoplethysmographic Signal Quality. In addition, they present certain preprocessing and segmentation approaches to enhance the accuracy further. However, I would recommend authors to adopt to certain clustering and autoencoder based approaches which have proven to be successful for such applications. Authors could mention on how they could utilize such methods for their research and how much it affects the accuracy of their classification model.

- Kebede, T. M., Djaneye-Boundjou, O., Narayanan, B. N., Ralescu, A., & Kapp, D. (2017, June). Classification of malware programs using autoencoders based deep learning architecture and its application to the microsoft malware classification challenge (big 2015) dataset. In 2017 IEEE National Aerospace and Electronics Conference (NAECON) (pp. 70-75). IEEE.

- Narayanan, B. N., Hardie, R. C., Kebede, T. M., & Sprague, M. J. (2019). Optimized feature selection-based clustering approach for computer-aided detection of lung nodules in different modalities. Pattern Analysis and Applications, 22(2), 559-571.

Author Response

To Reviewer #1:

Thank the first reviewer for his/her valuable comments that make better this manuscript. The texts in this revised manuscript have been corrected/ modified by red words.

Comments and Suggestions for Authors

Authors present a fuzzy neural network based approach for classification of Photoplethysmographic Signal Quality. In addition, they present certain preprocessing and segmentation approaches to enhance the accuracy further. However, I would recommend authors to adopt to certain clustering and autoencoder based approaches which have proven to be successful for such applications. Authors could mention on how they could utilize such methods for their research and how much it affects the accuracy of their classification model.

Ans: We have added a paragraph in Discussion to describe the performance of the autoencorder for our system.

7. Discussion

Kebede et al. [21] used a multilayer neural network with an autoencoder to boot the accuracy of the classification of malware programs. They transferred the malware programs to the images, and their results showed that the autoencoder could increase the performance of the multilayer neural network for the classification. Although the proposed SoNFIN belongs to one of multilayer neural networks, it employs the neural network to do the fuzzy inference. Thus, it is not suitable to implement an autoencoder in the TSK model. But, if we consider each PPG pulse and its differential signal as an image, a deep learning architecture will be used to do the classification of SQI. However, the autoencoder and the feature selection-based clustering approach could be utilized in the deep learning architecture to enhance its performance of classification [22].

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper discuss about using photoplethysmographic signals together with fuzzy neural networks to estimate SV accurately.

The paper has methodically presented the proposed hardware and the mathematics/algorithms behind their work. It has compared the SV measurements of the proposed device and an industry standard device. The results are acceptable within a permissible margin of error.

However, in the conclusion the authors have not quantitatively summarized the comparison results before making a conclusion(it is there in discussion, but rather than saying relatively small error, specify the error range/percentage). This could be included in the revised version.

Author Response

To Reviewer #2:

Thank the second reviewer for his/her valuable comments that make better this manuscript. The texts in this revised manuscript have been corrected/ modified by red words.

The paper has methodically presented the proposed hardware and the mathematics/algorithms behind their work. It has compared the SV measurements of the proposed device and an industry standard device. The results are acceptable within a permissible margin of error.

However, in the conclusion the authors have not quantitatively summarized the comparison results before making a conclusion(it is there in discussion, but rather than saying relatively small error, specify the error range/percentage). This could be included in the revised version.

Ans: We have added the error percentage to describe how to quantify the level of SQI and results of classification in Discussion and Conclusions.

7. Discussion

The main peak of the pulse and the dicrotic notch on the down slope of each cycle are the two primary characteristics for the contour of PPG pulse. However, when PPG pulses are used to measure the heart rate and oxygen saturation, previous studies always focus on whether the main peak of the PPG pulse is clear and the baseline is drifted or not [16, 14, 10]. The reason is that the amplitude of pulsatile waveform and the baseline are used to measure the peripheral oxygen saturation [3]. Thus, they used the rule-based algorithms to determine the SQI of PPG pulse. The quality level of PPG pulse was defined by experts with a manual fashion. However, in the validation of their algorithms, a direct comparison of performance between two published algorithms is restricted because the cognitive abilities of experts are different. Thus, what the main characteristics of PPG pulse for the high SQI depends on what the measuring physiological parameter are. In our study, we adopt a kind of expert system, SoNFIN, to evaluate the quality level of individual PPG pulse, and the SVs measured by our ICG device are compared with those (as reference) by the medis® CS2000. We use three levels of the error percentage (that is, less than 40%, between 40% and 45%, and larger than 45%) to define each PPG pulse belonging to high, middle, or low SQI categorization. This approach can readily quantify the quality level of individual PPG pulse. The current results show that the proposed algorithm can effectively pick up PPG pulses belonging to high SQI, and increase the accuracy in the SV measurement of ICG.

7. Conclusions

A rule-based algorithm combined with a SoNFIN was developed in the study, and it can be applied to successfully determine the SQI of each PPG pulse. In order to quantify the level of SQI for each PPG pulse, the error percentage of measured SV for each heartbeat was used to define the level of SQI for each PPG pulse. The PPG pulse with high SQI was used to measure SV in impedance plethysmography. The validation of our proposed algorithm was performed by comparing the errors of SVs measured by our proposed device and medis © CS2000. It was found that when PPG pulses with high SQI were adopted to measure SVs, their error percentage would be below 40% and the statistic error would be decreased from -18 ± 22.0 ml to 6.4 ±12.8 ml. The results suggest that the proposed algorithm incorporating the SoNFIN could be employed to find out the SQI of a PPG pulse. The SV measurement with the pulse with high SQI would increase the accuracy in the ICG applications.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

They have answered my questions. Paper is ready to be published now.