MCI Conversion Prediction Using 3D Zernike Moments and the Improved Dynamic Particle Swarm Optimization Algorithm
, Mohammadreza Gholami 1, Masoud Moradi 2, Iman Beheshti 3,* and Hasan Demirel 2,†
on behalf of the Alzheimer’s Disease Neuroimaging Initiative
Round 1
Reviewer 1 Report
Table 1: averages and SDs/SEMs are displayed. I am unsure if it is SDs/SEMs can the authors please make this clear. There is also a huge age difference between males and females and for this reason averages may be needed for each group, and not just an overall figure summarising both genders.
Equations from page 5 onwards needs to be more clear, it currently appears blurry.
Authors should perhaps think of adding some recent patient data to their model and compare patients over time.
Author Response
Responses to comments from Reviewer #1
Response: · We would like to thank the reviewer’s insightful comments. We deeply appreciate the time and efforts you spent on our manuscript.
Table 1: averages and SDs/SEMs are displayed. I am unsure if it is SDs/SEMs can the authors please make this clear. There is also a huge age difference between males and females and for this reason averages may be needed for each group, and not just an overall figure summarizing both genders.
Response:
- Thanks to the reviewer for picking up this point. In the Table 1, we have clarified that All data are presented in mean ± standard deviation mode. Please note that we identified the MCI patients with available baseline MRI scan and at least three years follow up. The MCI groups were imbalance in term of gender. However, there was no significant difference between males and females in both groups in terms of age, disease severity.
Equations from page 5 onwards needs to be more clear, it currently appears blurry.
Response:
- All equations and variables in the manuscript are updated using mathtype.
Authors should perhaps think of adding some recent patient data to their model and compare patients over time.
Response:
- It is an excellent suggestion to keep track of whether patients belong to the same class throughout time. However, the main objective of this research study was to develop a new pattern recognition method for classifying sMCI patients from pMCI patients on the basis of the baseline MRI data only, when both MCI groups have similar clinical patterns.
Reviewer 2 Report
The article presents a new framework for predicting Mild Cognitive Impairment (MCI) conversion to Alzheimer's disease (AD) using 3D-Zernike Moment (3D-ZM) and Improved Dynamic Particle Swarm Optimization (IDPSO). The proposed framework extracts statistical features from 3D-MRI scans and uses IDPSO to find informative features for MCI conversion prediction. The study was conducted on a large sample of MCI patients, and the experimental results indicate that the proposed method has a strong ability to distinguish progressive-MCI (pMCI) patients from stable-MCI (sMCI) patients, with an accuracy of more than 75%. The article concludes that the proposed framework can be useful in determining possible treatment trajectories for AD patients.
Overall, the article presents a novel approach to predict MCI conversion using 3D-ZM and IDPSO, and the experimental results suggest that the proposed framework is effective in identifying patients who are at risk for MCI conversion. It is a well-written and well-argued piece of research that makes a significant contribution to the field of MCI and AD research. However, it is important to note that the study was conducted on a specific dataset (ADNI), and further validation on other datasets is needed to confirm the generalizability of the proposed framework.
Author Response
Responses to comments from Reviewer #2
The article presents a new framework for predicting Mild Cognitive Impairment (MCI) conversion to Alzheimer's disease (AD) using 3D-Zernike Moment (3D-ZM) and Improved Dynamic Particle Swarm Optimization (IDPSO). The proposed framework extracts statistical features from 3D-MRI scans and uses IDPSO to find informative features for MCI conversion prediction. The study was conducted on a large sample of MCI patients, and the experimental results indicate that the proposed method has a strong ability to distinguish progressive-MCI (pMCI) patients from stable-MCI (sMCI) patients, with an accuracy of more than 75%. The article concludes that the proposed framework can be useful in determining possible treatment trajectories for AD patients.
Overall, the article presents a novel approach to predict MCI conversion using 3D-ZM and IDPSO, and the experimental results suggest that the proposed framework is effective in identifying patients who are at risk for MCI conversion. It is a well-written and well-argued piece of research that makes a significant contribution to the field of MCI and AD research. However, it is important to note that the study was conducted on a specific dataset (ADNI), and further validation on other datasets is needed to confirm the generalizability of the proposed framework.
Response: · We would like to thank the reviewer’s insightful comments. We deeply appreciate the time and efforts you spent on our manuscript. We have added the following passage into the discussion section:“This study was conducted based on the ADNI dataset, which was created using highly standardized protocols to distinguish individuals at different AD stages. For our MCI conversion prediction framework to be adapted in clinical settings, further validation using other datasets with different protocols is needed.”
Reviewer 3 Report
In this manuscript, the Beheshti group aims to develop an MCI conversion prediction framework based on 3D-ZM that generates features from 3D-MRI scans which can be analyzed and filtered by an IDPSO algorithm for the identification of MCI patients who are at risk of developing dementia due to AD. Their model achieved an accuracy exceeding 75% in the test run, comparable to some of the current recognition methods. In general, the manuscript is clearly written. The methodology was adequately explained, and the test showed positive results. My specific comments are as follow.
1. The final accuracy of the method introduced in the manuscript reached 75.66%. However, according to the literature, other independent methods also achieved similar or higher accuracy in pMCI to sMCI prediction using sMRI DATA. What is the advantage of the current method over the other published ones?
2. Based on Figure 1, the authors used the same dataset for training as well as final validation. Have the authors use this method to analyze an independent set of MCI patient data to confirm the accuracy of prediction?
Author Response
Responses to comments from Reviewer #3
In this manuscript, the Beheshti group aims to develop an MCI conversion prediction framework based on 3D-ZM that generates features from 3D-MRI scans which can be analyzed and filtered by an IDPSO algorithm for the identification of MCI patients who are at risk of developing dementia due to AD. Their model achieved an accuracy exceeding 75% in the test run, comparable to some of the current recognition methods. In general, the manuscript is clearly written. The methodology was adequately explained, and the test showed positive results. My specific comments are as follow.
Response: · We would like to thank the reviewer’s insightful comments. We deeply appreciate the time and efforts you spent on our manuscript.
- The final accuracy of the method introduced in the manuscript reached 75.66%. However, according to the literature, other independent methods also achieved similar or higher accuracy in pMCI to sMCI prediction using sMRI DATA. What is the advantage of the current method over the other published ones?
Response:
- The reviewer is quite correct that the state-of-the-art MCI conversion prediction approaches have reported similar prediction accuracies. In this study, we proposed alternative and new approach based on based on 3D-ZM that generates statistical features from 3D-MRI scans and IDPSO that finds an informative sub-set of Zernike features for MCI conversion prediction. Our accuracy of 75.66% for identifying pMCI patients from sMCI patients is well comparable with the with state-of-the-art approaches using baseline MRI data (Table 4).
- Based on Figure 1, the authors used the same dataset for training as well as final validation. Have the authors use this method to analyze an independent set of MCI patient data to confirm the accuracy of prediction?
Response:
- Please note that we used a standard validation strategy (k-fold cross validation), in which reduce the variance of the performance estimate and avoid of overfitting, to assess the reliability of proposed method. In each iteration, the training set was completely independent from the respective test set and feature selection (with IDPSO) was performed only based on the training set.
