Hybrid Quantum–Classical Architectures in Medical Imaging: A Taxonomy-Based Survey of COVID-19 Models

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper surveys the emerging field of hybrid quantum–classical (HQC) architectures for medical image analysis, with a particular focus on COVID-19 classification as a representative application.

The paper provides a comprehensive review for the existing research on this topic as well as the recent progress. I should recommend it for publication after addressing the following questions or concerns.

The several sections proceeding the abstract seem not most typical format for a research article. However, if it suits the journal, its clearance needs to be improved, such as the vagueness for terms, such as “translation limits”, “empirical performance pattern”.

The authors mentioned the issue of “fragmented and difficult to compare” (Sec. 2.5). How this problem can be resolved or mitigated given it is caused by small or biased data base?

Regarding the quantum formulation (Sec. 3), I wonder whether the quantum state (e.g., Eq. 7) is the entrance for the clinic data come into play? In this case, what is the typical dimensions of Hilbert space and whether is it necessary to solve the time-dependent quantum evolution? This is typically computationally expensive. If will it cause difficulty in practice?

In Table 3, the authors provide the qubit requirements and circuit depths for different methods. What does “shallow” or “small” mean precisely in this context? The authors could also comment on the practical feasibility of these methods, for instance, qubits ~ 4.

Finally, the authors are encouraged to discuss possible extensions of the present method beyond COVID-19, for example in Sec. 7.

 

 

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript surveyed hybrid quantum-classical architectures for COVID-19 medical imaging and organized existing models according to the functional placement of the quantum module, including front-end/quanvolutional processing, quantum feature extraction, and quantum classification or decision modules. The architecture-centered taxonomy was a useful contribution. The manuscript was also appropriately cautious about NISQ constraints, simulation-based evaluation, limited hardware validation, and the lack of clinical validation.

However, several important issues were identified and should be addressed.

1. The review methodology was not sufficiently transparent. The manuscript used the term “systematic review” in the keywords and repeatedly presented the work as a systematic or taxonomy-driven synthesis. However, the search strategy was not described in sufficient detail. The authors were advised to provide the databases searched, search strings, time window, inclusion and exclusion criteria, screening process, final number of included studies, and whether PRISMA or a similar protocol was followed. If the manuscript was intended as a narrative or taxonomy-based survey rather than a systematic review, the terminology should be revised accordingly.

2. The taxonomy assignment criteria should be made more explicit. The proposed classification into quanvolution/front-end processing, quantum feature extraction, and quantum classification or decision modules was useful. However, the criteria for assigning studies to these categories remained mostly descriptive. The authors were advised to clarify how mixed architectures were handled, whether classification was performed independently by more than one reviewer, and how disagreements were resolved.

3. The performance comparison was limited by heterogeneity across datasets and protocols. The surveyed studies differed in modality, datasets, binary versus multi-class tasks, split strategies, validation protocols, simulator or hardware use, and baseline models. The comparative tables should more explicitly list these conditions, including dataset source, modality, task formulation, patient-wise versus image/slice-wise split, external validation, simulator or hardware backend, and baseline model. Without this information, it was difficult to judge whether observed performance differences were attributable to the quantum component.

4. The quality assessment of medical imaging evidence should be strengthened. For COVID-19 image classification, accuracy alone was not sufficient. The authors were advised to place greater emphasis on patient-level splitting, data leakage risk, class imbalance, external validation, sensitivity, specificity, AUROC, and confidence intervals. Slice-level random splitting, for example, may have inflated performance if images from the same patient appeared in both training and testing sets.

5. Quantum resource reporting should be more complete and consistent. Since this was a review of hybrid quantum-classical architectures, conventional performance metrics alone were not sufficient to assess feasibility. Where available, the authors should add a unified summary of quantum-resource information, including the number of qubits, circuit depth, encoding strategy, ansatz or gate structure, number of shots, simulator or hardware backend, noise model, error-mitigation strategy, number of trainable quantum parameters, and quantum calls per image. These details were important for evaluating whether the reported methods were practically feasible under NISQ constraints and whether the claimed advantages could be reproduced beyond idealized simulation settings.

6.  Table 1 was very wide and difficult to read. The authors may consider splitting or simplifying it so that the main comparison points are clearer.

7. If arXiv or preprint references were retained, their formatting should be made consistent, and version or access information should be added where appropriate.

8. As mentioned before, there were several issues regarding the references.

Comments on the Quality of English Language

The English was generally understandable, but several technical descriptions could be improved for clarity and precision. Some equations and symbols were not rendered clearly, several tables were difficult to read, and some reference entries required formatting corrections. Moderate language editing and formatting revision were recommended.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript presents a comprehensive survey and taxonomy of Hybrid Quantum–Classical (HQC) architectures for medical image analysis. The study categorizes HQC pipelines into three architectural archetypes: Quanvolution-based hybrid architectures, Classical feature extractors with quantum classifier heads and Quantum feature extractors with classical classifiers. The topic is timely and relevant given the increasing interest in quantum machine learning for healthcare applications. However, the manuscript requires substantial revisions regarding clarity, methodological rigor, reproducibility discussion before it can be considered for publication.

 

• Many variables and mathematical symbols are introduced without sufficiently clear definitions. The manuscript should ensure that all notation is clearly explained when first introduced to improve readability and facilitate understanding of the proposed framework.
• The authors should more clearly articulate the novel analytical contribution of the proposed taxonomy, specifically:
• What new insights the taxonomy provides to the HQC research community
• How it differs from and advances beyond existing survey and review papers
• The manuscript would benefit from the inclusion of a comprehensive comparative analysis table summarizing key characteristics of the reviewed studies, including:
• Computational complexity
• Number of qubits utilized
• Quantum encoding strategies
• Dataset sizes and modalities
• Reported performance accuracies
• Scalability and hardware limitations

Such a comparison would significantly improve the practical value and readability of the survey.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors revised the manuscript carefully and addressed most of the major concerns listed in the previous review. The revised version provided a clearer review methodology, a more explicit taxonomy, and more complete discussion of validation protocols, dataset heterogeneity, and quantum-resource reporting. Overall, the manuscript was much improved and was close to being acceptable. Several minor revisions are suggested before publication.

1. Some tables are still dense and may be difficult to read. The authors may further check the formatting and make sure the tables are clear in the final published layout.
2. Some long paragraphs could be shortened to improve readability, especially in the sections discussing validation rigor and deployment feasibility.
3. The authors should make sure that all reference entries are consistently formatted and that all in-text citations match the final reference list.

In general, the authors responded well to the previous comments, and the revised manuscript provided a clearer and more useful survey.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The author did all the required comments

Author Response

We sincerely thank Reviewer 3 for the positive evaluation and valuable feedback throughout the review process.