Verified Rigor RecordRigor Reviewed

Rigor-Reviewed Manuscript

Archimedes optimisation algorithm quantum dilated convolutional neural network for road extraction in remote sensing images · 10.1016/j.heliyon.2024.e26589

Name: Verified Rigor Record — Archimedes optimisation algorithm quantum dilated convolutional neural network for road extraction in remote sensing images · 10.1016/j.heliyon.2024.e26589
Item: Archimedes optimisation algorithm quantum dilated convolutional neural network for road extraction in remote sensing images · 10.1016/j.heliyon.2024.e26589
Rating: 73
Author: Αlpha¹

Paper source

Read full text

Αlpha¹ Editorial

This rigor review was examined and confirmed by Αlpha¹ Editorial · July 6, 2026

Αlpha¹ Rigor Review

Rigor Reviewed

WARN

Warnings to address · Computational Study

Eight dimensions — click a segment for details

Critical

High-priority

Model-reported confidence: moderateReviewer agreement: lowEngine: 7.6.0Source: full text

Not verified in this report

No reported statistical tests were found to recompute.

This Αlpha¹ Rigor Review uses AI Rigor Reviewers trained on a curated corpus of high-fidelity and retracted papers, with expert supervision and curation. It can still make mistakes; verify each finding against the source before relying on it.

The paper presents a novel deep learning method for road extraction with a clear motivation and baseline comparisons, but has significant gaps in reproducibility-critical reporting: no software environment identified, no code shared, and no inferential statistics. Copyedit issues (broken table headers, grammar errors) further reduce polish.

Three independent reviewers evaluated the full manuscript. The study is purely computational, so biological variables, ethical approvals, and statistical analysis (no inferential tests) are not applicable. The key reviewer divergence was on study design (one pass, two warn) and key resources (one warn, two fail); the synthesized status resolves these by weighing the reproducibility gap.

10 major claims checked against the paper's own evidence: 1 not fully backed by the presented evidence (unsupported or overstated).

6 integrity concerns flagged (0 high).

Rigor

Eight Dimensions2 pass · 2 warn · 1 fail

10 major claims1 unsupported4 partial5 supported

Prior work is cited, including Shao et al., RADANet, Li et al., and others, establishing the challenges in road extraction. The rationale links the limitations of existing CNNs in capturing fine-grained details and contextual information to the proposed quantum dilated convolution approach. The paper addresses limitations of prior work by stating that traditional CNNs struggle with these aspects, and the proposed method is designed to overcome them.

Sub-criteria

prior work citedADEQUATE

premise rationaleADEQUATE

limitations addressedADEQUATE

Per reviewer (3) — the votes behind the verdict

PASSReviewer 1· 85% conf

The premise is adequately established: prior work is cited, the rationale linking quantum dilated convolution and AOA tuning to improved road extraction is logical, and limitations of prior work are acknowledged.

Evidence

paraphrase[Section 1 (Introduction), paragraph 10]

“The authors review several prior methods (Shao et al., RADANet, Li et al., Yan et al., Surendran et al., etc.) and note that 'Conventional convolutional neural networks (CNNs) achieve remarkable results in road extraction tasks. The research gap in the existing works are often struggle with capturing fine-grained details and contextual information.'”

direct quote[Section 1 (Introduction), paragraph 11]

“The AOA-QDCNNRE technique is mainly based on the QDCNN model, which integrates the concept of quantum computing (QC) with dilated convolutions for enhancing the network's capability to capture either local or global contextual data.”

PASSReviewer 2· 85% conf

The paper adequately establishes the scientific premise by citing prior work on road extraction using CNNs, identifying a research gap in capturing fine-grained details and contextual information, and proposing a novel method to address this gap.

Evidence

direct quote[Section 1, paragraph 3-4]

“Shao et al. [] created a road extraction network that integrates an embedded attention system to tackle the task of extracting road networks from a large amount of remote sensing imagery.”

direct quote[Abstract, paragraph 1]

“Traditional convolutional neural networks (CNNs) demonstrate impressive performance in road extract tasks; however, they frequently encounter difficulties in capturing intricate details and contextual information.”

direct quote[Section 4, paragraph 1]

“The limitation of the AOA-QDCNNRE approach is only tested in simulation experiments.”

PASSReviewer 3· 90% conf

The paper establishes a clear scientific premise by citing prior work and providing a logical rationale for the proposed method.

Evidence

direct quote[Introduction, paragraphs 1-3]

“Shao et al. [] created a road extraction network that integrates an embedded attention system...”

direct quote[Introduction, paragraph 4]

direct quote[Abstract]

The paper uses benchmark datasets and does not discuss sample size justification (power analysis) or data curation (inclusion/exclusion). Controls are adequately addressed by comparing with multiple baseline methods (CNN, U-Net, GL-Dense-U-Net, RDRCNN). Independent replication is partially supported by evaluation on two datasets (Massachusetts Road and GF-2 Road). Randomization, blinding, and other bench-science criteria are not applicable.

Sub-criteria

controlsADEQUATE

power analysisNOT REPORTED

blinding levelsN/A

outlier handlingN/A

randomization unitN/A

inclusion exclusionNOT REPORTED

randomization methodN/A

replicate distinctionN/A

independent replicationADEQUATE

Per reviewer (3) — the votes behind the verdict

WARNReviewer 1· 80% conf

The study design is computational and lacks several design elements expected in a rigorous computational study: no data curation or inclusion/exclusion criteria, no description of how the training/validation/test split was performed, no mention of multiple runs or cross-validation, and no power analysis or dataset size rationale.

Evidence

direct quote[Section 3 (Results and discussion), paragraph 1]

“the AOA-QDCNNRE method with other DL models on the Massachusetts Road database”

absence[Section 3 (Results and discussion), paragraph 1]

No data split or curation criteria reported.

absence[Section 3 (Results and discussion)]

No independent replication or multiple runs reported.

WARNReviewer 2· 85% conf

The study design is computational; many standard experimental rigor criteria are not applicable, but among those that are applicable (power analysis, controls, inclusion/exclusion, independent replication), only controls and independent replication are reported adequately.

Evidence

absence[Section 3, paragraph 1]

No sample size justification or power analysis is provided.

direct quote[Section 3, paragraph 2]

“The result stated the improved road extraction outcomes of the AOA-QDCNNRE technique.”

direct quote[Section 3, paragraph 1]

“the road extraction outcome of the AOA-QDCNNRE method is executed on 2 datasets, namely, Massachusetts Road and GF-2 Road Dataset.”

PASSReviewer 3· 80% conf

The study design is appropriate for a computational paper, with adequate use of benchmark datasets and baseline comparisons.

Evidence

direct quote[Section 3, Results and discussion]

“the AOA-QDCNNRE method is executed on 2 datasets, namely, Massachusetts Road [] and GF-2 Road [] Dataset.”

paraphrase[Table 1 and Table 2]

“Tables 1 and 2 compare the proposed method against CNN, U-Net, GL-Dense-U-Net, RDRCNN, and RDRCNN+post-process.”

The study uses remote sensing images (Massachusetts Road and GF-2 Road datasets) for road extraction. No human or animal subjects, cell lines, or laboratory organisms are involved. Therefore, all biological variable criteria are not applicable.

Sub-criteria

demographicsN/A

sex reportedN/A

sex justifiedN/A

age weight healthN/A

housing conditionsN/A

species strain sourceN/A

Per reviewer (3) — the votes behind the verdict

N/AReviewer 1· 100% conf

The study is purely computational and uses remote sensing images; no biological variables are relevant.

Evidence

paraphrase[Abstract and Section 1]

“The paper describes road extraction from remote sensing images.”

N/AReviewer 2· 100% conf

The study uses remote sensing images, not biological subjects, so no biological variables are reported.

Evidence

direct quote[Abstract, paragraph 1]

“Archimedes optimisation algorithm quantum dilated convolutional neural network for road extraction in remote sensing images”

N/AReviewer 3· 100% conf

No biological variables are relevant as the study is purely computational.

Evidence

direct quote[Ethics declarations]

“This article does not contain any studies with human participants or animals performed by any of the authors.”

The paper explicitly states 'This article does not contain any studies with human participants or animals performed by any of the authors.' It uses two benchmark datasets: Massachusetts Road Dataset (available on Kaggle and University of Toronto websites) and GF-2 Road Dataset (available on Paperswithcode). These are public, de-identified datasets. There is no new data collection or use of non-public subject-level data. Therefore, ethics approvals are genuinely not applicable.

Sub-criteria

iacuc statementN/A

informed consentN/A

irb ethics statementN/A

regulatory complianceN/A

Per reviewer (3) — the votes behind the verdict

N/AReviewer 1· 100% conf

The study uses only public, fully de-identified remote sensing datasets and no human or animal subjects, tissue, or non-public identifiable data; ethics approval is genuinely not applicable.

Evidence

direct quote[Ethics declarations]

“This article does not contain any studies with human participants or animals performed by any of the authors.”

direct quote[Data availability statement]

“Datasets are available in the following open access datasets. https://www.kaggle.com/datasets/balraj98/massachusetts-buildings-dataset ... https://paperswithcode.com/dataset/gid”

N/AReviewer 2· 100% conf

The study uses only public benchmark datasets and states it does not involve human participants or animals, so ethical approvals are not applicable.

Evidence

direct quote[Ethics declarations]

“This article does not contain any studies with human participants or animals performed by any of the authors.”

N/AReviewer 3· 100% conf

Ethical approvals are not required as the study uses only publicly available, de-identified benchmark datasets.

Evidence

direct quote[Ethics declarations]

“This article does not contain any studies with human participants or animals performed by any of the authors.”

The study is computational and uses bespoke software (the AOA-QDCNNRE model), but no software names, versions, or URLs are provided. The data availability statement gives dataset URLs, but no code repository is mentioned. All other sub-criteria (antibodies, cell lines, etc.) are not applicable.

Sub-criteria

mycoplasma testingN/A

reagents identifiedN/A

organisms identifiedN/A

antibodies identifiedN/A

cell line authenticationN/A

software tools identifiedNOT REPORTED

Per reviewer (3) — the votes behind the verdict

WARNReviewer 1· 85% conf

The study uses public datasets and reports the software framework (QDCNN, AOA) but does not identify software with version numbers or provide access to custom code, and no other biological/chemical resources are used.

Evidence

absence[Section 2 and Data availability statement]

Software is described but not versioned and no code is shared.

direct quote[Data availability statement]

“https://www.kaggle.com/datasets/balraj98/massachusetts-buildings-dataset ... https://paperswithcode.com/dataset/gid”

FAILReviewer 2· 90% conf

The paper does not identify any software tools, libraries, or versions used for implementation, and no custom code is shared, so the only applicable dimension (software_tools_identified) is not reported.

Evidence

absence[Section 2, paragraph 1]

No software tools or versions are identified.

absence[Data availability statement]

No code repository is provided.

FAILReviewer 3· 90% conf

The paper fails to identify the software framework and versions used for implementation, which is a critical omission for a computational study.

Evidence

absence[Entire paper]

No software framework is identified.

No p-values, confidence intervals, or statistical tests are reported. The paper simply compares performance metrics across methods. All sub-criteria related to inferential statistics are not applicable.

Sub-criteria

tests namedN/A

exact p valuesN/A

effect sizes ciN/A

data presentationINADEQUATE

software identifiedN/A

assumptions verifiedN/A

mathematical plausibilityN/A

Per reviewer (3) — the votes behind the verdict

WARNReviewer 1· 85% conf

The paper reports classification metrics (accuracy, precision, recall, F1, IoU) for model comparison, but does not name any statistical tests, verify assumptions, report exact p-values or effect sizes/CI, and the software for statistical analysis is not identified. Data presentation uses tables and bar charts with per-group values, adequate for the context.

Evidence

absence[Section 3 (Results and discussion)]

No statistical tests are reported for model comparisons.

absence[Section 3 (Results and discussion)]

Software for statistical analysis is not identified.

paraphrase[Section 3 (Results and discussion), Tables 1-3]

“Tables 1 and 2 present accuracy, precision, recall, F1 score, and IoU for each model on both datasets.”

N/AReviewer 2· 95% conf

The paper reports only descriptive statistics (accuracy, precision, etc.) without any inferential statistical tests, so the dimension is not applicable.

Evidence

paraphrase[Section 3, Tables 1 and 2]

“The tables show accuracy, precision, recall, F1 score, and IoU for each method, but no statistical tests comparing them.”

N/AReviewer 3· 90% conf

The paper does not report inferential statistics; it only presents descriptive performance metrics, so statistical analysis is not applicable.

Evidence

absence[Section 3, Results and discussion]

No inferential statistics are reported.

The paper provides a data availability statement with URLs to the Massachusetts Road and GF-2 Road datasets in public repositories. However, no code for the proposed AOA-QDCNNRE method is provided, and no software environment details are given. This limits reproducibility.

Sub-criteria

code sharingNOT REPORTED

accession numbersN/A

repository depositADEQUATE

data availability statementADEQUATE

Per reviewer (3) — the votes behind the verdict

PASSReviewer 1· 85% conf

A data availability statement is present with concrete URLs for both datasets. Code sharing is not applicable since the paper describes a custom method but provides no code repository; however, the paper is a methods paper describing the algorithm, and code sharing is commendable but not universally required. The data availability statement is adequate, and code sharing is missing but not required for this study type.

Evidence

direct quote[Data availability statement]

“https://www.kaggle.com/datasets/balraj98/massachusetts-buildings-dataset ... https://paperswithcode.com/dataset/gid”

absence[Data availability statement and manuscript]

No custom code repository provided.

WARNReviewer 2· 85% conf

The data availability statement provides concrete URLs to public datasets, but no code repository or persistent identifiers are provided for the datasets, and the code for the proposed method is not shared.

Evidence

direct quote[Data availability statement]

“Datasets are available in the following open access datasets. https://www.kaggle.com/datasets/balraj98/massachusetts-buildings-dataset https://www.cs.toronto.edu/∼vmnih/data/mass_roads/train/sat/index.html https://paperswithcode.com/dataset/gid”

absence[Data availability statement]

No code repository is provided.

WARNReviewer 3· 90% conf

Data availability is adequately stated with URLs to public datasets, but custom code is not shared, which is a common practice in computational studies.

Evidence

direct quote[Data availability statement]

absence[Entire paper]

No code is provided.

The methods are described in detail with equations. All reported outcomes (metrics across datasets) are presented. Funding and conflict of interest statements are provided. Limitations are mentioned (simulation-only testing), but could be more thorough. Conclusions are proportional to the results.

Sub-criteria

funding coiADEQUATE

trial registrationN/A

reporting guidelineN/A

methods completenessADEQUATE

all outcomes reportedADEQUATE

limitations discussedINADEQUATE

conclusions proportionalADEQUATE

Per reviewer (3) — the votes behind the verdict

WARNReviewer 1· 85% conf

Methods are described in sufficient detail for replication, limitations are briefly discussed, conclusions are proportional to the evidence, and funding/COI statements are present. However, no reporting guideline is referenced (not common for computational studies), and not all expected outcomes are pre-specified. The study is not a clinical trial, so trial registration is not applicable.

Evidence

direct quote[Section 2 (The proposed model)]

“The QDCNN model presents a road extraction methodology that incorporates a quantum layer, convolutional layer, and dilation layer in a three-tier-based framework.”

direct quote[Section 4 (Conclusion)]

“The limitation of the AOA-QDCNNRE approach is only tested in simulation experiments.”

absence[Full manuscript]

No reporting guideline referenced.

PASSReviewer 2· 85% conf

The paper reports methods in sufficient detail, discusses limitations, declares funding and conflicts of interest, and the conclusions are proportional to the evidence.

Evidence

direct quote[Section 2, paragraphs 2-7]

“The QDCNN model presents a road extraction methodology that incorporates a quantum layer, convolutional layer, and dilation layer in a three-tier-based framework.”

direct quote[Section 4, paragraph 1]

“The limitation of the AOA-QDCNNRE approach is only tested in simulation experiments.”

direct quote[Funding, Declaration of competing interest]

“The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.”

PASSReviewer 3· 90% conf

The paper is generally transparent, with detailed methods, complete outcome reporting, and funding/COI statements, though limitations are only briefly discussed.

Evidence

direct quote[Section 2, The proposed model]

“The QDCNN model presents a road extraction methodology that incorporates a quantum layer, convolutional layer, and dilation layer in a three-tier-based framework.”

paraphrase[Tables 1, 2, and 3]

“Tables 1 and 2 report accuracy, precision, recall, F1-score, IoU for both datasets. Table 3 reports computation time.”

direct quote[Section 4, Conclusion]

“The limitation of the AOA-QDCNNRE approach is only tested in simulation experiments.”

26 copyedit issues flagged (2 major): mostly typo, grammar, consistency.

Checked 36 references: 7 verified — 1 retracted, 9 unresolved, 19 not checked.

Reproducibility

4 data/code links checked; 2 live, 2 dead.

Αlpha¹ Rigor Action ItemsFinal pass before journal submission

The published paper has notable rigor and reporting weaknesses that an informed reader should weigh carefully. The most consequential issues are the lack of software/code sharing (key resource fail) and missing training details (data split, replication strategy), which severely limit reproducibility. The paper would warrant an erratum or correction to add these missing details; until then, its conclusions should be interpreted with caution.

1.
HIGHrigor
Add an explicit justification or remove the claim that AOA-based hyperparameter tuning improves road extraction outcomes; the paper provides no experiment comparing QDCNN with vs. without AOA. Add an ablation study or remove the unsupported claim.
The claim audit found this claim to be 'unsupported'; a headline conclusion without supporting evidence is a validity threat that could lead to retraction.
2.
HIGHdata code
Provide custom code (AOA-QDCNNRE) in a public repository (GitHub, Zenodo) with a DOI, and add the link to the Data Availability Statement.
The paper fails 'key_resources' because no software/code is shared; this is the most critical reproducibility gap for a computational methods paper.
3.
HIGHother
Cite a retracted paper (HQDCNet: hybrid quantum dilated convolution neural network for detecting covid-19... doi:10.1007/s11042-023-15515-6). Remove or replace this reference with a valid alternative, and verify it is not used to support the method.
Citing a retracted paper is a serious academic integrity issue that must be corrected.
4.
HIGHother
Check the 9 references flagged as 'not_found_in_registry' (e.g., 'NIGAN: a framework for mountain road extraction...', 'Mixer U-Net...'). Verify each exists as a legitimate publication or replace them.
Unresolvable references are a fabrication signal; they must be verified to maintain the integrity of the literature review.
5.
HIGHdata code
Correct the data availability URL: the Kaggle link points to the 'Massachusetts Buildings Dataset', not the 'Massachusetts Road Dataset' used in the paper. Replace it with the correct road dataset URL or remove it.
A mismatched data link misleads readers and undermines reproducibility; this error was flagged by both the copyedit and integrity checks.
6.
HIGHreporting
Add a description of the training/validation/test split strategy for the Massachusetts Road and GF-2 Road datasets, including sample sizes for each subset.
This is a standard requirement for computational studies; its absence weakens the study design dimension.
7.
HIGHreporting
Report whether the model was run multiple times with different random seeds and provide mean ± standard deviation metrics across runs.
Without independent replication, it is impossible to assess the stability of the reported high performance; this is a key design gap.
8.
HIGHrigor
Add confidence intervals or error bars to the performance metrics in Tables 1 and 2 (e.g., via bootstrap or multiple runs) to quantify uncertainty.
The reported point estimates (e.g., 99.72% accuracy) have no measure of variability, limiting the reliability of conclusions; this also addresses a 'data too clean' integrity concern.
9.
HIGHreporting
Identify the software framework (e.g., TensorFlow 2.x, PyTorch 1.x) and hardware (GPU model, OS) used, with version numbers, in Section 2.
This is essential for reproducibility and directly addresses the 'key_resources' fail.
10.
MEDIUMcopyedit
Fix broken table headers in Tables 1 and 2: change 'A c c u y' to 'Accuracy', 'P r e c n' to 'Precision', 'R e c a l' to 'Recall', 'F 1 S c o r e' to 'F1 Score'.
These copyedit issues (major severity) reduce readability and professionalism.
11.
MEDIUMcopyedit
Fix grammar errors: 'The research gap in the existing works are often struggle with...' (Section 1) — rephrase to 'The research gap in existing works is that they often struggle with...'
Subject-verb agreement errors and awkward phrasing detract from clarity.
12.
MEDIUMreporting
Expand the limitations section (Section 4) beyond a single sentence. Discuss potential overfitting, dataset-specific tuning, and the need for real-world validation.
The current limitations statement is minimal; a thorough discussion would strengthen reporting transparency.
13.
MEDIUMreporting
Provide hyperparameter values used in AOA-QDCNNRE training (learning rate, batch size, number of epochs, optimizer) in Section 2.
Methods completeness is rated 'reported_but_inadequate' due to missing implementation details; these are needed for replication.
14.
MEDIUMrigor
Add a brief description of data curation or preprocessing steps (e.g., image resizing, augmentation, any exclusion criteria) in Section 2.
This addresses the 'inclusion_exclusion' gap and improves reproducibility.
15.
LOWcopyedit
Capitalize 'Archimedes Optimisation Algorithm' consistently in the title and throughout the text.
Consistency in proper nouns improves professionalism.

Full text (35,218 chars)Analyzed: Jul 6, 2026

Αlpha¹ assesses methodological rigor, not the importance of the findings. See how we calibrate →