Quantum Transfer Learning to Boost Dementia Detection

Himanshu Thapliyal; Sounak Bhowmik; Talita Perciano

REVIEW 4 major objections 6 minor 15 references

Quantum Transfer Learning to Boost Dementia Detection

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Quantum transfer learning can lift a weak dementia classifier from 73 percent to 91 percent accuracy on MRI data.

desk verdict A routine application of the authors' own QTL pipeline to OASIS-2, with a useful weak-baseline setup but an evaluation protocol that cannot support the headline accuracy claim. read the letter →

arxiv 2507.12485 v1 pith:QNBC2UY7 submitted 2025-07-14 quant-ph cs.AIcs.LG

Sounak Bhowmik , Talita Perciano , Himanshu Thapliyal This is my paper

classification quant-phcs.AIcs.LG

keywords dementiadetectionquantumtransferlearningvariationalcircuithybridquantum-classicalmodelMRIclassificationOASIS-2datasetnoisysimulationbinary

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to show that a weak classical convolutional network for dementia detection can be made much more accurate by replacing its final dense layers with a trained quantum circuit, a strategy called quantum transfer learning. On the OASIS-2 MRI dataset, the best hybrid configuration reaches about 91 percent test accuracy and 94 percent recall, compared with about 73 percent for the untuned baseline and about 86 percent for classical fine-tuning. The authors also report that this performance survives simulation with a realistic noisy device model. If true, this would mean that lightweight classical models need not be discarded when they underperform: a quantum readout head can upgrade them while leaving the convolutional feature extractor frozen.

What carries the argument

The load-bearing component is the dressed quantum network (DQN), a hybrid layer that replaces the classical dense head. A pre-net maps the CNN's latent vector to the chosen number of qubits, a tanh activation and scaling by pi/2 prepare the values for angle embedding, and a parameterized quantum circuit built from Rz rotations, a ring of CNOT entangling gates, and controlled Ry rotations processes the embedded state. Measurements of the circuit give expectation values that a post-net turns into a single logit for binary classification. The same frozen convolutional feature extractor is used throughout, so the comparison isolates the effect of the quantum head against the classical fine-tuned head.

What would settle it

Rerun the same experiment with a stratified train/test split by subject, a fixed seed, and a class-balanced test set; if the QTL accuracy falls to the level of classical fine-tuning, the central claim fails.

Watch

Extended reading notes

Core claim

The central claim is that quantum transfer learning can transform a suboptimal classical model into a better dementia classifier. Concretely, the paper reports that a 6-qubit, 4-repetition variational circuit, used as the classification head of a frozen small convolutional network, achieves 91.29 percent test accuracy, 0.8750 precision, 0.9412 recall, 0.9069 F1, and 0.9232 AUC on the OASIS-2 binary dementia/non-dementia task. The comparable classically fine-tuned model achieves 86.36 percent accuracy and 0.8739 recall, and the untuned baseline achieves 73.10 percent accuracy and 0.7394 recall. The paper further claims the quantum model remains reliable under a depolarizing noisy-device simulation, showing resilience without explicit error mitigation.

Load-bearing premise

Every reported accuracy depends on the assumption that the randomly chosen 30 percent of subjects used for testing is balanced and representative; the split used no fixed random seed and no stratification, and the two most image-heavy subjects were kept entirely in training.

Editorial extensions

If this is right

A weak, resource-constrained classical model can be upgraded by swapping only its final layer for a trained quantum circuit, without retraining the convolutional feature extractor.
The high recall of 0.94 means fewer demented patients are missed, which matters more than raw accuracy in a screening setting.
Because the gain persists under a depolarizing noise model, near-term quantum hardware without full error correction may be usable for this kind of hybrid classification.
The reported numbers give a concrete benchmark: 6 qubits and 4 circuit repetitions outperform both the classical baseline and classical fine-tuning on this dataset.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Editorial inference: the classically fine-tuned head was retrained from scratch after freezing the convolutional layers, so part of the quantum gain may come from reinitialization rather than from quantum mechanics; a control experiment that reinitializes the classical head with the same protocol would isolate the quantum contribution.
Editorial inference: the single random, non-stratified 30-percent test split means the reported margins could shift if the split is unlucky; testing across multiple seeds would show whether the quantum advantage is stable.
Editorial inference: because the paper observes a barren-plateau-like flat landscape, scaling the circuit to more qubits may erase the accuracy gain; a natural extension is to test whether the advantage persists with 10 or more qubits or with a different ansatz.
Editorial inference: the slight drop in AUC for QTL versus classical fine-tuning (0.9232 versus 0.9399) suggests the quantum model is not uniformly better; its claim is specifically about threshold-based metrics.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague,

This paper applies the authors' existing quantum transfer learning (QTL) pipeline from Ref. [1] to the OASIS-2 dementia dataset. The genuinely useful part is the setup: they intentionally start from a weak CNN baseline, compare against classical fine-tuning, and include a noisy-device simulation. The subject-level split that keeps the two dominant subjects (IDs 1 and 2) entirely in the training set shows real awareness of leakage, and acknowledging barren plateaus in a short conference paper is fair.

But the central claim—that QTL boosts a weak classifier by 24% over baseline and 5.7% over CTL—is not supported by the reported evaluation. The stress-test concern is correct and load-bearing: Section 4.2 selects the 6-qubit/4-repetition configuration after a grid search over 24 options, using test accuracies on the same test set that later produces the headline number. The authors even say performance was 'almost equivalent' across configurations, which means the gap between the chosen run and the runner-up is likely noise, and picking the max after seeing the test set inflates the gain. The CTL comparison is also weakened by selecting the best of several Glorot initializations on the same test set. With single runs, no error bars, and no separate validation set, the 5.7-point quantum-specific advantage is indistinguishable from selection noise.

A second red flag: the noisy-simulator accuracy (91.29%) comes out higher than the noiseless result (90.50%) for the same configuration. That is not impossible, but it is the kind of result that needs explanation before it can be believed. The missing seed and unclear cross-validation description in Section 3 add to the unreliability.

The paper is honest about its limits in places, and there is no circularity beyond the selection bias. But as it stands, the metric that matters—whether QTL really improves upon a reasonable classical baseline—is unproven. The novelty is modest: this is the authors' own method applied to a known benchmark, and prior hybrid QML dementia papers already exist.

Who gets value from this? A reader interested in seeing how QTL pipelines are being adapted to medical imaging might skim it for the architectural details. But I would not cite the empirical numbers.

Recommendation: the paper deserves a serious referee only if the venue expects heavy revision. The evaluation can be fixed with repeated seeds, a held-out validation set for hyperparameter selection, stratified subject-level splits, error bars, and a proper explanation of the noisy-vs-noiseless discrepancy. Without those, the central claim is not credible.

Best, [You]

Referee Report

4 major / 6 minor

Summary. The manuscript proposes a quantum transfer learning (QTL) pipeline for binary dementia detection on the OASIS-2 MRI dataset. A weak classical CNN baseline is first trained; its convolutional layers are frozen and its final dense layers are replaced by a dressed quantum network consisting of a classical pre-net, angle embedding, a variational quantum circuit with varying qubit counts and ansatz repetitions, and a classical post-net. The authors report that the best QTL configuration (6 qubits, 4 repetitions) reaches 91.29% test accuracy and 0.94 recall, versus 73.10% for the baseline and 86.36% for classical fine-tuning, and they present results from an IonQ Forte-1 noisy simulator to support robustness claims. The paper also reports a grid search over 24 QTL configurations and notes the occurrence of barren plateaus.

Significance. If the central claim were established, the paper would provide a useful empirical demonstration that a quantum fine-tuning head can improve a deliberately weak classical model on a medical imaging task, with a realistic NISQ noise model. The comparison against a classical fine-tuned counterpart on a public benchmark is a sensible experimental design, and the explicit reporting of the grid-search range and the barren-plateau observation are helpful. However, the reported quantitative gains are not currently reliable: the headline configuration is selected on the test set, all metrics come from single runs without error bars, and the noisy-simulator accuracy is higher than the noiseless accuracy for the same configuration. The paper does not provide code, random seeds, or a validation-based selection protocol, so the central claim is not yet reproducible or statistically supported.

major comments (4)

[Section 4.2, Table 3] The model selection protocol uses the test set twice. The authors state that they chose the 6-qubit, 4-repetition setup 'based on the test accuracies recorded in the ideal simulation' after a grid search over 24 configurations, and then report the accuracy of that same configuration on the same test set. Selecting the maximum over 24 runs on the test set biases the reported 91.29% upward, and the bias is especially concerning because the authors themselves note that performance was 'almost equivalent' across configurations, meaning the maximum may reflect selection noise rather than a genuinely superior model. A separate validation split should be used for selection, with the test set evaluated only once for the final reported numbers; alternatively, nested cross-validation should be used.
[Section 3.1] The data split is not described with enough detail to rule out leakage or imbalance. The test set is formed by a random 30% of subjects excluding IDs 1 and 2, but no random seed is reported and no stratification by class or by subject-level image count is described. Because IDs 1 and 2 contribute the largest numbers of images and are placed entirely in the training set, the test set may not be representative of the full dataset. The authors should report the seed, the class distribution of the train and test sets, and ideally perform subject-level stratified cross-validation to confirm that all reported accuracies are stable.
[Section 4.2, Table 3] The noisy-simulator result (91.29% test accuracy) is higher than the noiseless ideal-simulation result for the same 6-qubit, 4-repetition configuration (90.50%). Under a depolarizing noise model one would expect degradation, not improvement, unless the discrepancy is within run-to-run variation or stems from a different inference procedure. This anomaly is load-bearing for the paper's robustness claim and should be explained, ideally with repeated runs under both simulators and with error bars.
[Section 4.2 and 3.4] All reported metrics appear to come from a single training run per configuration. The classical fine-tuning baseline is described as the 'best-performing model' among several Glorot initializations, but the number of initializations and the spread of results are not given. Given the paper's own observation of flat optimization landscapes and near-equivalent QTL performance across configurations, the claimed 5.71-percentage-point improvement of QTL over CTL cannot be distinguished from seed noise without repeated initializations, standard deviations, or confidence intervals.

minor comments (6)

[Section 4.1] The text says 'Performance improved along all the other matrices by roughly 20%'; 'matrices' should be 'metrics'.
[Section 3.3] The notation for the tanh output range is written as '[1, -1]' and then '[π/2, -π/2]'; the conventional ordering is '[-1, 1]' and '[-π/2, π/2]'.
[Section 3.4] The phrase 'cross-validated and trained the QTL-based models' in Section 3.5 is ambiguous, and the earlier statement in Section 3 that the authors 'assess the overall performance by 4-fold cross-validation, training on the entire dataset, and validating on a separate test dataset' appears internally contradictory; the exact protocol should be stated once and consistently.
[Figure 4] The caption describes 'An ansatz' with Rz gates, a C-NOT ring, and controlled Ry gates, but the text says two different ansatzes were used; it should be clarified which ansatz is shown and how the second one differs.
[Section 3.1] The description of Figure 2 says demented and non-demented images from IDs other than 1 and 2 are 'shown in the third column,' but the figure is not self-explanatory; labeling the columns directly would improve clarity.
[General] The paper would benefit from releasing the code and specifying all random seeds, simulator versions, and the exact preprocessing steps for the OASIS-2 images, as these details are necessary for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation; empirical claim rests on held-out evaluation, with a test-set-model-selection caveat that is a validity issue, not circularity.

full rationale

This paper makes no formal derivation that could be circular: the central claim is an empirical comparison of a baseline CNN, a classically fine-tuned model, and several quantum transfer learning configurations on a held-out test split of OASIS-2. The quantum transfer learning framework is taken from the authors' prior work (reference [1]), but that citation only introduces the hybrid pipeline; it does not supply the dementia accuracies or the comparison numbers. No equation in Section 2 defines the reported test accuracy in terms of itself, and no fitted parameter is renamed as a prediction. The one statistical concern is that Section 4.2 selects the 6-qubit, 4-repetition configuration using test accuracies from the ideal simulation and then reports the same test set's noisy accuracy as the headline result; this is a model-selection bias that can inflate the reported 91.29% and the apparent gain over classical fine-tuning, but it is not a definitional circularity because the reported noisy accuracy is a separate evaluation on the same held-out set rather than the same quantity used for selection by construction. The comparison models (baseline and CTL) are also evaluated on the same test set, so the relative comparison is affected by the same protocol. Under the strict definition of circularity used here, there is no self-definitional equivalence, no self-citation chain carrying the result, and no ansatz smuggled in via citation; the contribution is an empirical benchmark, and any weakness in the model-selection protocol is a methodological validity issue rather than a circularity.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The central empirical claim rests on a small public dataset, a random subject split, a hand-picked best configuration, and a noisy simulator. No new physical entities are introduced. The main free parameter is the best-of-grid QTL configuration, and the main auxiliary assumptions are dataset quality, feature transferability, and simulator fidelity.

free parameters (1)

Best QTL configuration (number of qubits and ansatz repetitions) = 6 qubits, 4 repetitions
Selected as the highest test accuracy among 24 configurations in Section 4.2, which is model selection on the test set and inflates reported performance.

assumptions (3)

domain assumption OASIS-2 preprocessed MRI images and their dementia labels are correct and representative.
All accuracy numbers inherit dataset quality; no visual or clinical validation is performed in the paper.
domain assumption Frozen convolutional features from the weak baseline remain informative for dementia classification after retraining the head.
Central to transfer learning; if the features are poor, both classical fine-tuning and QTL improvements would not transfer.
domain assumption The IonQ Forte-1 noisy simulator with a depolarizing noise channel approximates real device behavior well enough to assess robustness.
Used in Section 3.5; no real hardware experiments are performed, so the noise resilience claim depends on this approximation.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Quantum Transfer Learning to Boost Dementia Detection." pith.science (2026). https://pith.science/paper/QNBC2UY7

@misc{pith2026250712485,
  author       = {Pith},
  title        = {Pith review of: Quantum Transfer Learning to Boost Dementia Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QNBC2UY7}},
  note         = {Machine review of arXiv:2507.12485}
}

read the original abstract

Dementia is a devastating condition with profound implications for individuals, families, and healthcare systems. Early and accurate detection of dementia is critical for timely intervention and improved patient outcomes. While classical machine learning and deep learning approaches have been explored extensively for dementia prediction, these solutions often struggle with high-dimensional biomedical data and large-scale datasets, quickly reaching computational and performance limitations. To address this challenge, quantum machine learning (QML) has emerged as a promising paradigm, offering faster training and advanced pattern recognition capabilities. This work aims to demonstrate the potential of quantum transfer learning (QTL) to enhance the performance of a weak classical deep learning model applied to a binary classification task for dementia detection. Besides, we show the effect of noise on the QTL-based approach, investigating the reliability and robustness of this method. Using the OASIS 2 dataset, we show how quantum techniques can transform a suboptimal classical model into a more effective solution for biomedical image classification, highlighting their potential impact on advancing healthcare technology.

Figures

Figures reproduced from arXiv: 2507.12485 by the authors.

**Figure 2.** Data distribution based on patient IDs. the distribution of the number of images across different subjects and classes in the dataset. Here we can see that a large number of non-demented (the first blue column) images are associated with ID=1. Similarly, a significantly large number of demented (the second yellow column) images are associated with ID=2. The brain images from other patients with IDs apart from {1, 2… view at source ↗

**Figure 4.** Ansatz comprises parameterised Rz gates, entan [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗

Figures from the paper (1 more)

**Figure 3.** Figure 3: The setup to apply quantum transfer learning to [PITH_FULL_IMAGE:figures/full_fig_p003_3.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

15 extracted references · 14 canonical work pages

[1]

Sounak Bhowmik and Himanshu Thapliyal. 2024. Transfer Learning Based Hybrid Quantum Neural Network Model for Surface Anomaly Detection. In 2024 IEEE Computer Society Annual Symposium on VLSI (ISVLSI) . IEEE, 634–639

work page 2024
[2]

Giacomo Cappiello and Filippo Caruso. 2024. Quantum AI for Alzheimer’s disease early screening. arXiv preprint arXiv:2405.00755 (2024)

arXiv 2024
[3]

Kuan-Cheng Chen, Yi-Tien Li, Tai-Yu Li, Chen-Yu Liu, Po-Heng Li, and Cheng-Yu Chen. 2024. CompressedMediQ: Hybrid Quantum Machine Learning Pipeline for High-Dimensional Neuroimaging Data. arXiv:2409.08584 [quant-ph] https: //arxiv.org/abs/2409.08584

work page Pith review arXiv 2024
[4]

Abebech Jenber Belay, Yelkal Mulualem Walle, and Melaku Bitew Haile

work page
[5]

Ryan Kim. 2023. Hybrid Quantum-Classical Machine Learning for Dementia De- tection. In 2023 IEEE MIT Undergraduate Research Technology Conference (URTC) . IEEE, 1–5

work page 2023
[6]

Moona Mazher, Abdul Qayyum, MKA Ahamed Khan, Steven Niederer, Mastaneh Mokayef, and CS Hassan. 2024. Hybrid Classical and Quantum Deep Learning Models for Medical Image Classification. (2024)

work page 2024
[7]

Jarrod R McClean, Sergio Boixo, Vadim N Smelyanskiy, Ryan Babbush, and Hartmut Neven. 2018. Barren plateaus in quantum neural network training landscapes. Nature communications 9, 1 (2018), 4812

work page 2018
[8]

Nikhil Pateria and Dilip Kumar. 2024. A comprehensive review on detection and classification of dementia using neuroimaging and machine learning. Multimedia Tools and Applications 83, 17 (2024), 52365–52403

work page 2024

Show all 15 references

[9]

D Patil, NL Rane, P Desai, and J Rane. 2024. Machine learning and deep learning: Methods, techniques, applications, challenges, and future research opportunities. Trustworthy Artificial Intelligence in Industry and Society (2024), 28–81

2024
[10]

Stefan H Sack, Raimel A Medina, Alexios A Michailidis, Richard Kueng, and Maksym Serbyn. 2022. Avoiding barren plateaus using classical shadows. PRX Quantum 3, 2 (2022), 020365

2022
[11]

Georgina Waldo-Benítez, Luis Carlos Padierna, Pablo Ceron, and Modesto A Sosa. 2024. Dementia classification from magnetic resonance images by machine learning. Neural Computing and Applications 36, 6 (2024), 2653–2664

2024
[12]

Georgina Waldo-Benítez, Luis Carlos Padierna, Pablo Ceron, and Modesto A. Sosa. 2021. OASIS2 images for dementia classification by machine learning. doi:10.21227/284g-hf84

2021 doi
[13]

Yihan Wang, Shu Liu, Alanna G Spiteri, Andrew Liem Hieu Huynh, Chenyin Chu, Colin L Masters, Benjamin Goudey, Yijun Pan, and Liang Jin. 2024. Understanding machine learning applications in dementia research and clinical practice: a review for biomedical scientists and clinicia...

2024
[14]

Manuela Weigold, Johanna Barzen, Frank Leymann, and Marie Salm. 2020. Data encoding patterns for quantum computing. In Proceedings of the 27th Conference on Pattern Languages of Programs . 1–11

2020
[2024]

Scientific Reports 14, 1 (2024), 14196

Deep Ensemble learning and quantum machine learning approach for Alzheimer’s disease detection. Scientific Reports 14, 1 (2024), 14196

2024

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Sounak Bhowmik and Himanshu Thapliyal. 2024. Transfer Learning Based Hybrid Quantum Neural Network Model for Surface Anomaly Detection. In 2024 IEEE Computer Society Annual Symposium on VLSI (ISVLSI) . IEEE, 634–639

work page 2024

[2] [2]

Giacomo Cappiello and Filippo Caruso. 2024. Quantum AI for Alzheimer’s disease early screening. arXiv preprint arXiv:2405.00755 (2024)

arXiv 2024

[3] [3]

Kuan-Cheng Chen, Yi-Tien Li, Tai-Yu Li, Chen-Yu Liu, Po-Heng Li, and Cheng-Yu Chen. 2024. CompressedMediQ: Hybrid Quantum Machine Learning Pipeline for High-Dimensional Neuroimaging Data. arXiv:2409.08584 [quant-ph] https: //arxiv.org/abs/2409.08584

work page Pith review arXiv 2024

[4] [4]

Abebech Jenber Belay, Yelkal Mulualem Walle, and Melaku Bitew Haile

work page

[5] [5]

Ryan Kim. 2023. Hybrid Quantum-Classical Machine Learning for Dementia De- tection. In 2023 IEEE MIT Undergraduate Research Technology Conference (URTC) . IEEE, 1–5

work page 2023

[6] [6]

Moona Mazher, Abdul Qayyum, MKA Ahamed Khan, Steven Niederer, Mastaneh Mokayef, and CS Hassan. 2024. Hybrid Classical and Quantum Deep Learning Models for Medical Image Classification. (2024)

work page 2024

[7] [7]

Jarrod R McClean, Sergio Boixo, Vadim N Smelyanskiy, Ryan Babbush, and Hartmut Neven. 2018. Barren plateaus in quantum neural network training landscapes. Nature communications 9, 1 (2018), 4812

work page 2018

[8] [8]

Nikhil Pateria and Dilip Kumar. 2024. A comprehensive review on detection and classification of dementia using neuroimaging and machine learning. Multimedia Tools and Applications 83, 17 (2024), 52365–52403

work page 2024