Pith. sign in

REVIEW 5 major objections 5 minor 39 references

This paper argues that subtracting the average Log-PSD of authentic MRI volumes from a volume's spectrum isolates the scanner hardware trace and the generative fingerprint, and that a token-mixing 3D network reading this residual can detect

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-04 03:14 UTC pith:LHSKPG3K

load-bearing objection Plausible frequency-domain detector for 3D MRI deepfakes, but the 'anatomical prior' framing is overreaching and the evaluation needs baseline comparisons and error bars. the 5 major comments →

arxiv 2607.29541 v2 pith:LHSKPG3K submitted 2026-07-31 cs.CV

The K-Space Signature: Frequency-Domain Representation Learning for Medical Deepfake Detection

Riccardo Raciti , Francesco Guarnera , Francesco Rundo , Luca Guarnera , Sebastiano Battiato This is my paper
classification cs.CV
keywords K-Space SignatureMedical deepfake detectionFrequency-domain forensicsLog-PSD residualMRI scanner attributionMLP-MixerArcFaceGenerative-model artifacts
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper tries to show that a simple spectral subtraction can separate the machine that took an MRI scan from the brain being scanned, and that this separation makes it possible to spot fully synthetic medical images. The proposed method, the K-Space Signature (KSS), replaces each volume by its logarithmic power spectrum and subtracts the average logarithmic spectrum of authentic training volumes. What remains is claimed to be a residual containing hardware noise and generative artifacts but almost no anatomy. A token-mixing 3D network with an angular-margin classification head then reads these residuals and, according to the authors, exceeds 0.99 accuracy and ROC-AUC on multi-generator synthetic data, and keeps up to 0.93 accuracy on scanners never seen in training. If true, the approach would give hospitals a way to verify that an MRI came from a real machine rather than from a generative model.

Core claim

The central discovery claim is that the residual KSS_i = P~_i − P~_global, where P~_global is the empirical mean Log-PSD over authentic training volumes, constitutes a representation that isolates non-anatomical traces: scanner hardware fingerprints and generative-model artifacts. The authors support this with pairwise cosine-similarity analysis showing strong intra-manufacturer clustering (GE models at 0.97, Siemens at 0.96) and marked inter-manufacturer orthogonality (negative correlations), and showing that synthetic volumes exhibit a strong positive correlation with Siemens hardware, interpreted as a learned bias from generative training distributions. Paired with a 3D MLP-Mixer plus Arc

What carries the argument

The load-bearing mechanism is the KSS residual: for a volume V, compute P~ = ln(1+|F(V)|^2) (the logarithmic power spectral density), then subtract the empirical mean P~_global computed over authentic training volumes. The subtraction is designed to cancel the shared low-frequency anatomical content, leaving a residual that concentrates hardware acquisition noise and generative artifacts. A second mechanism is the 3D MLP-Mixer, which mixes tokens across the whole volumetric spectrum (rather than local convolutional patches) and is paired with an ArcFace spherical metric head (an additive angular margin loss) that forces real and synthetic signatures into separated angular clusters. The paper

Load-bearing premise

The method's central assumption is that the average Log-PSD of authentic training volumes is a purely anatomical prior; if scanner hardware traces do not cancel out in that average, the subtracted baseline contaminates the residual and the claimed hardware-invariant generalization weakens.

What would settle it

Train the KSS pipeline with the global baseline computed from only two of the three manufacturers, then test on the third. If detection accuracy on that held-out vendor's authentic volumes (or on synthetic volumes matched to that vendor's spectrum) drops far below the reported ~0.93, the 'anatomy-only' prior is falsified and the residual is carrying scanner content rather than purely generative traces.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If the KSS residual indeed suppresses anatomy, forensic models trained on it cannot cheat by memorizing ventricle shapes or cortical folding; any correct classification is attributable to acquisition or generation traces.
  • The reported performance (Accuracy and ROC-AUC above 0.99) indicates that state-of-the-art 3D generative models, including diffusion-based ones, leave detectable spectral traces when the global baseline is subtracted.
  • The strong positive correlation between synthetic volumes and one vendor's hardware implies that generative models inherit and replicate the spectral properties of their training corpora; this could be used to attribute a synthetic volume to the scanner family that dominated the generator's training data.
  • Leave-one-scanner-out experiments suggest that a detector trained on one manufacturer's scans transfers to other manufacturers, up to the reported ~0.93 accuracy, provided the decision threshold is chosen appropriately.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • A natural extension, not pursued in the paper, is to use the KSS residual to localize where in the spectrum the generative trace lives; the reverse-mapping visualizations hint at band-selective signatures that could support partial-forgery detection rather than whole-volume binary classification.
  • If the anatomy-only assumption on P~_global is relaxed, an even stronger test would be to compute the baseline from a different patient population or a non-brain modality; the framework's portability to CT or PET will hinge on whether an analogous 'global anatomical prior' exists for those modalities.
  • The threshold-collapse cases (e.g., Med vs Philips) suggest practical deployments should use calibrated thresholds or anomaly scores rather than a fixed 0.5 cutoff, since the ranking of real vs synthetic is preserved even when raw accuracy drops.
  • The reported correlation between synthetic data and Siemens hardware implies a potential privacy leak: a generative model trained on a specific hospital's Siemens-heavy data may embed that vendor bias, and KSS analysis could be used to infer the composition of training data of an unknown generative model.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper introduces the K-Space Signature (KSS), a frequency-domain representation for detecting fully synthetic 3D MRI volumes. The method computes the Log-PSD of each volume, subtracts the empirical mean Log-PSD over authentic training volumes (Eq. 4), and classifies the residual KSS tensor with a 3D MLP-Mixer equipped with an ArcFace metric-learning head. Experiments on ADNI/PPMI real scans and MAISI/Med-DDPM/SuperSynth synthetic volumes report near-perfect closed-set detection and moderate-to-high open-set accuracy on unseen scanners, alongside scanner-attribution and visual-explainability analyses. The paper claims that KSS isolates hardware and generative traces while suppressing anatomy, and that the MLP-Mixer-ArcFace combination yields exceptional detection and zero-shot generalization.

Significance. If the empirical claims hold, KSS is a simple, cheap spectral residual with potentially useful generalization benefits across scanners. The manuscript has several strengths: strict patient-level splitting, extensive supplementary ablations (23 baseline formulations, interpolation robustness, RAW-vs-REG spectral integrity, zero-shot gender classification), and a commitment to release code. The frequency-domain residual idea is plausible and worth pursuing. However, the central interpretability claim—that the global mean is an 'anatomical prior'—is not established, and the experimental comparison is incomplete: there are no comparisons to existing medical deepfake detectors, no error bars or repeated-seed statistics, and several accuracy/AUC inconsistencies. The contribution is potentially valuable but not yet rigorously supported.

major comments (5)
  1. [§2, Eq. (4)–(5)] The central claim that KSS 'isolates hardware and generative traces while suppressing anatomy' is not established. P~global is the empirical mean Log-PSD over authentic training volumes from six scanner models, so it is a hardware-weighted average, not purely anatomical. For a volume from scanner s, the residual is KSS_i ≈ (A_i − A_mean) + (H_s − H_mean); both anatomical deviations and scanner hardware remain. Moreover, Fig. 3 shows synthetic volumes are strongly correlated with Siemens spectra; if Siemens is a substantial component of Rtrain, subtracting P~global partially cancels the generative signature the method aims to detect. The ablation in Supplementary Table 2 is a 6-class scanner-classification task, not a deepfake-detection task, so it does not show that the global prior's scanner composition is immaterial for detection. Please either demonstrate (e.g., by recomputing P~globa
  2. [§3, 'Medical Deepfake Detection'; Table 3] The headline 'exceptional detection performance' is not benchmarked against any existing medical deepfake detector. The manuscript cites Li et al. (2025) and Grabovski et al. (2025) as related methods but never compares to them; the baselines are generic 3D CNNs and MLP-Mixer variants only. Also, no ablation isolates the effect of the log transform from the subtraction: 'K-Space' magnitude is compared to 'KSS', but the latter differs by both log-scaling and mean subtraction. Add comparisons to at least one medical-deepfake-specific method and a Log-PSD-without-subtraction baseline to support the claim that the KSS representation itself is responsible for the reported gains.
  3. [§3, 'Experimental Setup'; Tables 1–3] No error bars, repeated seeds, or significance tests are reported. Given the small open-set (120 volumes) and near-saturated closed-set numbers, single-run results are not convincing. Report mean ± standard deviation over at least 5 random seeds (ideally with different patient splits) and, for key zero-shot comparisons, a paired significance test.
  4. [Abstract; Table 1; Table 3] The reported Accuracy and AUC values are not coherent as presented. In Table 3, the KSS MLP-Mixer-ArcFace model trained on MAISI has Open Dataset Accuracy 0.973 but ROC-AUC 0.769. In Table 1, the Med/PHILIPS row reports Target Fakes Accuracy 55% despite PR-AUC and ROC-AUC 100%, and Open Set Accuracy 49% with PR-AUC 81% and ROC-AUC 68%. Such patterns imply the Open Set test sets are imbalanced or the decision threshold is chosen without calibration, neither of which is documented. The abstract's 'up to 0.93 Accuracy' does not match Table 1, where the maximum Open Set Accuracy is 99% and the maximum PR-AUC is 93%; the 0.93 value appears to be PR-AUC. Please clarify class balance, threshold selection, and correct the abstract.
  5. [§3, 'Robustness and Scanner-Invariant Generalization'; Table 1] The protocol does not state whether P~global is recomputed from the single training manufacturer or from the full multi-scanner authentic training set. If the global prior of Eq. (4) is used, then the KSS residuals seen during training contain contributions from the held-out scanner manufacturers, undermining the zero-shot generalization claim. The supplementary Zero-Shot Domain Adaptation section explicitly recomputes P~global on source domains; the main leave-one-manufacturer-out experiment must do the same (or justify why it need not).
minor comments (5)
  1. [Fig. 3 and accompanying text] The phrase 'mathematically orthogonal inter-vendor domains' is inaccurate for negative correlations (e.g., -0.75 to -0.77); orthogonality corresponds to zero correlation. Suggest 'oppositely oriented' or 'near-orthogonal' with an explicit reference to the zero baseline.
  2. [§2, Eq. (8); Supplementary B] The ArcFace loss is written for a K-class problem, but the cross-dataset experiments use BCEWithLogitsLoss for binary classification. Clarify how the ArcFace margin is implemented for K=2 and why BCE is used for the main results.
  3. [§3, 'Pairwise Similarity Evaluation', Eq. (9)] The cosine similarity is computed on averaged P~ tensors. Specify the number of volumes averaged per scanner and whether the same computation on the KSS tensors (after mean subtraction) would change the qualitative conclusions.
  4. [Fig. 4] The x-axis is categorical with arbitrary ordering; the overlaid linear trend lines are misleading. Consider per-generator box plots or a clearer visualization that does not imply a monotone relationship.
  5. [Supplementary Table 3] The 'Open Dataset' column should specify its composition—how many real scans from unseen scanners and how many synthetic samples from each generator—and the class ratio. This is needed to interpret Accuracy/AUC inconsistencies.

Circularity Check

0 steps flagged

No significant circularity: KSS is a label-free centering transform and detection is independently tested.

full rationale

The derivation chain is not circular. P~_global (Eq. 4) is the empirical mean Log-PSD over authentic training volumes and is computed without using class labels; Eq. (5) is a fixed centering transform, KSS_i = P~_i − P~_global. Detection then trains a classifier on these residuals and evaluates on patient-disjoint held-out volumes and on scanner-disjoint open-set data, so the reported accuracy/AUC is an independent test of the representation, not a re-statement of the fitted mean. The paper's self-citations (Guarnera et al. 2020/2024; Giudice et al. 2021) are contextual references to prior deepfake-detection methodology and are not load-bearing for the KSS construction or for any claimed uniqueness of the baseline. The concern that P~_global may encode hardware as well as anatomy is a correctness/generalization risk—the paper itself notes the need for 'a representative cohort of authentic training volumes'—but it is not a circular reduction: no equation makes the prediction equal to the fitted mean, and the mean is not optimized against authenticity labels. The robustness ablations in the supplement provide empirical checks rather than definitional identities. Therefore no circular step is present.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 1 invented entities

The method's central representation rests on one fitted tensor (P~global) and a set of domain assumptions about anatomy, hardware noise, and generative failure modes. No external benchmark or formal proof independently establishes that P~global is anatomy-only; the paper's ablations provide internal consistency but not external grounding.

free parameters (4)
  • Global anatomical prior P~global = Empirical mean Log-PSD tensor over authentic training volumes (Eq. 4)
    Central to KSS extraction; assumed to be anatomy-only, but it averages over six scanner models and thus contains a mixture of hardware traces.
  • ArcFace margin m = 0.5
    Hand-chosen angular margin used in the metric-learning head.
  • ArcFace scale s = 30.0
    Hand-chosen hypersphere scale used in the metric-learning head.
  • MLP-Mixer architecture hyperparameters = patch=16, dim=128, depth=6, token-dim=256, channel-dim=512, dropout=0.4
    Fixed architecture choices that affect reported performance; not fitted to the test labels but chosen by the authors.
axioms (5)
  • standard math The DFT of a fully reconstructed spatial MRI volume is mathematically equivalent to raw k-space (Eq. 2).
    Standard Fourier analysis; used to justify calling the frequency representation k-space.
  • domain assumption The aggregate Log-PSD over a large diverse authentic dataset converges to a global anatomical prior that is scanner-invariant (Eq. 4).
    This is the core assumption that subtracting P~global removes anatomy while preserving hardware and generative traces.
  • domain assumption Generative models cannot faithfully replicate the physical noise of clinical MRI scanners.
    Underpins the claim that spectra reveal synthetic content; cited from PRNU literature but not proven for the specific generators used.
  • domain assumption MLP-Mixer's global token-mixing is better suited than CNN locality for diffuse spectral artifacts.
    Motivates the architecture choice; supported only by the paper's own experiments.
  • domain assumption Rigid registration with Lanczos interpolation preserves the spectral geometry of hardware traces.
    All volumes are registered before KSS extraction; the supplementary ablation supports this but it remains a preprocessing assumption.
invented entities (1)
  • K-Space Signature (KSS) representation no independent evidence
    purpose: A Log-PSD residual intended to isolate non-anatomical hardware and generative traces.
    KSS is a constructed feature space whose claimed disentanglement is validated only within this paper's own experiments and similarity matrices.

pith-pipeline@v1.3.0-alltime-deepseek · 32037 in / 9365 out tokens · 481556 ms · 2026-08-04T03:14:49.030451+00:00 · methodology

0 comments
read the original abstract

In medical imaging, generative models are increasingly deployed to synthesize realistic data and augment limited datasets. Unfortunately, while beneficial for privacy-preserving data sharing, these synthesized images can be repurposed for malicious intents, threatening public health through the creation of Medical Deepfakes. To address this threat, we introduce the K-Space Signature (KSS), a novel forensic framework that isolates hardware and generative traces within the spectral domain. By shifting analysis to the frequency domain, the KSS suppresses macroscopic anatomical variance by subtracting an empirical global anatomical prior computed in the Logarithmic Power Spectral Density (Log-PSD) space. To effectively process these globally distributed spectral artifacts without the local spatial bias inherent to Convolutional Neural Networks, we pair the KSS representation with a novel 3D MLP-Mixer architecture equipped with an ArcFace metric-learning head. Extensive experiments on multi-center 3D MRI datasets demonstrate that this combined approach achieves exceptional detection performance, exceeding 0.99 Accuracy and ROC-AUC on multi-generator synthetic datasets. Furthermore, the framework exhibits robust zero-shot generalization, maintaining strong discriminative power (up to 0.93 Accuracy) on independent datasets acquired from entirely unseen scanners. To ensure full reproducibility, the complete source code and pre-trained models will be made publicly available upon acceptance.

Figures

Figures reproduced from arXiv: 2607.29541 by Francesco Guarnera, Francesco Rundo, Luca Guarnera, Riccardo Raciti, Sebastiano Battiato.

Figure 1
Figure 1. Figure 1: Overview of the proposed framework. (1) Global Modeling: A universal anatomical prior, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Real and synthetic MRI across domains. Rows: spatial anatomy (top), K-Space (middle), and KSS traces (bottom). [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Cross-model cosine similarity matrix of isolated [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Model accuracy comparison (Closed Dataset vs. Open Dataset) across architectures (ResNet3D, ShuffleNet3D, [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Triplanar reconstructions demonstrating that [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Extended cross-domain similarity matrix compar [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Spectral footprint of the registration pipeline. The [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Cross-interpolation Cosine Similarity matrix of the [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: 1D radial profiles of the global anatomical base [PITH_FULL_IMAGE:figures/full_fig_p015_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Visual explainability and feature disentanglement of the KSS. The figure is organized by manufacturer: the first [PITH_FULL_IMAGE:figures/full_fig_p018_11.png] view at source ↗

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

39 extracted references

  1. [1]

    Neuroimaging Clinics , volume=

    The Alzheimer's disease neuroimaging initiative , author=. Neuroimaging Clinics , volume=. 2005 , publisher=

    2005

  2. [2]

    and Jennings, D

    Marek, K. and Jennings, D. and Siderowf, A. and Tanner, C. and Simuni, T. and Coffey, C. and Kieburtz, K. and PPMI investigators , title =. Progress in Neurobiology , year =

  3. [3]

    IEEE Journal of Biomedical and Health Informatics , volume=

    Conditional Diffusion Models for Semantic 3D Brain MRI Synthesis , author=. IEEE Journal of Biomedical and Health Informatics , volume=. 2024 , publisher=

    2024

  4. [4]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    MAISI-v2: Accelerated 3D High-Resolution Medical Image Synthesis with Rectified Flow and Region-specific Contrastive Loss , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  5. [5]

    Medical image analysis , volume=

    SynthSeg: Segmentation of brain MRI scans of any contrast and resolution without retraining , author=. Medical image analysis , volume=. 2023 , publisher=

    2023

  6. [6]

    2019 IEEE conference on multimedia information processing and retrieval (MIPR) , pages=

    Do GANs Leave Artificial Fingerprints? , author=. 2019 IEEE conference on multimedia information processing and retrieval (MIPR) , pages=. 2019 , organization=

    2019

  7. [7]

    The New England journal of medicine , volume=

    Implementing Machine Learning in Health Care — Addressing Ethical Challenges , author=. The New England journal of medicine , volume=

  8. [8]

    Medical image analysis , volume=

    Adversarial attack vulnerability of medical image analysis systems: Unexplored factors , author=. Medical image analysis , volume=. 2021 , publisher=

    2021

  9. [9]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    FaceForensics++: Learning to Detect Manipulated Facial Images , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  10. [10]

    Scientific Reports , volume=

    Effects of MRI scanner manufacturers in classification tasks with deep learning models , author=. Scientific Reports , volume=. 2023 , publisher=

    2023

  11. [11]

    Electronic Imaging , volume=

    A deep learning approach to mri scanner manufacturer and model identification , author=. Electronic Imaging , volume=. 2020 , publisher=

    2020

  12. [12]

    2022 International conference on computational science and computational intelligence (CSCI) , pages=

    Deep-RSI: Deep learning for Radiographs Source Identification , author=. 2022 International conference on computational science and computational intelligence (CSCI) , pages=. 2022 , organization=

    2022

  13. [13]

    IEEE Transactions on Information Forensics and Security , volume=

    Digital Camera Identification From Sensor Pattern Noise , author=. IEEE Transactions on Information Forensics and Security , volume=. 2006 , publisher=

    2006

  14. [14]

    2017 , publisher=

    MRI from Picture to Proton , author=. 2017 , publisher=

    2017

  15. [15]

    International conference on machine learning , pages=

    Leveraging Frequency Analysis for Deep Fake Image Recognition , author=. International conference on machine learning , pages=. 2020 , organization=

    2020

  16. [16]

    Advances in neural information processing systems , volume=

    Mlp-mixer: An ALL-ML Architecture for Vision , author=. Advances in neural information processing systems , volume=

  17. [17]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    ArcFace: Additive Angular Margin Loss for Deep Face Recognition , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  18. [18]

    IEEE journal of selected topics in signal processing , volume=

    Media Forensics and DeepFakes: An Overview , author=. IEEE journal of selected topics in signal processing , volume=. 2020 , publisher=

    2020

  19. [19]

    Information fusion , volume=

    Deepfakes and beyond: A Survey of face manipulation and fake detection , author=. Information fusion , volume=. 2020 , publisher=

    2020

  20. [20]

    International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=

    Toward Medical Deepfake Detection: A Comprehensive Dataset and Novel Method , author=. International Conference on Medical Image Computing and Computer-Assisted Intervention , pages=. 2025 , organization=

    2025

  21. [21]

    ACM Transactions on Intelligent Systems and Technology , volume=

    Back-in-Time Diffusion: Unsupervised Detection of Medical Deepfakes , author=. ACM Transactions on Intelligent Systems and Technology , volume=. 2025 , publisher=

    2025

  22. [22]

    Expert Systems with Applications , volume=

    DeepFake detection based on high-frequency enhancement network for highly compressed content , author=. Expert Systems with Applications , volume=. 2024 , publisher=

    2024

  23. [23]

    2021 IEEE 23rd International Workshop on Multimedia Signal Processing (MMSP) , pages=

    Leveraging High-Frequency Components for Deepfake Detection , author=. 2021 IEEE 23rd International Workshop on Multimedia Signal Processing (MMSP) , pages=. 2021 , organization=

    2021

  24. [24]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Frequency-Aware Deepfake Detection: Improving Generalizability through Frequency Space Domain Learning , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  25. [25]

    IEEE transactions on medical imaging , volume=

    Zero-Shot Medical Image Translation via Frequency-Guided Diffusion Models , author=. IEEE transactions on medical imaging , volume=. 2023 , publisher=

    2023

  26. [26]

    NeuroImage , volume=

    Unsupervised MR harmonization by learning disentangled representations using information bottleneck theory , author=. NeuroImage , volume=. 2021 , publisher=

    2021

  27. [27]

    Medical image analysis , volume=

    Disentangled representation learning in cardiac image analysis , author=. Medical image analysis , volume=. 2019 , publisher=

    2019

  28. [28]

    Science , volume=

    Adversarial attacks on medical machine learning , author=. Science , volume=. 2019 , publisher=

    2019

  29. [29]

    Cancers , volume=

    Adversarial Attacks on Medical Image Classification , author=. Cancers , volume=. 2023 , publisher=

    2023

  30. [30]

    American journal of roentgenology , volume=

    An Introduction to the Fourier Transform: Relationship to MRI , author=. American journal of roentgenology , volume=. 2008 , publisher=

    2008

  31. [31]

    Magnetic resonance in medicine , volume=

    Radiofrequency Pulse Design Using Nonlinear Gradient Magnetic Fields , author=. Magnetic resonance in medicine , volume=. 2015 , publisher=

    2015

  32. [32]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops , pages=

    DeepFake Detection by Analyzing Convolutional Traces , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops , pages=

  33. [33]

    ACM Transactions on Multimedia Computing, Communications and Applications , volume=

    Mastering Deepfake Detection: A Cutting-edge Approach to Distinguish GAN and Diffusion-model Images , author=. ACM Transactions on Multimedia Computing, Communications and Applications , volume=. 2024 , publisher=

    2024

  34. [34]

    2026 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) , pages=

    Deepfake Detection that Generalizes Across Benchmarks , author=. 2026 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) , pages=. 2026 , organization=

    2026

  35. [35]

    Expert Systems with Applications , volume=

    Enhanced Deepfake Detection with DenseNet and Cross-ViT , author=. Expert Systems with Applications , volume=. 2025 , publisher=

    2025

  36. [36]

    Yang, Yongqi and Qian, Zhihao and Zhu, Ye and Russakovsky, Olga and Wu, Yu , booktitle=. D\^

  37. [37]

    Journal of Imaging , volume=

    Fighting Deepfakes by Detecting GAN DCT Anomalies , author=. Journal of Imaging , volume=. 2021 , publisher=

    2021

  38. [38]

    Neuroimage , volume=

    A reproducible evaluation of ANTs similarity metric performance in brain image registration , author=. Neuroimage , volume=. 2011 , publisher=

    2011

  39. [39]

    Advances in neural information processing systems , volume=

    PyTorch: An Imperative Style, High-Performance Deep Learning Library , author=. Advances in neural information processing systems , volume=