pith. sign in

arxiv: 2605.19279 · v1 · pith:JBDVHPT7new · submitted 2026-05-19 · 💻 cs.CV

FPED: A Functional-Network Prior-Guided Mixture-of-Experts Framework for Interpretable Brain Decoding

Yudan Ren , Pengcheng Shi , Zihan Ma , Xiaowei He , Xiao Li This is my paper

Pith reviewed 2026-05-20 07:09 UTC · model grok-4.3

classification 💻 cs.CV
keywords fMRI decodingbrain networksmixture of expertssemantic reconstructioninterpretable AIfunctional connectivityvisual semantics
0
0 comments X

The pith

A prior-guided mixture-of-experts model treats brain functional networks as experts to decode visual semantics from fMRI with competitive performance and added interpretability.

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

The paper introduces FPED to overcome the flattening of fMRI signals that ignores brain network structure and other functional contributions. It builds a mixture-of-experts system where each expert corresponds to a distinct functional brain network, guided by neurobiological priors. Adaptive routing learns how these networks complement each other in processing high-level visual semantics. This yields reconstruction performance on par with larger models while using only 0.68 billion parameters. The routing patterns align with known brain functions, offering direct neuroscientific insights.

Core claim

FPED explicitly models different functional brain networks as specialized experts and employs adaptive routing to capture their complementary contributions to visual semantic understanding. Unlike conventional homogeneous decoding paradigms, the framework incorporates neurobiologically grounded priors to enable structured and interpretable network-level representation learning. This approach achieves highly competitive semantic reconstruction performance with only 0.68B parameters, and the learned routing dynamics reveal biologically meaningful correspondence between functional brain networks and modality-specific semantic processing.

What carries the argument

Mixture-of-experts framework with functional brain networks as experts and neurobiologically grounded priors guiding adaptive routing.

If this is right

  • Semantic reconstruction from fMRI can respect the brain's distributed network topology instead of flattening signals.
  • Routing dynamics provide a transparent view into how different brain networks contribute to visual understanding.
  • Smaller parameter counts suffice when models are structured around biological priors.
  • Brain decoding can serve as a bridge to develop more biologically inspired AI systems.

Where Pith is reading between the lines

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

  • This structure could allow better generalization across subjects by respecting individual network variations.
  • Similar expert modeling might apply to decoding other cognitive processes like language or memory.
  • Future work could test if these routing patterns predict behavioral measures of perception.

Load-bearing premise

That neurobiologically grounded priors can be used to structure a mixture-of-experts framework such that adaptive routing between functional network experts captures complementary contributions to visual semantic understanding without disrupting inherent brain topology.

What would settle it

A direct comparison showing that routing weights do not correlate with established functional connectivity maps from neuroscience, or that performance is not competitive when priors are removed.

Figures

Figures reproduced from arXiv: 2605.19279 by Pengcheng Shi, Xiao Li, Xiaowei He, Yudan Ren, Zihan Ma.

Figure 1
Figure 1. Figure 1: Overview of the FPED framework. Stage 1 establishes a mapping from multi￾network fMRI signals to CLIP embedding space. Stage 2 incorporates spatiotemporal routing mechanisms to enable dynamic gating of features across temporal brain activity variations. 3 Method Our proposed FPED framework comprises two sequential stages for progressive fMRI-to-image reconstruction, as illustrated in [PITH_FULL_IMAGE:figu… view at source ↗
Figure 2
Figure 2. Figure 2: Brain-to-image reconstruction performance. We employed multiple metrics to comprehensively evaluate reconstruction quality. Low-level features were assessed using PixCorr, SSIM, and AlexNet (L5), while high-level semantic features were evaluated using Inception, CLIP, EffNet￾B, and SwAV [18]. We compared against state-of-the-art baselines: Brain-Diffuser [35], Mind￾Bridge, and MindEye2 for brain decoding, … view at source ↗
Figure 3
Figure 3. Figure 3: Semantic heatmaps of brain-network-based experts. Each row displays the orig￾inal image alongside heatmaps from seven experts (V, SM, DA, VA, L, C, DM). Spatial Semantic Specialization Visualization of semantic focus ( [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of MoE routing weights across modalities. Modality-Specific Routing Behavior The quantification of routing weights ( [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
read the original abstract

Visual image reconstruction from functional Magnetic Resonance Imaging (fMRI) is a fundamental task in brain decoding, providing a crucial pathway for understanding human perceptual mechanisms and developing advanced brain-computer interfaces (BCIs). However, most current methods simply flatten fMRI signals from localized visual cortices into one-dimensional (1D) vectors, mapping them directly into latent spaces such as that of Contrastive Language-Image Pre-training (CLIP). This paradigm not only disrupts the inherent network topology of the brain-leading to limited neuroscientific interpretability-but also overlooks the synergistic contributions of other distributed functional networks in processing high-level visual semantics. To address these limitations, we propose FPED, a Functional-Network Prior-Guided Mixture of Experts (MoE) framework for interpretable brain decoding. FPED explicitly models different functional brain networks as specialized experts and employs adaptive routing to capture their complementary contributions to visual semantic understanding. Unlike conventional homogeneous decoding paradigms, our framework incorporates neurobiologically grounded priors to enable structured and interpretable network-level representation learning. Experimental results demonstrate that FPED achieves highly competitive semantic reconstruction performance with only 0.68B parameters. The learned routing dynamics reveal biologically meaningful correspondence between functional brain networks and modality-specific semantic processing, providing transparent neuroscientific interpretability. This suggests that brain network-aware expert modeling is a promising direction for bridging neural decoding and biologically inspired artificial intelligence.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces FPED, a Functional-Network Prior-Guided Mixture-of-Experts (MoE) framework for interpretable brain decoding from fMRI. It explicitly models different functional brain networks as specialized experts, employs adaptive routing guided by neurobiologically grounded priors to capture complementary contributions to visual semantic understanding, and reports competitive semantic reconstruction performance with only 0.68B parameters along with routing dynamics that reveal biologically meaningful correspondences to modality-specific semantic processing.

Significance. If the empirical results hold under rigorous validation, this work could meaningfully advance interpretable brain decoding by preserving brain network topology rather than flattening signals, while integrating distributed functional networks into decoding. The neurobiologically grounded MoE structure offers a concrete path toward more transparent and biologically plausible models for BCIs and perceptual mechanism studies.

major comments (2)
  1. [§4] §4 (Experimental Results): The central claim of 'highly competitive semantic reconstruction performance' with 0.68B parameters is load-bearing for the contribution, yet the manuscript supplies no quantitative metrics, baseline comparisons, error bars, or statistical tests in the visible experimental description, preventing evaluation of whether the result actually supports competitiveness or the interpretability gains.
  2. [§3.2] §3.2 (Routing Mechanism): The assumption that neurobiological priors structure the MoE routing to capture complementary contributions without disrupting inherent brain topology is central to the interpretability claim, but the manuscript does not provide a concrete test (e.g., ablation removing the prior or topology-preservation metric) to confirm this does not introduce artifacts.
minor comments (2)
  1. [Abstract] The abstract would benefit from a one-sentence mention of the specific datasets and evaluation metrics used to ground the performance claim.
  2. [§3] Notation for the expert routing function and prior incorporation should be introduced with a clear equation early in §3 to aid reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. The comments correctly identify areas where additional rigor is needed to support the performance and interpretability claims. We have revised the manuscript to address both major points as described below.

read point-by-point responses

  1. Referee: [§4] §4 (Experimental Results): The central claim of 'highly competitive semantic reconstruction performance' with 0.68B parameters is load-bearing for the contribution, yet the manuscript supplies no quantitative metrics, baseline comparisons, error bars, or statistical tests in the visible experimental description, preventing evaluation of whether the result actually supports competitiveness or the interpretability gains.

    Authors: We agree that the experimental results section requires more explicit quantitative support. In the revised manuscript we have expanded §4 to include specific semantic reconstruction metrics (CLIP similarity and other standard measures), direct numerical comparisons against recent baselines, error bars computed over multiple runs, and statistical significance tests. These additions allow direct evaluation of the competitiveness claim at the stated parameter count. revision: yes

  2. Referee: [§3.2] §3.2 (Routing Mechanism): The assumption that neurobiological priors structure the MoE routing to capture complementary contributions without disrupting inherent brain topology is central to the interpretability claim, but the manuscript does not provide a concrete test (e.g., ablation removing the prior or topology-preservation metric) to confirm this does not introduce artifacts.

    Authors: The referee correctly notes the absence of a direct validation test. We have added an ablation study that removes the neurobiological priors from the routing and reports the resulting change in both reconstruction performance and routing interpretability. We have also introduced a topology-preservation metric that compares learned routing weights against established functional connectivity patterns; the revised results show that the prior-guided routing improves rather than disrupts this alignment. revision: yes

Circularity Check

0 steps flagged

No significant circularity; framework is a modeling choice with empirical validation

full rationale

The paper introduces FPED as a new Mixture-of-Experts architecture that incorporates neurobiological priors for routing between functional-network experts. No equations, derivations, or parameter-fitting steps are presented that reduce any claimed prediction or result back to the inputs by construction. The performance claims and routing interpretations are reported as outcomes of experiments rather than theorems or self-referential definitions. Self-citations, if present, are not load-bearing for the central construction, and the approach remains self-contained as an independent modeling proposal against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract, the central claim rests on the domain assumption that functional brain networks provide useful priors for expert specialization and that routing dynamics will align with biological semantics.

axioms (1)
  • domain assumption Functional brain networks can be treated as specialized experts whose complementary contributions are captured by adaptive routing in an MoE architecture.
    This premise structures the entire framework and is invoked to justify improved interpretability over flattening methods.

pith-pipeline@v0.9.0 · 5791 in / 1243 out tokens · 46762 ms · 2026-05-20T07:09:17.859752+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

43 extracted references · 43 canonical work pages

  1. [1]

    arXiv preprint arXiv:2503.15978 (2025)

    Liu,P.,Dong,G.,Guo,D.,Li,K.,Li,F.,Yang,X.,Wang,M.,Ying,X.:Asurveyon fMRI-based brain decoding for reconstructing multimodal stimuli. arXiv preprint arXiv:2503.15978 (2025)

    work page arXiv 2025

  2. [2]

    Brain Sciences12(2), 228 (2022)

    Du, B., Cheng, X., Duan, Y., Ning, H.: fMRI brain decoding and its applications in brain-computer interface: A survey. Brain Sciences12(2), 228 (2022)

    work page 2022

  3. [3]

    NeuroImage56(2), 400–410 (2011)

    Naselaris, T., Kay, K.N., Nishimoto, S., Gallant, J.L.: Encoding and decoding in fMRI. NeuroImage56(2), 400–410 (2011)

    work page 2011

  4. [4]

    Journal of Integrative Neuroscience14(2), 155–168 (2015)

    Zafar, R., Malik, A.S., Kamel, N., Dass, S.C., Abdullah, J.M., Reza, F., Abdul Karim, A.H.: Decoding of visual information from human brain activity: A review of fMRI and EEG studies. Journal of Integrative Neuroscience14(2), 155–168 (2015)

    work page 2015

  5. [5]

    In: Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, pp

    Takagi, Y., Nishimoto, S.: High-resolution image reconstruction with latent diffu- sion models from human brain activity. In: Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, pp. 14453–14463 (2023)

    work page 2023

  6. [6]

    Brain Captioning: Decoding human brain activity into images and text, May 2023

    Ferrante, M., Ozcelik, F., Boccato, T., VanRullen, R., Toschi, N.: Brain cap- tioning: Decoding human brain activity into images and text. arXiv preprint arXiv:2305.11560 (2023)

    work page arXiv 2023

  7. [7]

    Advances in Neural Information Processing Systems 36, 24705–24728 (2023)

    Scotti, P., Banerjee, A., Goode, J., Shabalin, S., Nguyen, A., Dempster, A., Ver- linde, N., et al.: Reconstructing the mind’s eye: fMRI-to-image with contrastive learning and diffusion priors. Advances in Neural Information Processing Systems 36, 24705–24728 (2023)

    work page 2023

  8. [8]

    arXiv preprint arXiv:2412.19487 (2024) 14 Ren et al

    Wang, Z., Zhao, Z., Zhou, L., Nachev, P.: UniBrain: A unified model for cross- subject brain decoding. arXiv preprint arXiv:2412.19487 (2024) 14 Ren et al

    work page arXiv 2024

  9. [9]

    Proceedings of the National Academy of Sciences114(18), 4793– 4798 (2017)

    Bonner, M.F., Epstein, R.A.: Coding of navigational affordances in the human visual system. Proceedings of the National Academy of Sciences114(18), 4793– 4798 (2017)

    work page 2017

  10. [10]

    Artificial Intelligence Review42(2), 275–293 (2014)

    Masoudnia, S., Ebrahimpour, R.: Mixture of experts: A literature survey. Artificial Intelligence Review42(2), 275–293 (2014)

    work page 2014

  11. [11]

    The organization of the human cerebral cortex estimated by intrinsic functional con- nectivity

    Yeo BT, Krienen FM, Sepulcre J, Sabuncu MR, Lashkari D, Hollinshead M, Roff- man JL, Smoller JW, Zollei L., Polimeni JR, Fischl B, Liu H, Buckner RL. The organization of the human cerebral cortex estimated by intrinsic functional con- nectivity. J Neurophysiol 106(3):1125-65 (2011)

    work page 2011

  12. [12]

    Nature Neuroscience25(1), 116–126 (2022)

    Allen, E.J., St-Yves, G., Wu, Y., Breedlove, J.L., Prince, J.S., Dowdle, L.T., et al.: A massive 7T fMRI dataset to bridge cognitive neuroscience and artificial intelligence. Nature Neuroscience25(1), 116–126 (2022)

    work page 2022

  13. [13]

    arXiv preprint arXiv:2505.15946 (2025)

    Wei, Y., Zhang, Y., Xiao, X., Wang, T., Wang, X., Calhoun, V.D.: MORE-Brain: Routed mixture of experts for interpretable and generalizable cross-subject fMRI visual decoding. arXiv preprint arXiv:2505.15946 (2025)

    work page arXiv 2025

  14. [14]

    Communications of the ACM63(11), 139–144 (2020)

    Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., et al.: Generative adversarial networks. Communications of the ACM63(11), 139–144 (2020)

    work page 2020

  15. [15]

    IEEE Transactions on Neu- ral Networks and Learning Systems30(8), 2310–2323 (2018)

    Du, C., Du, C., Huang, L., He, H.: Reconstructing perceived images from human brain activities with Bayesian deep multiview learning. IEEE Transactions on Neu- ral Networks and Learning Systems30(8), 2310–2323 (2018)

    work page 2018

  16. [16]

    NeuroImage228, 117602 (2021)

    Ren, Z., Li, J., Xue, X., Li, X., Yang, F., Jiao, Z., Gao, X.: Reconstructing seen im- age from brain activity by visually-guided cognitive representation and adversarial learning. NeuroImage228, 117602 (2021)

    work page 2021

  17. [17]

    Advances in Neural Information Processing Systems35, 29624–29636 (2022)

    Lin, S., Sprague, T., Singh, A.K.: Mind reader: Reconstructing complex images from brain activities. Advances in Neural Information Processing Systems35, 29624–29636 (2022)

    work page 2022

  18. [18]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

    Wang, S., Liu, S., Tan, Z., Wang, X.: MindBridge: A cross-subject brain decoding framework. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 11333–11342 (2024)

    work page 2024

  19. [19]

    Mindeye2: Shared-subject models enable fmri-to-image with 1 hour of data.arXiv preprint arXiv:2403.11207, 2024

    Scotti, P.S., Tripathy, M., Villanueva, C.K.T., Kneeland, R., Chen, T., Narang, A., et al.: MindEye2: Shared-subject models enable fMRI-to-image with 1 hour of data. arXiv preprint arXiv:2403.11207 (2024)

    work page arXiv 2024

  20. [20]

    Brain Research Re- views61(2), 144–153 (2009)

    Farivar, R.: Dorsal-ventral integration in object recognition. Brain Research Re- views61(2), 144–153 (2009)

    work page 2009

  21. [21]

    Archives of General Psychiatry59(11), 1011–1020 (2002)

    Doniger, G.M., Foxe, J.J., Murray, M.M., Higgins, B.A., Javitt, D.C.: Impaired visual object recognition and dorsal/ventral stream interaction in schizophrenia. Archives of General Psychiatry59(11), 1011–1020 (2002)

    work page 2002

  22. [22]

    Trends in Cognitive Sciences26(12), 1119–1132 (2022)

    Ayzenberg, V., Behrmann, M.: Does the brain’s ventral visual pathway compute object shape?. Trends in Cognitive Sciences26(12), 1119–1132 (2022)

    work page 2022

  23. [23]

    Frontiers in Human Neuroscience15, 757128 (2022)

    Machner, B., Braun, L., Imholz, J., Koch, P.J., Münte, T.F., Helmchen, C., Sprenger, A.: Resting-state functional connectivity in the dorsal attention net- work relates to behavioral performance in spatial attention tasks and may show task-related adaptation. Frontiers in Human Neuroscience15, 757128 (2022)

    work page 2022

  24. [24]

    Journal of Cognitive Neuroscience33(6), 965–983 (2021)

    Rajan, A., Meyyappan, S., Liu, Y., Samuel, I.B.H., Nandi, B., Mangun, G.R., Ding,M.:Themicrostructureofattentionalcontrolinthedorsalattentionnetwork. Journal of Cognitive Neuroscience33(6), 965–983 (2021)

    work page 2021

  25. [25]

    Li, W., Mai, X., Liu, C.: The default mode network and social understanding of others:Whatdobrainconnectivitystudiestellus.FrontiersinHumanNeuroscience 8, 52017 (2014) FPED: Brain Decoding with MoE and Functional Networks 15

    work page 2014

  26. [26]

    Neuron73(3), 415–434 (2012)

    DiCarlo, J.J., Zoccolan, D., Rust, N.C.: How does the brain solve visual object recognition?. Neuron73(3), 415–434 (2012)

    work page 2012

  27. [27]

    Human Brain Mapping44(7), 2921–2935 (2023)

    Ye, Z., Qu, Y., Liang, Z., Wang, M., Liu, Q.: Explainable fMRI-based brain de- coding via spatial temporal-pyramid graph convolutional network. Human Brain Mapping44(7), 2921–2935 (2023)

    work page 2023

  28. [28]

    In2013 IEEE International Conference on Acoustics, Speech and Signal Processing, pp

    Yu, D., Kaisheng Yao, Hang Su, Gang Li, and Frank Seide: KL-divergence regular- ized deep neural network adaptation for improved large vocabulary speech recog- nition. In2013 IEEE International Conference on Acoustics, Speech and Signal Processing, pp. 7893–7897. IEEE (2013)

    work page 2013

  29. [29]

    arXiv preprint arXiv:2309.13850 (2023)

    Nguyen, H., Akbarian, P., Yan, F., Ho, N.: Statistical perspective of top-k sparse softmax gating mixture of experts. arXiv preprint arXiv:2309.13850 (2023)

    work page arXiv 2023

  30. [30]

    InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

    Fei, B., Lyu, Z., Pan, L., Zhang, J., Yang, W., Luo, T., Zhang, B., Dai, B.: Gener- ative diffusion prior for unified image restoration and enhancement. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 9935–9946 (2023)

    work page 2023

  31. [31]

    Wang, J., Yue, Z., Zhou, S., Chan, K.C., Loy, C.C.: Exploiting diffusion prior for real-world image super-resolution.International Journal of Computer Vision 132(12), 5929–5949 (2024)

    work page 2024

  32. [32]

    155–161 (2019)

    Minnema, G., Herbelot, A.: From brain space to distributional space: The perilous journeys of fMRI decoding.Proceedings of the 57th Annual Meeting of the Asso- ciation for Computational Linguistics: Student Research Workshop, pp. 155–161 (2019)

    work page 2019

  33. [33]

    IEEE Signal Processing Magazine42(5), 22–35 (2025)

    Zhou, X., Liu, C., Chen, Z., Wang, K., Ding, Y., Jia, Z., Wen, Q.: Brain foundation models: A survey on advancements in neural signal processing and brain discovery. IEEE Signal Processing Magazine42(5), 22–35 (2025)

    work page 2025

  34. [34]

    Wang, H., Lu, J., Li, H., Li, X.: ZEBRA: Towards Zero-Shot Cross-Subject Gen- eralization for Universal Brain Visual Decoding.arXiv preprint arXiv:2510.27128 (2025)

    work page arXiv 2025

  35. [35]

    16–26 (2023)

    Chen, X., Lei, B., Pun, C.M., Wang, S.: Brain diffuser: An end-to-end brain im- age to brain network pipeline.Chinese Conference on Pattern Recognition and Computer Vision (PRCV), pp. 16–26 (2023)

    work page 2023

  36. [36]

    Situating the default-mode network along a principal gradient of macroscale cortical organization

    Margulies DS, Ghosh SS, Goulas A, Falkiewicz M, Huntenburg JM, Langs G, Bezgin G, et al. Situating the default-mode network along a principal gradient of macroscale cortical organization. Proc Natl Acad Sci USA 113(44):12574-12579 (2016)

    work page 2016

  37. [37]

    Principles of neural science

    Kandel ER, Schwartz JH, Jessell TM, Siegelbaum S, Hudspeth AJ, Mack S, eds. Principles of neural science. Vol. 4. New York: McGraw-Hill (2000)

    work page 2000

  38. [38]

    Somatotopy in human primary motor and somatosensory hand representations revisited

    Hluštík P, Solodkin A, Gullapalli RP, Noll DC, Small SL. Somatotopy in human primary motor and somatosensory hand representations revisited. Cereb Cortex 11(4):312-321 (2001)

    work page 2001

  39. [39]

    Control of goal-directed and stimulus-driven attention in the brain

    Corbetta M, Shulman GL. Control of goal-directed and stimulus-driven attention in the brain. Nat Rev Neurosci 3(3):201-215 (2002)

    work page 2002

  40. [40]

    Contributions of the amygdala to emotion processing: from animal models to human behavior

    Phelps EA, LeDoux JE. Contributions of the amygdala to emotion processing: from animal models to human behavior. Neuron 48(2):175-187 (2005)

    work page 2005

  41. [41]

    Distinct brain networks for adaptive and stable task control in humans

    Dosenbach NUF, Fair DA, Miezin FM, Cohen AL, Wenger KK, Dosenbach RAT, Fox MD, et al. Distinct brain networks for adaptive and stable task control in humans. Proc Natl Acad Sci USA 104(26):11073-11078 (2007)

    work page 2007

  42. [42]

    When brain-inspired AI meets AGI

    Zhao L, Zhang L, Wu Z, Chen Y, Dai H, Yu X, Liu Z, et al. When brain-inspired AI meets AGI. Meta-Radiology 1(1):100005 (2023)

    work page 2023

  43. [43]

    BrainMCLIP: Brain image decoding with multi-layer feature fusion of CLIP

    Xia T, Ma Z, Zhang Y, Wang X, Liu Q, He X, Ren Y. BrainMCLIP: Brain image decoding with multi-layer feature fusion of CLIP. Meta-Radiology 100219 (2026)

    work page 2026