MambaKAN: An Interpretable Framework for Alzheimer's Disease Diagnosis via Selective State Space Modeling of Dynamic Functional Connectivity.

Code ↔ Paper

17 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 17 matches

[1] § 3. Methods › 3.5. Joint Training Strategy › 3.5.1. Two-Phase Training Protocol ↔ train_stage1.py, lines 1–30 · score 0.92 · lowest validation reconstruction, VAE Unsupervised Pre, noise robust latent, best checkpoint, pre training, Adam
[2] § 3. Methods › 3.4. Kolmogorov–Arnold Network Classifier › 3.4.1. KAN Architecture Motivation ↔ models/kan_classifier.py, lines 1–17 · score 0.84 · Kolmogorov Arnold Networks, learnable weights, visualizable curve, intrinsically interpretable, inspired, MLPs
[3] § 3. Methods › 3.3. Mamba Selective State Space Temporal Encoder › 3.3.3. Selective State Space Model (S6) ↔ models/mamba_encoder.py, lines 13–22 · score 0.82 · Selective State Space, depthwise convolution, selectivity mechanism, Mamba block, S6, Model
[4] § 3. Methods › 3.5. Joint Training Strategy › 3.5.1. Two-Phase Training Protocol ↔ train_stage2.py, lines 1–44 · score 0.76 · End Joint Fine, trained jointly, MambaKAN, tuning, pipeline, accuracy
[5] § 3. Methods › 3.4. Kolmogorov–Arnold Network Classifier › 3.4.5. MambaKAN Classifier Structure ↔ models/kan_classifier.py, lines 180–280 · score 0.73 · layer KAN, context vector, Mamba encoder, KAN classifier, learnable, weights
[6] § 3. Methods › 3.1. Overview of MambaKAN ↔ train_stage2.py, lines 1–44 · score 0.72 · VAE encoding, MambaKAN, KAN classification, Mamba temporal, pipeline, joint
[7] § 5. Interpretability Analysis › 5.3. Layer 3: Gradient-Based Brain Region Attribution ↔ analysis.py, lines 230–357 · score 0.67 · bar charts, attribution scores, chord, heatmap, connectivity, Gradient
[8] § 3. Methods › 3.5. Joint Training Strategy › 3.5.2. Differential Learning Rates ↔ train_stage2.py, lines 46–74 · score 0.63 · warmup epochs, fine tuning, frozen, trained, KAN, Mamba
[9] § 3. Methods › 3.3. Mamba Selective State Space Temporal Encoder › 3.3.1. Rationale for Mamba over LSTM and Transformer ↔ models/mamba_encoder.py, lines 152–179 · score 0.63 · parallel scan, hardware aware, efficient, matrices, Selective, Mamba
[10] § 3. Methods › 3.4. Kolmogorov–Arnold Network Classifier › 3.4.2. Rationale for KAN over MLP ↔ models/kan_classifier.py, lines 1–17 · score 0.63 · provides intrinsic interpretability, latent dimension, MLP, mapping, linear, logits
[11] § 5. Interpretability Analysis › 5.3. Layer 3: Gradient-Based Brain Region Attribution ↔ analysis.py, lines 230–357 · score 0.59 · chord diagram, attribution scores, connectivity, Gradient, Brain, class
[12] § 3. Methods › 3.4. Kolmogorov–Arnold Network Classifier › 3.4.3. B-Spline Edge Activations ↔ models/kan_classifier.py, lines 29–66 · score 0.59 · spline coefficients, spline basis functions, uniform, learnable, weight
[13] § 3. Methods › 3.3. Mamba Selective State Space Temporal Encoder › 3.3.5. Temporal Context Aggregation ↔ models/mamba_encoder.py, lines 182–235 · score 0.58 · stacked Mamba blocks, temporal context, sequence, vector
[14] § 5. Interpretability Analysis › 5.3. Layer 3: Gradient-Based Brain Region Attribution ↔ analysis.py, lines 1–41 · score 0.55 · attribution matrix, pairwise, ROI, brain, map, Gradient
[15] § 3. Methods › 3.2. Variational Autoencoder for Per-Window Feature Extraction › 3.2.3. VAE Loss Function ↔ models/vae.py, lines 51–55 · score 0.54 · KL divergence, reconstruction loss, VAE
[16] § 3. Methods › 3.5. Joint Training Strategy › 3.5.3. Regularization ↔ models/mamba_encoder.py, lines 13–22 · score 0.54 · depthwise convolution, Mamba block, Dropout, space
[17] § 4. Experiments › 4.2. Evaluation Metrics ↔ analysis.py, lines 1–41 · score 0.52 · ROC curve, MambaKAN, AUC

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The paper is loaded by your browser, with JavaScript: read it at doi.org or on Europe PMC.

The authors' code

Python · 297 lines · 10 KB · no license · 4 matches

The registry keeps no copy of this file: its repository has no license, so its authors keep all their rights to it. Your browser shows it from its source, with JavaScript.

It can be read at the source: models/kan_classifier.py.

Overview

Authors: Libin Gao¹, Zhongyi Hu²

ORCID iDs: Libin Gao, Zhongyi Hu

Institutions: Zhejiang Industry & Trade Vocational College (China); Wenzhou University (China)

Journal: Brain sciences, volume 16, issue 4, article 421

Dates: received 25 March 2026; accepted 16 April 2026; published online 17 April 2026

Type: Research article · Language: English

License: CC BY

Identifiers: DOI 10.3390/brainsci16040421 · PMID 42041829 · PMCID PMC13114612 · OpenAlex W7154712876

Open access: gold, a free copy (OpenAlex)

Status: code verified

Categories: fMRI (modality), Alzheimer's / dementia (population)

Methods: Spectral & time-frequency, Statistics, Preprocessing, Connectivity, Machine learning, fMRI & imaging

Keywords: Alzheimer’s disease, dynamic functional connectivity, state space model, Mamba, Kolmogorov–Arnold network, variational autoencoder, interpretability, resting-state fMRI, deep learning, brain connectivity

Topic: Functional Brain Connectivity Studies (Cognitive Neuroscience, Neuroscience), according to OpenAlex

Funding: Wenzhou major scientific and technological innovation project plan “reveal the list” (ZG2023022); Wenzhou Fundamental Scientific Research Program (Y20240185, G20240065); General Program of Wenzhou Municipal Science and Technology Bureau (S2023013); General Program of Zhejiang Provincial Department of Education (Y202250720); Teacher Scientific Research Project of Zhejiang Industry & Trade Vocational College (G250101)

Citations: cited by 1 paper (Europe PMC); 24 references in the paper

Abstract

Background/Objectives: Alzheimer’s disease (AD) is an irreversible neurodegenerative disorder that imposes a profound burden on global public health. While resting-state functional magnetic resonance imaging (rs-fMRI)-based dynamic functional connectivity (dFC) analysis has demonstrated promise in capturing time-varying brain network abnormalities, existing deep learning methods suffer from three fundamental limitations: (1) an inability to model temporal dependencies across dynamic connectivity windows, (2) reliance on post hoc black-box explainability tools, and (3) misalignment between feature learning and classification objectives. Methods: To address these challenges, we propose MambaKAN, an end-to-end interpretable framework integrating a Variational Autoencoder (VAE), a Selective State Space Model (Mamba), and a Kolmogorov–Arnold Network (KAN). The VAE encodes each dFC snapshot into a compact latent representation, preserving nonlinear connectivity patterns. The Mamba encoder captures long-range temporal dynamics across the sequence of latent representations via input-selective state transitions. The KAN classifier provides intrinsic interpretability through learnable B-spline activation functions, enabling direct visualization of how latent features influence diagnostic decisions without post-hoc approximation. The entire pipeline is trained end-to-end with a joint loss function that aligns feature learning with classification. Results: Evaluated on the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset across five classification tasks (CN vs. AD, CN vs. EMCI, EMCI vs. LMCI, LMCI vs. AD, and four-class), MambaKAN achieves accuracies of 95.1%, 89.8%, 84.0%, 86.7%, and 70.5%, respectively, outperforming strong baselines including LSTM, Transformer, and MLP-based variants. Conclusions: Comprehensive ablation studies confirm the indispensable contribution of each module, and the three-layer interpretability analysis reveals key temporal patterns and brain regions associated with AD progression.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repository

Its files are read in the Code ↔ Paper reader above, with 17 matches between paragraphs and lines of code.

l1binn/MambaKAN

License: none: the authors keep all their rights

State: the link answers, verified on 29 September 2026

Evidence: files inventoried

Commit: 0ff30c0a3602f29e7eb6b90543f4ef7978b34707, 17 April 2026

Languages: Python (9)

Size: 11 files, 9 scripts

Software Heritage: not archived

Found in: “Data Availability Statement”

Holds: README

Not found: license file, CITATION.cff, environment file, tests, continuous integration, documentation

Tools: PyTorch (8 files), Matplotlib (1 file), NumPy (1 file), scikit-learn (1 file), SciPy (1 file)

Availability: 1 check, the latest on 29 September 2026: the link answers

29 September 2026: the link answers

10 files, not copied: shown from their source

OSCR keeps no copy of these files: this repository has no license that allows it. The reader above shows each one from its source, fetched by your browser at commit 0ff30c0, when its fingerprint is the one OSCR verified. How this works.

analysis.py — Python, 661 lines, 4 matches, shown from its source
demo.py — Python, 107 lines, shown from its source
models/__init__.py — Python, 4 lines, shown from its source
models/kan_classifier.py — Python, 297 lines, 4 matches, shown from its source
models/mamba_encoder.py — Python, 249 lines, 4 matches, shown from its source
models/proposed.py — Python, 215 lines, shown from its source
models/vae.py — Python, 65 lines, 1 match, shown from its source
train_stage1.py — Python, 133 lines, 1 match, shown from its source
train_stage2.py — Python, 263 lines, 3 matches, shown from its source
README.md — Text, 261 lines, shown from its source

At the source: github.com/l1binn/MambaKAN

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
9 scripts, each with its path and the digest of its content;
17 matches between paragraphs of the paper and lines of the code (method lexical-v1);
neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data Availability Statement

The ADNI dataset is publicly available at https://adni.loni.usc.edu/ (accessed on 14 April 2026) upon registration. The core model and algorithm code is publicly available at https://github.com/l1binn/MambaKAN.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 29 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 2 authors, 10 keywords, 5 funders, 20 references.

Cite

This paper

Gao, L., & Hu, Z. (2026). MambaKAN: An Interpretable Framework for Alzheimer's Disease Diagnosis via Selective State Space Modeling of Dynamic Functional Connectivity. Brain sciences, 16(4), 421. https://doi.org/10.3390/brainsci16040421

BibTeX

@article{gao2026mambakan, author = {Gao, Libin and Hu, Zhongyi}, title = {{MambaKAN: An Interpretable Framework for Alzheimer's Disease Diagnosis via Selective State Space Modeling of Dynamic Functional Connectivity}}, journal = {Brain sciences}, year = {2026}, month = apr, volume = {16}, number = {4}, pages = {421}, publisher = {Multidisciplinary Digital Publishing Institute (MDPI)}, issn = {2076-3425}, doi = {10.3390/brainsci16040421}, url = {https://doi.org/10.3390/brainsci16040421}, pmid = {42041829}, pmcid = {PMC13114612} }

RIS

TY - JOUR AU - Gao, Libin AU - Hu, Zhongyi TI - MambaKAN: An Interpretable Framework for Alzheimer's Disease Diagnosis via Selective State Space Modeling of Dynamic Functional Connectivity T2 - Brain sciences J2 - Brain Sci PY - 2026 DA - 2026/04/17 VL - 16 IS - 4 SP - 421 SN - 2076-3425 PB - Multidisciplinary Digital Publishing Institute (MDPI) DO - 10.3390/brainsci16040421 UR - https://doi.org/10.3390/brainsci16040421 LA - en ER -

CSL-JSON

{ "id": "10.3390/brainsci16040421", "type": "article-journal", "title": "MambaKAN: An Interpretable Framework for Alzheimer's Disease Diagnosis via Selective State Space Modeling of Dynamic Functional Connectivity", "container-title": "Brain sciences", "author": [ { "family": "Gao", "given": "Libin" }, { "family": "Hu", "given": "Zhongyi" } ], "container-title-short": "Brain Sci", "volume": "16", "issue": "4", "page": "421", "DOI": "10.3390/brainsci16040421", "PMID": "42041829", "PMCID": "PMC13114612", "ISSN": "2076-3425", "publisher": "Multidisciplinary Digital Publishing Institute (MDPI)", "URL": "https://doi.org/10.3390/brainsci16040421", "language": "en", "issued": { "date-parts": [ [ 2026, 4, 17 ] ] } }

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.