pith. sign in

arxiv: 2606.07798 · v1 · pith:GMPXZ344new · submitted 2026-06-05 · 💻 cs.AI · cs.LG· q-bio.NC

Reconstructing and forecasting disease trajectories of patients with Alzheimer's disease using routine data in resource-constrained settings

Ratnadeep Das , Atri Chatterjee , Sitikantha Roy This is my paper

Pith reviewed 2026-06-27 21:35 UTC · model grok-4.3

classification 💻 cs.AI cs.LGq-bio.NC
keywords Alzheimer's diseasedisease trajectory forecastingneural ODEvariational autoencodercognitive score predictionirregular time seriesresource-constrained settingsGRU encoder
0
0 comments X

The pith

A GRU-Neural ODE variational autoencoder reconstructs past and forecasts future Alzheimer's cognitive scores from irregular routine visits alone.

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

The paper establishes that a single neural architecture can handle irregular clinical visit data to both fill in missing past cognitive states and predict future ones for Alzheimer's patients. It does so without any imaging or biomarker inputs, while also producing uncertainty estimates around each prediction. A sympathetic reader would care because this approach relies only on data collected in ordinary doctor visits, which opens the possibility of wider use where expensive scans are unavailable. The model was tested on 1,727 patients tracked for up to ten years in the ADNI dataset, reaching mean absolute errors of 1.35 on the CDR-SB scale and 2.28 on the MMSE scale.

Core claim

The GNOVA framework combines a GRU encoder that accepts any number of visits at irregular times, a Neural ODE decoder that produces continuous trajectories for interpolation and extrapolation, and a variational autoencoder component that supplies calibrated uncertainty. This unified model performs bidirectional prediction of CDR-SB and MMSE scores using only routine clinical variables, achieving the stated errors on the full ADNI cohort while identifying age, BMI, and APOE4 status as the strongest predictors through ablation.

What carries the argument

GNOVA (GRU-Neural ODE Variational Autoencoder): a hybrid architecture whose GRU encoder processes variable-length irregular inputs, Neural ODE decoder generates continuous-time estimates, and VAE layer quantifies predictive uncertainty.

If this is right

  • Clinicians can reconstruct incomplete patient histories from past routine visits.
  • Future cognitive states can be anticipated at any chosen future time point.
  • Each prediction comes with a well-calibrated uncertainty interval derived from the variational component.
  • No neuroimaging or CSF biomarkers are required, lowering the barrier for deployment.
  • Age, BMI, and APOE4 status emerge as the dominant routine predictors after feature ablation.

Where Pith is reading between the lines

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

  • The same architecture could be retrained on local clinic records to test whether the reported errors survive domain shift.
  • Adding other routinely collected variables such as blood pressure or medication lists might further reduce error without new modalities.
  • The continuous-time decoder suggests the framework could be extended to irregular multi-disease trajectories if similar longitudinal data become available.

Load-bearing premise

Results obtained on the ADNI research cohort will hold for patients in actual resource-constrained clinics whose visit schedules, demographics, and data quality differ from the study set.

What would settle it

Apply the trained GNOVA model to an independent cohort collected from a low-resource clinic and measure whether the mean absolute errors remain at or below 1.35 for CDR-SB and 2.28 for MMSE.

Figures

Figures reproduced from arXiv: 2606.07798 by Atri Chatterjee, Ratnadeep Das, Sitikantha Roy.

Figure 1
Figure 1. Figure 1: GNOVA Architecture The complete architecture consists of four blocks. The details of each block are given below. (i) Static Encoder Block: It consists of a multi-layer perceptron (MLP) that nonlinearly trans￾forms the 14-dimensional baseline static feature vector xstatic into a fixed-dimensional represen￾tation hstatic (encoder_hidden_dim), using ReLU activation. 6 [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Trajectory of the patients with Input time points - 0,1 [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Trajectory of the patients with Input time points - 2,5 [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Trajectory of the patients with Input time points - 6,7 [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
read the original abstract

Alzheimer's disease is a progressive neurodegenerative disorder, and its progression varies substantially across patients. Existing work aims to forecast patients' future cognitive state, with minimal focus on reconstructing the state from past visits. Furthermore, in current research, quantifying predictive uncertainty remains underexplored and relies on costly modalities such as MRI, PET, and CSF, limiting their deployment in resource-limited settings. In this research, our primary objectives are: First, bidirectional prediction of cognitive scores from irregular visits to present the complete disease trajectory. Second, to enable interpolation and extrapolation capabilities to assist clinicians in informed prognostic decision making, and third, to provide a well-calibrated uncertainty estimate for all predictions, and finally, to achieve the objectives using the modalities available during routine visits. We propose a unified framework, GNOVA: A GRU-Neural ODE Variational Autoencoder. The architecture combines a Gated Recurrent Unit encoder and a Neural ODE decoder within a variational autoencoder framework. In our work, we forecast the CDR-SB and MMSE Scores. The GRU encoder allows for any number of inputs at any time point. The Neural-ODE decoder performs continuous estimation, allowing interpolation and extrapolation at any desired time point. The Variational autoencoder allows for uncertainty estimation in predictions. We worked with 1,727 patients from the ADNI dataset over 10 years; the model achieved mean absolute errors of 1.35 and 2.28 for CDR-SB and MMSE scores, respectively, without requiring any neuroimaging or biomarker data. Feature-ablation studies revealed that age, BMI, and APOE4 status were strong predictors. The proposed framework enables the reconstruction of incomplete patient histories and the anticipation of future cognitive states.

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

3 major / 1 minor

Summary. The manuscript proposes GNOVA, a GRU-Neural ODE Variational Autoencoder, for reconstructing and forecasting Alzheimer's disease trajectories via CDR-SB and MMSE scores. It processes irregular longitudinal visits from 1,727 ADNI patients over 10 years, claims MAEs of 1.35 (CDR-SB) and 2.28 (MMSE) using only routine-visit modalities without neuroimaging or biomarkers, supplies uncertainty estimates via the VAE, and identifies age, BMI, and APOE4 status as strong predictors via ablation.

Significance. If the performance claims are verified with proper held-out evaluation and the inputs are restricted to truly routine data, the continuous-time bidirectional modeling could support prognostic decisions in low-resource clinics. The combination of GRU encoding for variable-length sequences and Neural ODE decoding for interpolation/extrapolation addresses a practical gap in longitudinal AD modeling; the explicit uncertainty quantification is a further strength.

major comments (3)
  1. [Abstract] Abstract: the central claim that predictions are obtained 'without requiring any neuroimaging or biomarker data' and 'using the modalities available during routine visits' is placed in tension by the subsequent statement that feature-ablation studies identified APOE4 status as a strong predictor. APOE genotyping is not part of standard routine clinical encounters in resource-constrained settings; the manuscript must clarify whether APOE4 is an input feature and, if so, how the 'routine data only' premise is maintained.
  2. [Abstract] Abstract / Evaluation section: no information is supplied on the train/test split, cross-validation procedure, temporal ordering of visits, or missing-data imputation strategy. Without these details the reported MAEs cannot be assessed for leakage or overfitting and therefore cannot support the performance claim.
  3. [Results] Results / Discussion: the manuscript reports performance on the ADNI research cohort but contains no external validation set or domain-shift experiments that would test generalization to the different visit frequencies, demographics, and data quality typical of resource-constrained clinical settings.
minor comments (1)
  1. [Abstract] Abstract: the phrasing 'In our work, we forecast the CDR-SB and MMSE Scores' should be expanded to make explicit that both reconstruction (interpolation) and forecasting (extrapolation) are performed.

Simulated Author's Rebuttal

3 responses · 1 unresolved

We thank the referee for the constructive feedback. We address each major comment point-by-point below, indicating where revisions will be made to improve clarity and rigor.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that predictions are obtained 'without requiring any neuroimaging or biomarker data' and 'using the modalities available during routine visits' is placed in tension by the subsequent statement that feature-ablation studies identified APOE4 status as a strong predictor. APOE genotyping is not part of standard routine clinical encounters in resource-constrained settings; the manuscript must clarify whether APOE4 is an input feature and, if so, how the 'routine data only' premise is maintained.

    Authors: We agree this creates an inconsistency. APOE4 status is an input feature in the reported experiments because it was available in ADNI and ablation confirmed its predictive value. To resolve the tension, we will revise the abstract, methods, and results to explicitly distinguish core routine inputs (age, BMI, visit history, demographics) from the optional APOE4 genetic marker. We will also add performance numbers from the ablation excluding APOE4, showing that the model remains functional with strictly routine data. This preserves the resource-constrained focus while being transparent. revision: yes

  2. Referee: [Abstract] Abstract / Evaluation section: no information is supplied on the train/test split, cross-validation procedure, temporal ordering of visits, or missing-data imputation strategy. Without these details the reported MAEs cannot be assessed for leakage or overfitting and therefore cannot support the performance claim.

    Authors: We concur that these details are required to evaluate validity. The revised manuscript will add a dedicated Evaluation subsection specifying: patient-level train/test split (70/30, no patient overlap), strict temporal ordering (only past visits used for prediction to prevent leakage), 5-fold cross-validation on the training set, and the model's native handling of irregular/missing visits via GRU encoding and Neural ODE decoding (no separate imputation step). These additions will allow direct assessment of the reported MAEs. revision: yes

  3. Referee: [Results] Results / Discussion: the manuscript reports performance on the ADNI research cohort but contains no external validation set or domain-shift experiments that would test generalization to the different visit frequencies, demographics, and data quality typical of resource-constrained clinical settings.

    Authors: This is a substantive limitation for claims about resource-constrained deployment. The present work establishes feasibility on the publicly available ADNI cohort. In revision we will expand the Discussion to analyze likely domain shifts (sparser visits, demographic differences, noisier data) and state the absence of external validation as a clear limitation, with explicit suggestions for future validation on clinical cohorts. New external experiments cannot be performed within the current study scope. revision: partial

standing simulated objections not resolved
  • External validation set or domain-shift experiments on independent datasets from resource-constrained clinical settings

Circularity Check

0 steps flagged

No circularity; empirical ML performance on held-out data

full rationale

The paper introduces the GNOVA architecture (GRU encoder + Neural ODE decoder in a VAE) and reports empirical MAEs of 1.35 (CDR-SB) and 2.28 (MMSE) on the ADNI cohort of 1,727 patients. These are presented as results of training and evaluation on held-out data rather than any closed-form derivation. No equations are shown that reduce a claimed prediction to a fitted input by construction, no self-citations are invoked as load-bearing uniqueness theorems, and feature-ablation results are post-hoc analyses rather than the basis of the primary performance numbers. The modeling pipeline is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The performance claim rests on the representativeness of ADNI data for resource-constrained settings and on the assumption that the VAE component produces well-calibrated uncertainty without additional verification details.

free parameters (2)
  • neural network weights and biases
    All parameters of the GRU, ODE, and VAE components are fitted to the ADNI training data.
  • hyperparameters for training and architecture
    Learning rate, hidden sizes, and ODE solver settings chosen during model development.
axioms (2)
  • domain assumption ADNI cohort distribution is representative of resource-constrained clinical populations
    Paper claims applicability to resource-constrained settings while training and testing exclusively on ADNI.
  • domain assumption Irregular visit times can be treated as continuous-time observations without additional bias correction
    Neural ODE decoder assumes continuous dynamics from discrete irregular inputs.

pith-pipeline@v0.9.1-grok · 5858 in / 1364 out tokens · 30271 ms · 2026-06-27T21:35:52.518060+00:00 · methodology

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

43 extracted references · 25 canonical work pages

  1. [1]

    Evans, K

    S. Evans, K. McRae-McKee, C. Hadjichrysanthou, M. M. Wong, D. Ames, O. Lopez, F. de Wolf, R. M. Anderson, Alzheimer’s disease progression and risk factors: A standardized comparison between six large data sets, Alzheimer’s & Dementia: Translational Research & Clinical Inter- ventions 5 (2019) 515–523.doi:10.1016/j.trci.2019.04.005

    work page doi:10.1016/j.trci.2019.04.005 2019

  2. [2]

    URLhttps://www.who.int/news-room/fact-sheets/detail/dementia

    World Health Organization, Dementia, accessed: 2025-05-19 (2024). URLhttps://www.who.int/news-room/fact-sheets/detail/dementia

    2025

  3. [3]

    Blennow, A review of fluid biomarkers for alzheimer’s disease: Moving from csf to blood, Neurology and Therapy 6 (2017) 15–24.doi:10.1007/s40120-017-0073-9

    K. Blennow, A review of fluid biomarkers for alzheimer’s disease: Moving from csf to blood, Neurology and Therapy 6 (2017) 15–24.doi:10.1007/s40120-017-0073-9

    work page doi:10.1007/s40120-017-0073-9 2017

  4. [4]

    Zhang, Y

    J. Zhang, Y. Zhang, J. Wang, Y. Xia, J. Zhang, L. Chen, Recent advances in alzheimer’s disease: mechanisms, clinical trials and new drug development strategies, Signal Transduction and Targeted Therapy 9 (2024) 211.doi:10.1038/s41392-024-01911-3

    work page doi:10.1038/s41392-024-01911-3 2024

  5. [5]

    Schaap, P

    T. Schaap, P. Thropp, D. Tosun, Timing of alzheimer’s disease biomarker progressions: A two-decade observational study from the alzheimer’s disease neuroimaging initiative (adni), Alzheimer’s and Dementia (12 2024).doi:10.1002/alz.14306. 23

    work page doi:10.1002/alz.14306 2024

  6. [6]

    V. J. Lowe, C. T. Mester, E. S. Lundt, J. Lee, S. Ghatamaneni, A. Algeciras-Schimnich, M. R. Campbell, J. Graff-Radford, A. Nguyen, H. Min, M. L. Senjem, M. M. Machulda, C. G. Schwarz, D. W. Dickson, M. E. Murray, K. K. Kandimalla, K. Kantarci, B. Boeve, P. Vemuri, D. T. Jones, D. Knopman, C. R. Jack, R. C. Petersen, M. M. Mielke, Amyloid pet detects the ...

    work page doi:10.1002/alz.14317 2024

  7. [7]

    Tzeng, Y.-W

    R.-C. Tzeng, Y.-W. Yang, K.-C. Hsu, H.-T. Chang, P.-Y. Chiu, Sum of boxes of the clinical dementia rating scale highly predicts conversion or reversion in predementia stages, Frontiers in Aging Neuroscience 14 (9 2022).doi:10.3389/fnagi.2022.1021792

    work page doi:10.3389/fnagi.2022.1021792 2022

  8. [8]

    mini-mental state

    M. F. Folstein, S. E. Folstein, P. R. McHugh, “mini-mental state”: A practical method for grading the cognitive state of patients for the clinician, Journal of Psychiatric Research 12 (1975) 189– 198.doi:10.1016/0022-3956(75)90026-6

    work page doi:10.1016/0022-3956(75)90026-6 1975

  9. [9]

    Mouchet, K

    J. Mouchet, K. A. Betts, M. V. Georgieva, R. Ionescu-Ittu, L. M. Butler, X. Teitsma, P. Delmar, T. Kulalert, J. Zhu, N. Lema, U. Desai, Classification, prediction, and concordance of cognitive and functional progression in patients with mild cognitive impairment in the united states: A latent class analysis, Journal of Alzheimer’s Disease 82 (2021) 1667–1...

    2021

  10. [10]

    B. Lei, M. Yang, P. Yang, F. Zhou, W. Hou, W. Zou, X. Li, T. Wang, X. Xiao, S. Wang, Deep and joint learning of longitudinal data for alzheimer’s disease prediction, Pattern Recognition 102 (6 2020).doi:10.1016/j.patcog.2020.107247

    work page doi:10.1016/j.patcog.2020.107247 2020

  11. [11]

    Jiang, Y

    S. Jiang, Y. Xie, G. A. Colditz, Functional ensemble survival tree: Dynamic prediction of alzheimer’s disease progression accommodating multiple time-varying covariates, Journal of the Royal Statistical Society. Series C: Applied Statistics 70 (2021) 66–79.doi:10.1111/rssc. 12449

    work page doi:10.1111/rssc 2021

  12. [12]

    Devanarayan, Y

    V. Devanarayan, Y. Ye, A. Charil, E. Andreozzi, P. Sachdev, D. A. Llano, L. Tian, L. Zhu, H. Hampel, L. Kramer, S. Dhadda, M. Irizarry, Predicting clinical progression trajectories of early alzheimer’s disease patients, Alzheimer’s and Dementia 20 (2024) 1725–1738.doi:10. 1002/alz.13565

    2024

  13. [13]

    Tabarestani, M

    S. Tabarestani, M. Aghili, M. Eslami, M. Cabrerizo, A. Barreto, N. Rishe, R. E. Curiel, D. Loewenstein, R. Duara, M. Adjouadi, A distributed multitask multimodal approach for the prediction of alzheimer’s disease in a longitudinal study, NeuroImage 206 (2 2020).doi: 10.1016/j.neuroimage.2019.116317

    work page doi:10.1016/j.neuroimage.2019.116317 2020

  14. [14]

    Peterson, O

    K. Peterson, O. Rudovic, R. Guerrero, R. W. Picard, Personalized gaussian processes for future prediction of alzheimer’s disease progression (2018).arXiv:1712.00181. URLhttps://arxiv.org/abs/1712.00181

    Pith/arXiv arXiv 2018

  15. [15]

    C. Puri, G. Kooijman, B. Vanrumste, S. Luca, Forecasting time series in healthcare with gaussian processes and dynamic time warping based subset selection, IEEE Journal of Biomedical and Health Informatics 26 (2022) 6126–6137.doi:10.1109/JBHI.2022.3214343

    work page doi:10.1109/jbhi.2022.3214343 2022

  16. [16]

    M. Liu, J. Zhang, C. Lian, D. Shen, Weakly supervised deep learning for brain disease prognosis using mri and incomplete clinical scores, IEEE Transactions on Cybernetics 50 (2020) 3381–3392. doi:10.1109/TCYB.2019.2904186. 24

    work page doi:10.1109/tcyb.2019.2904186 2020

  17. [17]

    Morar, H

    U. Morar, H. Martin, W. Izquierdo, P. Forouzannezhad, E. Zarafshan, R. E. Curiel, M. Roselli, D. Loewenstein, R. Duara, E. Unger, M. Adjouadi, A deep-learning approach for the prediction of mini-mental state examination scores in a multimodal longitudinal study, in: Proceedings - 2020 International Conference on Computational Science and Computational Int...

    arXiv 2020

  18. [18]

    Z. Yuan, X. Li, Z. Hao, Z. Tang, X. Yao, T. Wu, Intelligent prediction of alzheimer’s disease via improved multifeature squeeze-and-excitation-dilated residual network, Scientific Reports 14 (12 2024).doi:10.1038/s41598-024-62712-w

    work page doi:10.1038/s41598-024-62712-w 2024

  19. [19]

    Goodfellow, Y

    I. Goodfellow, Y. Bengio, A. Courville, Deep Learning, MIT Press, 2016. URLhttp://www.deeplearningbook.org

    2016

  20. [20]

    K. Cho, B. van Merrienboer, C. Gulcehre, D. Bahdanau, F. Bougares, H. Schwenk, Y. Bengio, Learning phrase representations using rnn encoder-decoder for statistical machine translation (2014).arXiv:1406.1078. URLhttps://arxiv.org/abs/1406.1078

    Pith/arXiv arXiv 2014

  21. [21]

    Kaushik, N

    M. Nguyen, T. He, L. An, D. C. Alexander, J. Feng, B. T. Yeo, Predicting alzheimer’s disease progression using deep recurrent neural networks, NeuroImage 222 (11 2020).doi:10.1016/j. neuroimage.2020.117203

    work page doi:10.1016/j 2020

  22. [22]

    W. Jung, E. Jun, H. I. Suk, A. D. N. Initiative, Deep recurrent model for individualized predic- tion of alzheimer’s disease progression, NeuroImage 237 (8 2021).doi:10.1016/j.neuroimage. 2021.118143

    work page doi:10.1016/j.neuroimage 2021

  23. [23]

    Liang, K

    W. Liang, K. Zhang, P. Cao, X. Liu, J. Yang, O. Zaiane, Rethinking modeling alzheimer’s disease progression from a multi-task learning perspective with deep recurrent neural network, Computers in Biology and Medicine 138 (11 2021).doi:10.1016/j.compbiomed.2021.104935

    work page doi:10.1016/j.compbiomed.2021.104935 2021

  24. [24]

    Mukherji, M

    D. Mukherji, M. Mukherji, N. Mukherji, Early detection of alzheimer’s disease using neuropsy- chological tests: a predict–diagnose approach using neural networks, Brain Informatics 9 (12 2022).doi:10.1186/s40708-022-00169-1

    work page doi:10.1186/s40708-022-00169-1 2022

  25. [25]

    Morar, H

    U. Morar, H. Martin, P. M. Robin, W. Izquierdo, E. Zarafshan, P. Forouzannezhad, E. Unger, M. Cabrerizo, R. E. C. Cid, M. Rosselli, A. Barreto, N. Rishe, D. E. Vaillancourt, S. T. DeKosky, D. Loewenstein, R. Duara, M. Adjouadi, Prediction of cognitive test scores from variable length multimodal data in alzheimer’s disease, Cognitive Computation 15 (2023) ...

    work page doi:10.1007/s12559-023-10169-w 2023

  26. [26]

    Poonam, R

    K. Poonam, R. Guha, P. P. Chakrabarti, Accurate prediction of alzheimer’s disease progression trajectory via a novel encoder-decoder lstm architecture, in: 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 2023, pp. 1–4. doi:10.1109/EMBC40787.2023.10340517

    work page doi:10.1109/embc40787.2023.10340517 2023

  27. [27]

    Poonam, R

    K. Poonam, R. Guha, P. P. Chakrabarti, Predicting alzheimer’s disease progression using a ver- satile sequence-length-adaptive encoder-decoder lstm architecture, IEEE Journal of Biomedical and Health Informatics 28 (2024) 4184–4193.doi:10.1109/JBHI.2024.3386801. 25

    work page doi:10.1109/jbhi.2024.3386801 2024

  28. [28]

    R. Das, A. Chatterjee, S. Roy, An interpretable bayesian framework for alzheimer’s disease prediction with uncertainty quantification, Neuroscience 589 (2025) 150–160.doi:https:// doi.org/10.1016/j.neuroscience.2025.10.021. URLhttps://www.sciencedirect.com/science/article/pii/S0306452225010139

    work page doi:10.1016/j.neuroscience.2025.10.021 2025

  29. [29]

    Chung, C

    J. Chung, C. Gulcehre, K. Cho, Y. Bengio, Empirical evaluation of gated recurrent neural networks on sequence modeling, arXiv preprint arXiv:1412.3555 (2014)

    Pith/arXiv arXiv 2014

  30. [30]

    R. T. Q. Chen, Y. Rubanova, J. Bettencourt, D. Duvenaud, Neural ordinary differential equa- tions (2019).arXiv:1806.07366. URLhttps://arxiv.org/abs/1806.07366

    Pith/arXiv arXiv 2019

  31. [31]

    Jeong, W

    S. Jeong, W. Jung, J. Sohn, H. I. Suk, Deep geometric learning with monotonicity constraints for alzheimer’s disease progression, IEEE Transactions on Neural Networks and Learning Systems (2024).doi:10.1109/TNNLS.2024.3394598

    work page doi:10.1109/tnnls.2024.3394598 2024

  32. [32]

    S. G. Mueller, M. W. Weiner, L. J. Thal, R. C. Petersen, C. Jack, W. Jagust, J. Q. Trojanowski, A. W. Toga, L. Beckett, The alzheimer’s disease neuroimaging initiative, Neuroimaging Clinics of North America 15 (2005) 869–877.doi:10.1016/j.nic.2005.09.008

    work page doi:10.1016/j.nic.2005.09.008 2005

  33. [34]

    D. P. Kingma, M. Welling, Auto-encoding variational bayes (2022).arXiv:1312.6114. URLhttps://arxiv.org/abs/1312.6114

    Pith/arXiv arXiv 2022

  34. [35]

    Akiba, S

    T. Akiba, S. Sano, T. Yanase, T. Ohta, M. Koyama, Optuna: A next-generation hyperparameter optimization framework, in: Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining, 2019, pp. 2623–2631

    2019

  35. [36]

    R. T. Chen, B. Amos, M. Nickel, Learning neural event functions for ordinary differential equa- tions, arXiv preprint arXiv:2011.03902 (2020)

    arXiv 2011

  36. [37]

    B. Lei, E. Liang, M. Yang, P. Yang, F. Zhou, E. L. Tan, Y. Lei, C. M. Liu, T. Wang, X. Xiao, S. Wang, Predicting clinical scores for alzheimer’s disease based on joint and deep learning, Expert Systems with Applications 187 (1 2022).doi:10.1016/j.eswa.2021.115966

    work page doi:10.1016/j.eswa.2021.115966 2022

  37. [38]

    Franco-Marina, J

    F. Franco-Marina, J. J. García-González, F. Wagner-Echeagaray, J. Gallo, O. Ugalde, S. Sánchez-García, C. Espinel-Bermúdez, T. Juárez-Cedillo, M. Ángel Villa Rodríguez, C. García-Peña, The mini-mental state examination revisited: ceiling and floor effects after score adjustment for educational level in an aging mexican population, International Psychogeri...

    work page doi:10.1017/s1041610209990822 2010

  38. [39]

    S. B. Sando, S. Melquist, A. Cannon, M. L. Hutton, O. Sletvold, I. Saltvedt, L. R. White, S. Lydersen, J. O. Aasly, Apoeε4 lowers age at onset and is a high risk factor for alzheimer’s disease; a case control study from central norway, BMC neurology 8 (1) (2008) 9

    2008

  39. [40]

    A. D. Roses, On the discovery of the genetic association of apolipoprotein e genotypes and common late-onset alzheimer disease, Journal of Alzheimer’s Disease 9 (s3) (2006) 361–366

    2006

  40. [41]

    J. Kim, J. M. Basak, D. M. Holtzman, The role of apolipoprotein e in alzheimer’s disease, Neuron 63 (3) (2009) 287–303. 26

    2009

  41. [42]

    S. H. Cho, M. Jang, H. Ju, M. J. Kang, J. M. Yun, J. W. Yun, Association of late-life body mass index with the risk of alzheimer disease: a 10-year nationwide population-based cohort study, Scientific Reports 12 (1) (2022) 15298

    2022

  42. [43]

    H. Xie, C. Zhang, Y. Wang, S. Huang, W. Cui, W. Yang, L. Koski, X. Xu, Y. Li, M. Zheng, et al., Distinct patterns of cognitive aging modified by education level and gender among adults with limited or no formal education: a normative study of the mini-mental state examination, Journal of alzheimer’s disease 49 (4) (2016) 961–969

    2016

  43. [44]

    Odaibo, Tutorial: Deriving the standard variational autoencoder (vae) loss function, arXiv preprint arXiv:1907.08956 (2019)

    S. Odaibo, Tutorial: Deriving the standard variational autoencoder (vae) loss function, arXiv preprint arXiv:1907.08956 (2019). 27

    Pith/arXiv arXiv 1907