pith. sign in

arxiv: 2605.16969 · v1 · pith:ZFMCS7UBnew · submitted 2026-05-16 · 💻 cs.AI

Brain Vascular Age Prediction Using Cerebral Blood Flow Velocity and Machine Learning Algorithms

Anni Zhao , Alex Bateh , Tyler Baldridge , Sandra Billinger , Xiao Hu This is my paper

Pith reviewed 2026-05-19 20:19 UTC · model grok-4.3

classification 💻 cs.AI
keywords transcranial Dopplercerebral blood flow velocityvascular age predictionmachine learning regressionaccelerated cerebrovascular agingstrokeAlzheimer's diseaseMOCAIP
0
0 comments X

The pith

Features from transcranial Doppler measurements of cerebral blood flow velocity enable machine learning models to predict brain vascular age and identify accelerated aging in diseased subjects.

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

This paper establishes that machine learning regression models, trained on features extracted from transcranial Doppler recordings of the middle cerebral artery in healthy subjects, can estimate cerebrovascular age. When applied to patients with brain diseases such as stroke and Alzheimer's, these models indicate varying degrees of accelerated aging compared to chronological age. A reader might care because this offers a potential non-invasive method to assess and track cerebrovascular health beyond standard chronological measures. The study also notes that the models slightly over-predict age in healthy subjects and that dataset imbalance affects performance.

Core claim

Regression models trained exclusively on healthy subjects using MOCAIP-generated features from TCD and heart rate variability predict subjects' chronological age, and when tested on diseased subjects, reveal accelerated cerebrovascular aging, with the differences between healthy and diseased performances suggesting TCD features' relevance for evaluating such acceleration.

What carries the argument

MOCAIP algorithm for morphological analysis and clustering of signals from TCD recordings, producing features combined with heart rate variability for input to regression models predicting age.

If this is right

  • Healthy subjects' cerebrovascular age is predicted to be 3.69 years above their chronological age on average.
  • Subjects with acute stroke, post-stroke, Alzheimer's disease, and mild cognitive impairment exhibit different levels of age acceleration.
  • TCD features may be relevant for evaluating accelerated cerebrovascular aging.
  • Imbalanced datasets affect the performance of machine-learning-based brain age prediction models.

Where Pith is reading between the lines

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

  • This method could potentially be used to monitor cerebrovascular aging in clinical settings for early intervention.
  • Further studies might explore combining TCD with other imaging modalities for more accurate predictions.
  • Calibrating the model to reduce the 3.69-year over-prediction in healthy subjects could improve its baseline accuracy.

Load-bearing premise

The assumption that regression models trained only on healthy subjects yield an unbiased baseline for normal cerebrovascular aging, even as they over-predict healthy subjects' ages by 3.69 years on average.

What would settle it

Observing no significant difference in predicted minus chronological age between diseased and healthy subjects when using the same trained models would falsify the claim of detecting accelerated aging via TCD features.

Figures

Figures reproduced from arXiv: 2605.16969 by Alex Bateh, Anni Zhao, Sandra Billinger, Tyler Baldridge, Xiao Hu.

Figure 1
Figure 1. Figure 1: Data distribution of healthy subjects and diseased subjects with Alzheimer’s and MCI diseases. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Data distribution of healthy subjects and diseased subjects with acute stroke and post stroke diseases. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Data distribution of healthy subjects and established subjects. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Data distribution of the training and testing healthy subjects for machine learning algorithms. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Averaged dominant pulse comparisons in age group [20,40] for healthy subjects and different disease groups. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Data processing and feature extraction procedures for brain vascular age prediction using CBv. Left: Data [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Group-level standardized feature differences for MOCAIP features extracted from the healthy and the diseased [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Brain vascular prediction results from healthy and AD subjects. [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Brain vascular prediction results from subjects with MCI and acute stroke. [PITH_FULL_IMAGE:figures/full_fig_p010_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Brain vascular prediction results from post stroke and established subjects. [PITH_FULL_IMAGE:figures/full_fig_p010_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: The positive and negative predicted brain vascular prediction gap for all subjects, including the healthy and [PITH_FULL_IMAGE:figures/full_fig_p011_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: The positive and negative predicted brain vascular age gap for AD and Established subjects. [PITH_FULL_IMAGE:figures/full_fig_p011_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: The positive and negative predicted brain vascular age gap for healthy and MCI subjects. [PITH_FULL_IMAGE:figures/full_fig_p012_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: The positive and negative predicted brain vascular age gap for post stroke and stroke subjects. [PITH_FULL_IMAGE:figures/full_fig_p013_14.png] view at source ↗
read the original abstract

Defining vascular age in terms of physiological function has become one focal point of the extensive studies to categorize and track chronological age. Transcranial Doppler (TCD) is a method by which cerebral blood flow velocity is measured along the major arteries feeding the human brain. This study aims to use features extracted from TCD to estimate chronological age and assess accelerated aging in subjects with various brain diseases. We predict subjects with various brain diseases to present with accelerated cerebrovascular aging when tested on various regression models trained by healthy subjects. 168 healthy subjects and 277 diseased subjects with bilateral TCD recordings of the middle cerebral artery were analyzed using the Morphological Analysis and Clustering of Intracranial Pressure (MOCAIP) algorithm. MOCAIP-generated features and heart rate variability features were used as input features for regression models to predict the brain vascular age. 66 subjects with acute stroke, 27 subjects with post stroke, 26 subjects with Alzheimer's disease, 23 subjects with mild cognitive impairment, and 135 established subjects were tested against the machine learning model to assess for accelerated cerebrovascular age. The trained model, on average, predicted healthy subjects' cerebrovascular age to be 3.69 years above their chronological age. Subjects with different disease conditions exhibited varying levels of age acceleration. The differences in healthy and diseased subjects' performances suggest that features generated using TCD may be relevant when evaluating accelerated cerebrovascular aging. Moreover, imbalanced datasets have been observed to affect the performance of machine-learning-based brain age prediction models.

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 / 2 minor

Summary. The manuscript claims that features extracted from transcranial Doppler (TCD) recordings of the middle cerebral artery via the MOCAIP algorithm, combined with heart rate variability features, can be used to train regression models on 168 healthy subjects to predict cerebrovascular (vascular) age. These models are then applied to 277 subjects with brain diseases (66 acute stroke, 27 post-stroke, 26 Alzheimer's, 23 mild cognitive impairment, and 135 'established' subjects) to demonstrate accelerated cerebrovascular aging, with the model over-predicting healthy subjects' age by an average of 3.69 years and showing varying acceleration levels across disease groups. The authors conclude that TCD/MOCAIP features are relevant for assessing accelerated cerebrovascular aging, while noting that imbalanced datasets affect model performance.

Significance. If the reported bias can be corrected and proper validation metrics provided, the work could contribute a non-invasive, TCD-based approach to quantifying cerebrovascular aging and its acceleration in neurological conditions, potentially complementing existing brain-age prediction methods that rely on MRI or other modalities. The use of MOCAIP for feature extraction from cerebral blood flow velocity is a specific strength that could enable broader clinical translation if reproducibility is demonstrated.

major comments (3)
  1. [Abstract] Abstract: The central claim that diseased subjects exhibit accelerated cerebrovascular aging relative to a healthy baseline is undermined by the reported 3.69-year over-prediction of age in the healthy training population itself. This systematic positive bias indicates that the learned mapping from TCD/MOCAIP features to age does not center on chronological age, so excess predictions in disease groups may partly reflect the same offset rather than disease-specific acceleration. Without explicit bias correction, calibration plots, or a demonstration that the offset is negligible after proper validation, the cross-group differences do not securely support the relevance of TCD features for accelerated aging assessment.
  2. [Abstract] Abstract and Methods (implied): No model performance metrics (MAE, RMSE, R²), cross-validation procedure, or statistical tests for group differences are reported, despite concrete numerical claims such as the 3.69-year offset and 'varying levels of age acceleration.' This absence prevents evaluation of whether the regression models are reliable or whether the observed differences exceed what would be expected from noise or imbalance.
  3. [Abstract] Abstract: The manuscript acknowledges that 'imbalanced datasets have been observed to affect the performance' but provides no details on how class imbalance was addressed during training (e.g., weighting, resampling, or stratified cross-validation) or whether results were reported separately for balanced subsets. Given that the healthy training set (n=168) is smaller than the combined diseased test set (n=277), this omission is load-bearing for interpreting the acceleration claims.
minor comments (2)
  1. [Abstract] The term 'established subjects' in the abstract is undefined and should be clarified (e.g., is this a control or specific disease subgroup?).
  2. [Abstract] The abstract states 'We predict subjects with various brain diseases to present with accelerated cerebrovascular aging' but does not specify the exact statistical comparison or null hypothesis used to support this prediction.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which highlight important aspects for improving the clarity and rigor of our analysis on TCD-based cerebrovascular age prediction. We address each major comment point by point below and will incorporate revisions to strengthen the manuscript where feasible.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that diseased subjects exhibit accelerated cerebrovascular aging relative to a healthy baseline is undermined by the reported 3.69-year over-prediction of age in the healthy training population itself. This systematic positive bias indicates that the learned mapping from TCD/MOCAIP features to age does not center on chronological age, so excess predictions in disease groups may partly reflect the same offset rather than disease-specific acceleration. Without explicit bias correction, calibration plots, or a demonstration that the offset is negligible after proper validation, the cross-group differences do not securely support the relevance of TCD features for accelerated aging assessment.

    Authors: We acknowledge the systematic positive bias observed in the healthy training cohort (average over-prediction of 3.69 years), which is explicitly reported in the manuscript. This offset may arise from the regression fitting process or inherent characteristics of the TCD/MOCAIP features. Our core claim focuses on relative differences and varying acceleration levels across disease groups compared to this healthy baseline. To address the concern, we will add calibration plots in the revised manuscript and implement a simple bias correction by subtracting the mean healthy offset from all predictions, thereby centering the healthy excess age at zero for clearer interpretation of disease-specific acceleration. We will also explicitly clarify in the text that reported accelerations represent excess beyond the observed healthy bias. This revision should better isolate disease-related effects while preserving the reported numerical findings. revision: yes

  2. Referee: [Abstract] Abstract and Methods (implied): No model performance metrics (MAE, RMSE, R²), cross-validation procedure, or statistical tests for group differences are reported, despite concrete numerical claims such as the 3.69-year offset and 'varying levels of age acceleration.' This absence prevents evaluation of whether the regression models are reliable or whether the observed differences exceed what would be expected from noise or imbalance.

    Authors: We agree that the absence of standard performance metrics and validation details limits the ability to fully assess model reliability. In the revised manuscript, we will expand the Methods section to describe the cross-validation procedure (e.g., 5-fold or leave-one-out CV on the healthy cohort) and report quantitative metrics including MAE, RMSE, and R² for the regression models. We will also add statistical analyses, such as ANOVA or pairwise t-tests with multiple-comparison corrections, to evaluate the significance of differences in predicted age acceleration between the healthy group and each disease subgroup. These additions will directly support the numerical claims and allow readers to judge whether group differences are statistically meaningful beyond potential noise. revision: yes

  3. Referee: [Abstract] Abstract: The manuscript acknowledges that 'imbalanced datasets have been observed to affect the performance' but provides no details on how class imbalance was addressed during training (e.g., weighting, resampling, or stratified cross-validation) or whether results were reported separately for balanced subsets. Given that the healthy training set (n=168) is smaller than the combined diseased test set (n=277), this omission is load-bearing for interpreting the acceleration claims.

    Authors: We recognize that dataset imbalance, particularly the disparity between the healthy training set (n=168) and the larger diseased cohort (n=277), requires explicit handling and discussion. In the revision, we will detail any techniques applied during model training, such as class weighting in the regression models or stratified sampling where relevant. If these were not previously used, we will re-evaluate and report performance on balanced subsets of the data or using imbalance-robust metrics. We will also expand the discussion to quantify the potential impact of imbalance on the observed acceleration differences, ensuring the claims are interpreted with appropriate caveats. revision: yes

Circularity Check

0 steps flagged

No significant circularity: vascular age predictions use independent diseased-subject data

full rationale

The paper extracts MOCAIP and HRV features from TCD recordings, fits regression models exclusively on the 168 healthy subjects to map features to chronological age, then applies the fitted models to the separate 277 diseased subjects to obtain predicted vascular ages and compute group-wise differences in (predicted minus chronological) age. This chain does not reduce to self-definition or fitted-input renaming because the diseased cohort supplies independent feature vectors whose outputs are not constrained by the healthy training distribution; the reported 3.69-year average over-prediction on healthy subjects is an observed validation statistic rather than a definitional identity, and the cross-group comparison therefore retains external empirical content.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The paper rests on standard domain assumptions about TCD signals reflecting vascular aging and on conventional machine-learning regression practices; no new physical entities are introduced. Free parameters consist of model hyperparameters and feature choices that are fitted to the healthy cohort. Because only the abstract is available, the ledger is necessarily incomplete.

free parameters (2)
  • Regression model hyperparameters
    Parameters of the various regression models trained on healthy-subject TCD and HRV features to predict chronological age.
  • Feature selection from MOCAIP output
    Choice of which morphological and clustering features extracted by MOCAIP are retained as inputs.
axioms (2)
  • domain assumption TCD velocity waveforms from the middle cerebral artery contain information about cerebrovascular aging processes
    Invoked when using these signals to train age-prediction models.
  • domain assumption Regression models trained on healthy subjects can serve as an unbiased reference for normal vascular aging
    Required for interpreting higher predicted ages in diseased subjects as acceleration.

pith-pipeline@v0.9.0 · 5807 in / 1542 out tokens · 57128 ms · 2026-05-19T20:19:43.888791+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

38 extracted references · 38 canonical work pages

  1. [1]

    Deep learning-derived cardiovascular age shares a genetic basis with other cardiac phenotypes.Scientific reports, 12(1):22625, 2022

    Julian Libiseller-Egger, Jody E Phelan, Zachi I Attia, Ernest Diez Benavente, Susana Campino, Paul A Friedman, Francisco Lopez-Jimenez, David A Leon, and Taane G Clark. Deep learning-derived cardiovascular age shares a genetic basis with other cardiac phenotypes.Scientific reports, 12(1):22625, 2022. ISSN 2045-2322

    work page 2022

  2. [2]

    Photoplethysmogram based vascular aging assessment using the deep convolutional neural network.Scientific Reports, 12(1):11377, 2022

    Hangsik Shin, Gyujeong Noh, and Byung-Moon Choi. Photoplethysmogram based vascular aging assessment using the deep convolutional neural network.Scientific Reports, 12(1):11377, 2022. ISSN 2045-2322

    work page 2022

  3. [3]

    Prediction of brain age based on the community structure of functional networks.Biomedical Signal Processing and Control, 79:104151, 2023

    Hongfang Han, Sheng Ge, and Haixian Wang. Prediction of brain age based on the community structure of functional networks.Biomedical Signal Processing and Control, 79:104151, 2023. ISSN 1746-8094

    work page 2023

  4. [4]

    Deep learning to predict future cognitive decline: a multimodal approach using brain mri and clinical data.Frontiers in Neuroimaging, 5:1726037, 2026

    Tamoghna Chattopadhyay, Pavithra Senthilkumar, Rahul H Ankarath, Christopher Patterson, Emma J Gleave, Sophia I Thomopoulos, Heng Huang, Li Shen, Lei You, and Degui Zhi. Deep learning to predict future cognitive decline: a multimodal approach using brain mri and clinical data.Frontiers in Neuroimaging, 5:1726037, 2026

    work page 2026

  5. [5]

    A deep ensemble hip- pocampal cnn model for brain age estimation applied to alzheimer’s diagnosis.Expert Systems with Applications, 195:116622, 2022

    Katia Maria Poloni, Ricardo José Ferrari, and Alzheimer’s Disease Neuroimaging Initiative. A deep ensemble hip- pocampal cnn model for brain age estimation applied to alzheimer’s diagnosis.Expert Systems with Applications, 195:116622, 2022. ISSN 0957-4174

    work page 2022

  6. [6]

    Predicting age from cortical structure across the lifespan

    Christopher R Madan and Elizabeth A Kensinger. Predicting age from cortical structure across the lifespan. European Journal of Neuroscience, 47(5):399–416, 2018. ISSN 0953-816X

    work page 2018

  7. [7]

    Mri-based age prediction using hidden markov models.Journal of neuroscience methods, 199(1):140–145, 2011

    Bing Wang and Tuan D Pham. Mri-based age prediction using hidden markov models.Journal of neuroscience methods, 199(1):140–145, 2011. ISSN 0165-0270

    work page 2011

  8. [8]

    Federation of brain age estimation in structural neuroimaging data

    Sunitha Basodi, Rajikha Raja, Bhaskar Ray, Harshvardhan Gazula, Jingyu Liu, Eric Verner, and Vince D Calhoun. Federation of brain age estimation in structural neuroimaging data. In2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), pages 3854–3857. IEEE, 2021. ISBN 172811179X

    work page 2021

  9. [9]

    Multimodal sparse classifier for adolescent brain age prediction.IEEE journal of biomedical and health informatics, 24(2):336–344, 2019

    Peyman Hosseinzadeh Kassani, Alexej Gossmann, and Yu-Ping Wang. Multimodal sparse classifier for adolescent brain age prediction.IEEE journal of biomedical and health informatics, 24(2):336–344, 2019. ISSN 2168-2194

    work page 2019

  10. [10]

    A review on brain age prediction models.Brain Research, 1823:148668,

    LK Soumya Kumari and R Sundarrajan. A review on brain age prediction models.Brain Research, 1823:148668,

    work page

  11. [11]

    Uk biobank-centric advances in brain age prediction: a comprehensive review.Reviews in the Neurosciences, 37(1):19–42, 2026

    Yanxue Li, Hongjian Gao, Lan Lin, Yutong Wu, and Xinyu Zhu. Uk biobank-centric advances in brain age prediction: a comprehensive review.Reviews in the Neurosciences, 37(1):19–42, 2026. ISSN 0334-1763

    work page 2026

  12. [12]

    Mri signatures of brain age and disease over the lifespan based on a deep brain network and 14 468 individuals worldwide.Brain, 143(7):2312–2324, 2020

    Vishnu M Bashyam, Guray Erus, Jimit Doshi, Mohamad Habes, Ilya M Nasrallah, Monica Truelove-Hill, Dhivya Srinivasan, Liz Mamourian, Raymond Pomponio, and Yong Fan. Mri signatures of brain age and disease over the lifespan based on a deep brain network and 14 468 individuals worldwide.Brain, 143(7):2312–2324, 2020. ISSN 0006-8950

    work page 2020

  13. [13]

    Changes in cerebral arterial pulsatility and hippocampal volume: a transcranial doppler ultrasonography study.Neurobiology of Aging, 108: 110–121, 2021

    Matt L Miller, Paolo Ghisletta, Bradley S Jacobs, Cheryl L Dahle, and Naftali Raz. Changes in cerebral arterial pulsatility and hippocampal volume: a transcranial doppler ultrasonography study.Neurobiology of Aging, 108: 110–121, 2021. ISSN 0197-4580. 14 arXivTemplateA PREPRINT

    work page 2021

  14. [14]

    Artificial intelligence-derived photoplethysmography age as a digital biomarker for cardiovascular health.Communications Medicine, 5(1):481,

    Guangkun Nie, Qinghao Zhao, Gongzheng Tang, Yaxin Li, and Shenda Hong. Artificial intelligence-derived photoplethysmography age as a digital biomarker for cardiovascular health.Communications Medicine, 5(1):481,

    work page

  15. [15]

    A reusable benchmark of brain-age prediction from m/eeg resting-state signals.Neuroimage, 262:119521, 2022

    Denis A Engemann, Apolline Mellot, Richard Höchenberger, Hubert Banville, David Sabbagh, Lukas Gemein, Tonio Ball, and Alexandre Gramfort. A reusable benchmark of brain-age prediction from m/eeg resting-state signals.Neuroimage, 262:119521, 2022. ISSN 1053-8119

    work page 2022

  16. [16]

    Lifespan brain age prediction based on multiple eeg oscillatory features and sparse group lasso.Frontiers in Aging Neuroscience, 17:1559067, 2025

    Shiang Hu, Xue Xiang, Xiaolong Huang, Yan Lu, Xulai Zhang, Dezhong Yao, and Pedro A Valdes-Sosa. Lifespan brain age prediction based on multiple eeg oscillatory features and sparse group lasso.Frontiers in Aging Neuroscience, 17:1559067, 2025. ISSN 1663-4365

    work page 2025

  17. [17]

    Association of pulsatile and mean cerebral blood flow velocity with age and neuropsychological performance.Physiology & behavior, 130:23–27, 2014

    Matthew P Pase, Natalie A Grima, Con Stough, Andrew Scholey, and Andrew Pipingas. Association of pulsatile and mean cerebral blood flow velocity with age and neuropsychological performance.Physiology & behavior, 130:23–27, 2014. ISSN 0031-9384

    work page 2014

  18. [18]

    Morphological clustering and analysis of continuous intracranial pressure.IEEE Transactions on Biomedical Engineering, 56(3):696–705, 2008

    Xiao Hu, Peng Xu, Fabien Scalzo, Paul Vespa, and Marvin Bergsneider. Morphological clustering and analysis of continuous intracranial pressure.IEEE Transactions on Biomedical Engineering, 56(3):696–705, 2008

    work page 2008

  19. [19]

    Cerebrovascular longitudinal atlas: Changes in cerebral arteries in unruptured intracranial aneurysm patients followed with mra.NeuroImage: Clinical, 46:103766, 2025

    Aichi Chien, Fernando Vinuela, Viktor Szeder, Geoffrey Colby, Reza Jahan, Anthony Wang, Satoshi Tateshima, Gary Duckwiler, and Noriko Salamon. Cerebrovascular longitudinal atlas: Changes in cerebral arteries in unruptured intracranial aneurysm patients followed with mra.NeuroImage: Clinical, 46:103766, 2025. ISSN 2213-1582

    work page 2025

  20. [20]

    Toward automated classification of pathological transcranial doppler waveform morphology via spectral clustering.PloS one, 15(2):e0228642, 2020

    Samuel G Thorpe, Corey M Thibeault, Nicolas Canac, Kian Jalaleddini, Amber Dorn, Seth J Wilk, Thomas Devlin, Fabien Scalzo, and Robert B Hamilton. Toward automated classification of pathological transcranial doppler waveform morphology via spectral clustering.PloS one, 15(2):e0228642, 2020. ISSN 1932-6203

    work page 2020

  21. [21]

    Preliminary study of dynamic cerebral autoregulation in acute ischemic stroke: association with clinical factors.Frontiers in Neurology, 9:1006, 2018

    Hongyin Ma, Zhen-Ni Guo, Hang Jin, Xiuli Yan, Jia Liu, Shan Lv, Peng Zhang, Xin Sun, and Yi Yang. Preliminary study of dynamic cerebral autoregulation in acute ischemic stroke: association with clinical factors.Frontiers in Neurology, 9:1006, 2018. ISSN 1664-2295

    work page 2018

  22. [22]

    Pictorial essay: transcranial doppler findings of the intracranial and extracranial diseases.Journal of Neurosonology and Neuroimaging, 11(1):2–21, 2019

    Jongyeol Kim. Pictorial essay: transcranial doppler findings of the intracranial and extracranial diseases.Journal of Neurosonology and Neuroimaging, 11(1):2–21, 2019. ISSN 2635-425X

    work page 2019

  23. [23]

    Continuous detection of cerebral vasodilatation and vasoconstriction using intracranial pulse morphological template matching.Plos one, 7(11):e50795, 2012

    Shadnaz Asgari, Nestor Gonzalez, Andrew W Subudhi, Robert Hamilton, Paul Vespa, Marvin Bergsneider, Robert C Roach, and Xiao Hu. Continuous detection of cerebral vasodilatation and vasoconstriction using intracranial pulse morphological template matching.Plos one, 7(11):e50795, 2012. ISSN 1932-6203

    work page 2012

  24. [24]

    Machine learning for health services researchers.Value in Health, 22(7):808–815, 2019

    Patrick Doupe, James Faghmous, and Sanjay Basu. Machine learning for health services researchers.Value in Health, 22(7):808–815, 2019. ISSN 1098-3015

    work page 2019

  25. [25]

    Predicting 30-days mortality for mimic-iii patients with sepsis-3: a machine learning approach using xgboost.Journal of translational medicine, 18(1):462, 2020

    Nianzong Hou, Mingzhe Li, Lu He, Bing Xie, Lin Wang, Rumin Zhang, Yong Yu, Xiaodong Sun, Zhengsheng Pan, and Kai Wang. Predicting 30-days mortality for mimic-iii patients with sepsis-3: a machine learning approach using xgboost.Journal of translational medicine, 18(1):462, 2020. ISSN 1479-5876

    work page 2020

  26. [26]

    Early prediction of unplanned 30-day hospital readmission: model development and retrospective data analysis.JMIR medical informatics, 9(3):e16306, 2021

    Peng Zhao, Illhoi Yoo, and Syed H Naqvi. Early prediction of unplanned 30-day hospital readmission: model development and retrospective data analysis.JMIR medical informatics, 9(3):e16306, 2021

    work page 2021

  27. [27]

    Machine learning-enabled 30-day readmission model for stroke patients.Frontiers in neurology, 12:638267, 2021

    Negar Darabi, Niyousha Hosseinichimeh, Anthony Noto, Ramin Zand, and Vida Abedi. Machine learning-enabled 30-day readmission model for stroke patients.Frontiers in neurology, 12:638267, 2021. ISSN 1664-2295

    work page 2021

  28. [28]

    Multimodal machine learning for 10-year dementia risk prediction: the framingham heart study

    Huitong Ding, Amiya Mandapati, Alexander P Hamel, Cody Karjadi, Ting FA Ang, Weiming Xia, Rhoda Au, and Honghuang Lin. Multimodal machine learning for 10-year dementia risk prediction: the framingham heart study. Journal of Alzheimer’s Disease, 96(1):277–286, 2023. ISSN 1387-2877

    work page 2023

  29. [29]

    E-catboost: An efficient machine learning framework for predicting icu mortality using the eicu collaborative research database.Plos one, 17(5):e0262895, 2022

    Nima Safaei, Babak Safaei, Seyedhouman Seyedekrami, Mojtaba Talafidaryani, Arezoo Masoud, Shaodong Wang, Qing Li, and Mahdi Moqri. E-catboost: An efficient machine learning framework for predicting icu mortality using the eicu collaborative research database.Plos one, 17(5):e0262895, 2022. ISSN 1932-6203

    work page 2022

  30. [30]

    Applying machine learning to carotid sonographic features for recurrent stroke in patients with acute stroke.Frontiers in Cardiovascular Medicine, 9:804410, 2022

    Shih-Yi Lin, Kin-Man Law, Yi-Chun Yeh, Kuo-Chen Wu, Jhih-Han Lai, Chih-Hsueh Lin, Wu-Huei Hsu, Cheng- Chieh Lin, and Chia-Hung Kao. Applying machine learning to carotid sonographic features for recurrent stroke in patients with acute stroke.Frontiers in Cardiovascular Medicine, 9:804410, 2022. ISSN 2297-055X

    work page 2022

  31. [31]

    Validation of random forest machine learning models to predict dementia-related neuropsychiatric symptoms in real-world data.Journal of Alzheimer’s Disease, 77(2):855–864,

    Javier Mar, Ania Gorostiza, Oliver Ibarrondo, Carlos Cernuda, Arantzazu Arrospide, Álvaro Iruin, Igor Larrañaga, Mikel Tainta, Enaitz Ezpeleta, and Ane Alberdi. Validation of random forest machine learning models to predict dementia-related neuropsychiatric symptoms in real-world data.Journal of Alzheimer’s Disease, 77(2):855–864,

    work page

  32. [32]

    Prediction of incident delirium using a random forest classifier.Journal of medical systems, 42(12):261, 2018

    John P Corradi, Stephen Thompson, Jeffrey F Mather, Christine M Waszynski, and Robert S Dicks. Prediction of incident delirium using a random forest classifier.Journal of medical systems, 42(12):261, 2018. ISSN 0148-5598. 15 arXivTemplateA PREPRINT

    work page 2018

  33. [33]

    Predicting intensive care unit readmission with machine learning using electronic health record data.Annals of the American Thoracic Society, 15(7):846–853, 2018

    Juan C Rojas, Kyle A Carey, Dana P Edelson, Laura R Venable, Michael D Howell, and Matthew M Churpek. Predicting intensive care unit readmission with machine learning using electronic health record data.Annals of the American Thoracic Society, 15(7):846–853, 2018. ISSN 2329-6933

    work page 2018

  34. [34]

    Accurate predictions on small data with a tabular foundation model.Nature, 637 (8045):319–326, 2025

    Noah Hollmann, Samuel Müller, Lennart Purucker, Arjun Krishnakumar, Max Körfer, Shi Bin Hoo, Robin Tibor Schirrmeister, and Frank Hutter. Accurate predictions on small data with a tabular foundation model.Nature, 637 (8045):319–326, 2025. ISSN 0028-0836

    work page 2025

  35. [35]

    Yunhua Li, Jianfeng Yang, Pan Xiao, Haibo Liu, Yingjun Zhou, Xiuqi Yang, Gangwen Chen, and Zhichao Zuo. Mri delta-radiomics and morphological feature-driven tabpfn model for preoperative prediction of lymphovascular invasion in invasive breast cancer.Technology in Cancer Research & Treatment, 24:15330338251362050, 2025. ISSN 1533-0346

    work page 2025

  36. [36]

    Time: Tabpfn-integrated multimodal engine for robust tabular-image learning.arXiv preprint arXiv:2506.00813,

    Jiaqi Luo, Yuan Yuan, and Shixin Xu. Time: Tabpfn-integrated multimodal engine for robust tabular-image learning.arXiv preprint arXiv:2506.00813, 2025

    work page arXiv 2025

  37. [37]

    Predicting dementia in parkinson’s disease on a small tabular dataset using hybrid lightgbm–tabpfn and shap.Digital Health, 10:20552076241272585, 2024

    Vinh Quang Tran and Haewon Byeon. Predicting dementia in parkinson’s disease on a small tabular dataset using hybrid lightgbm–tabpfn and shap.Digital Health, 10:20552076241272585, 2024. ISSN 2055-2076

    work page 2024

  38. [38]

    Scikit-learn: Machine learning in python

    Fabian Pedregosa, Gaël Varoquaux, Alexandre Gramfort, Vincent Michel, Bertrand Thirion, Olivier Grisel, Mathieu Blondel, Peter Prettenhofer, Ron Weiss, and Vincent Dubourg. Scikit-learn: Machine learning in python. the Journal of machine Learning research, 12:2825–2830, 2011. ISSN 1532-4435. 16

    work page 2011