Pith. sign in

REVIEW 3 major objections 5 minor 25 references

Beyond Random Partitioning: Unsupervised Spatio-Temporal Stratification for Cohort Balancing in Longitudinal Medical Imaging

T0 review · 3 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read Random splitting of clinical cohorts secretly creates artificial distribution shifts; this paper replaces it with unsupervised, phenotype-aware stratified partitioning that shrinks cross-subset intensity bias from 34.1% to under 2.1%.

desk verdict A clear, honest methods paper that shows stratified splitting reduces imbalance on the stratification features; the main open question is whether that matters downstream, which the authors themselves flag. read the letter →

arxiv 2608.00073 v1 pith:LD23LPKC submitted 2026-07-29 cs.CV cs.LG

Qinghui Liu , Jon André Ottesen , Atle Bj{o}rnerud , Kyrre Eeg Emblem This is my paper · ORCID
classification cs.CVcs.LG
keywords cohortpartitioninglongitudinalimagingstratifiedsamplingcovariateshiftK-meansclusteringMRIintensitytemporaltrajectoriesdomain
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that the conventional shuffle-and-train habit in medical imaging is unsafe for small longitudinal cohorts, because random splits routinely push rare intensity profiles and irregular follow-up schedules into one subset, creating an artificial domain shift at the evaluation boundary. It proposes a standard operating procedure: audit the cohort's spatial grid, intensity distributions, and temporal trajectories; map each patient to a six-dimensional fingerprint; cluster patients into phenotypic strata with K-means; and then sample proportionally within each cluster to build train, validation, and test sets. On 149 patients with longitudinal contrast-enhanced brain MRI, this reduces the maximum cross-subset intensity imbalance from 34.1% under random splitting to under 2.1%, and keeps follow-up intervals aligned around the population mean. The authors position the protocol as a preemptive, auditable form of domain-shift mitigation that can be combined with later intensity-normalization and adaptation methods.

What carries the argument

The engine is a six-dimensional subject fingerprint: four intensity statistics (minimum, maximum, mean, and standard deviation of voxel intensities in the baseline scan) plus two temporal features (maximum follow-up interval in hours and number of follow-up encounters). These are z-score standardized, clustered by K-means with the number of clusters chosen by an elbow/curvature criterion (K=4 for this cohort), and then each cluster is proportionally sampled so that every subset reproduces the cohort's cluster mix. This intra-cluster proportionate stratified sampling is what guarantees that rare hyperintense phenotypes and long, densely sampled follow-up trajectories appear in every subset in

What would settle it

Train a longitudinal model on a cohort split both ways and compare performance on a truly independent external test set: if the stratified split does not reduce the gap between internal validation and external test performance, or if a held-out covariate like age or lesion burden remains unbalanced across the stratified subsets, the core claim is falsified.

Watch

Extended reading notes

Core claim

The central claim is that covariate shift in longitudinal clinical datasets can be engineered away at the partitioning stage, without labels, by aligning summary statistics of intensity and temporal sampling across subsets. The paper shows that naive random allocation on a 149-patient cohort produces validation and test sets whose mean maximum intensity exceeds the training mean by more than 34%, whereas the proposed six-dimensional, K-means-based stratified protocol confines the same deviation to under 2.1% while equalizing follow-up span and sequence length across splits. The balancing is achieved deterministically and survives Monte Carlo stress tests over ten seeds and three split ratios

Load-bearing premise

The protocol assumes that aligning six summary statistics—four intensity moments and two temporal measures—across subsets is a sufficient proxy for eliminating covariate shift; if the real driver of a model's failure is some covariate not captured by those six numbers, the balanced partitions may not protect the evaluation.

Editorial extensions

If this is right

  • Splitting a longitudinal cohort without this kind of stratification systematically under-represents high-intensity outlier patients in training and over-exposes them at evaluation; the paper quantifies this as a >34% intensity shift.
  • The protocol reduces cross-subset intensity bias to <2.1% and temporal follow-up alignment to within roughly 0.5 hours across holdout and 5-fold cross-validation regimes.
  • Because the features are unsupervised baseline statistics and temporal intervals, the procedure avoids label leakage while making the partition auditable.
  • The framework is split-ratio agnostic and supports stratified k-fold cross-validation, not just a single holdout.
  • The authors recommend running spatial grid audits and intensity fingerprinting before any model training, and applying augmentation only after stratification.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the six summary statistics are not sufficient to capture the covariates that actually drive model performance (for example, tumor volume, age, or scanner site), balancing them may not remove the evaluation risk; the paper itself flags downstream benefit as open.
  • A natural test would be to run a downstream longitudinal model under both random and stratified splits on a task with a known covariate, and check whether stratified splits improve true generalization or merely decorrelate the reported metrics.
  • The framework could be extended to multi-modal or multi-center cohorts by adding modality-level intensity fingerprints and site identities to the clustering space, which would also let it serve as a harmonization-audit tool.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper argues that naive random partitioning of small longitudinal medical imaging cohorts introduces covariate shift and temporal sampling imbalance across train/validation/test subsets. It introduces a Tripartite Dataset Analytics Framework that audits spatial grid consistency, extracts a six-dimensional per-subject fingerprint (Imin, Imax, Imean, Istd, Δtmax, M), and maps longitudinal intervals. On top of this, it formalizes an 'unsupervised spatio-temporal cohort-balancing SOP' that applies elbow-optimized K-means clustering in this feature space (Eq. 4–6) followed by intra-cluster proportionate stratified sampling. On a 149-patient longitudinal contrast-enhanced T1-weighted brain MRI cohort, the paper reports that the SOP reduces the maximum cross-subset intensity bias from 34.1% under random shuffling to under 2.1%, with Monte Carlo stress tests across ten seeds and three split ratios showing tighter alignment than random partitioning. The paper explicitly states in §4.4 that downstream model benefit is not established.

Significance. If the balance claim were validated on features independent of the clustering objective and if downstream evaluation benefit were demonstrated, the SOP would be a useful, reproducible, and auditable pre-training cohort engineering step. The proposed pipeline is coherent and addresses a real gap: small longitudinal cohorts are often split without regard to distributional alignment. The paper also ships a concrete algorithmic recipe (Eq. 4–6) and an explicit stress-testing protocol. However, the current empirical support is substantially weakened by the fact that the reported balance metrics are computed on the same feature vector used to construct clusters and stratified splits, making the headline reduction partly a mathematical consequence of the optimization. The paper's own §4.4 concedes that the link to downstream model reliability is an open question. This limits the significance of the present results to a methodology proposal with an incomplete validation.

major comments (3)
  1. [§3.1, §3.2, Table 1, Eq. (4)] The headline claim that the SOP reduces maximum cross-subset intensity bias from 34.1% to under 2.1% is measured on the same six features (Imin, Imax, Imean, Istd, Δtmax, M) that are used to construct the K-means clusters (Eq. 4–6) and to perform proportionate stratified sampling. The clustering objective explicitly minimizes within-cluster variance over these features, and the sampling step enforces cluster balance across subsets. Therefore, the reduction in cross-subset imbalance for Imax and Δt is a direct property of the optimization target, not an independent demonstration of covariate balance. To support the claim that the SOP mitigates covariate shift, the evaluation must include held-out covariates not used in the clustering vector (e.g., age, sex, lesion volume, scanner, clinical variables) or a downstream model whose performance is compared across SOP-based and random splits.
  2. [Abstract, §3.2, §3.3] The 34.1% versus 2.1% figures are presented in Table 1 as point estimates from a single split, with no confidence intervals or standard errors. The stress tests in Figures 6 and 7 are descriptive (violin plots of deviations) and do not report formal uncertainty quantification or statistical comparisons. The claim that the SOP is 'seed-invariant and scale-invariant' would be much more convincing if the authors reported, across the ten seeds and three split configurations, the distribution of the maximum cross-subset bias (e.g., mean and 95% CI) and a paired statistical test against random shuffling. As it stands, the reader cannot assess whether the 2.1% figure is robust or a favorable draw.
  3. [§4.4] The paper explicitly states that 'whether stratified partitioning of this kind translates into a measurable downstream evaluation benefit is itself dataset-dependent' and that balancing on a fixed set of summary features can under-represent rare sub-populations. These are not merely future-work remarks; they directly qualify the abstract's assertion that random partitioning 'exposes downstream models to out-of-distribution evaluation.' The central practical motivation of the SOP—protecting the evaluation boundary—is therefore supported only by a proxy whose connection to model performance is untested. The authors should either temper the abstract's causal language or add an experiment (e.g., a segmentation or classification task with a standard architecture) that compares evaluation stability under SOP-based versus random splits.
minor comments (5)
  1. [§2.2] The text says 'minimum intensities converged tightly at a fixed value of −0.338 (σ≈0)' but the earlier figure legend uses a window [−0.337, 1]. Please reconcile these values and clarify whether the quoted number is exact or rounded.
  2. [§3.1, Eq. (4)] The paper calls the vector 'six-dimensional', but Imin has near-zero variance in this cohort (§2.2). While retaining it for generalizability is reasonable, the effective dimensionality is five; a brief note would avoid overstating the feature diversity.
  3. [§3.3, Eq. (7)] Equation (7) writes Δ = ||μ_subset − μ_population||, but the text describes it as an 'absolute deviation' along a single feature. If the norm is over the feature vector, this is inconsistent with the univariate descriptions in Figure 6 and 7. Please define the norm explicitly and use a consistent notation.
  4. [§3.3] The seed set is written as 'S ∈ {12,42,...,9999}'; the ellipsis is ambiguous. Please list the exact ten seed values or describe the sampling rule.
  5. [General] Figure 2 uses violin plots with boxplots and jittered points, but the caption does not define the kernel bandwidth or whether the violin width is scaled by sample size. Adding this information would improve reproducibility.

Circularity Check

1 steps flagged · score 6.0 of 10

The headline bias-reduction result is measured on the same six features that the SOP explicitly optimizes and stratifies, so the 34.1%→2.1% reduction is largely by construction; §4.4 concedes downstream benefit remains untested.

  1. fitted input called prediction [Abstract; §3.2 Table 1; §3.3 Eq. (7)]
    "the protocol reduces the maximum cross-subset intensity bias from 34.1% under conventional random shuffling to under 2.1% ... In contrast, our spatio-temporal stratified SOP substantially mitigates these sampling anomalies ... confines cross-subset variation in the multi-parametric intensity metrics to under 2.1% (Train Imax = 1.544 vs. Test Imax = 1.531). ... partition error is quantified as the absolute deviation of a subset’s empirical feature mean from the full cohort mean, Δ=∥μ_subset−μ_population∥, (7) measured independently along the peak spatial intensity axis (Imax) and the temporal f"

    The SOP's construction uses exactly these features: Eq. (4) puts Imax and Δtmax in xi, Eq. (5) minimizes within-cluster variance over z(xi), and Eq. (6) partitions the cohort to minimize that objective; the allocation protocol then does intra-cluster proportionate stratified sampling. The validation metric Δ in Eq. (7) is the deviation of subset means of Imax and Δt from the population mean—i.e., the same quantities the clustering and stratified allocation were designed to equalize across subsets. The 34.1%→2.1% improvement is therefore a direct consequence of the optimization target, not an independent demonstration that covariate shift is controlled. §4.4 acknowledges that downstream benefits are 'an open empirical question,' so the central claim rests on this by-construction metric.

full rationale

The paper's framework is largely self-contained and does not rely on load-bearing self-citations: [9] is cited only for the clinical observation that neuro-oncology follow-up patterns are irregular, which is not central to the algorithm. However, the central empirical validation is circular in the specific sense of Pattern 2. The six-dimensional feature vector xi in Eq. (4) consists of Imin, Imax, Imean, Istd, Δtmax, and M. K-means (Eqs. 5–6) minimizes within-cluster variance over exactly this vector, and the deterministic protocol uses intra-cluster proportionate stratified sampling so each subset receives the same share of every cluster. Consequently, subset means of these six features are forced close to the population means (up to sampling error within clusters); the reported 'maximum cross-subset intensity bias' reduction from 34.1% to under 2.1% and the Δ<0.04 / Δ<0.5h envelopes are measurements of the same features being optimized. The comparison to random shuffling is real and shows random splits are worse, but it does not test whether the six chosen features are the right covariates or whether balancing them protects a downstream model. The paper itself flags this in §4.4: 'whether stratified partitioning of this kind translates into a measurable downstream evaluation benefit is itself dataset-dependent... We view this as an open empirical question.' Thus the central claim is partly by construction, meriting a 6, not a 0: the algorithm is coherent and the random-baseline comparison is informative, but the headline 'covariate shift' reduction is not independently evidenced beyond the fitted feature set.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central numerical claims rest on a small set of choices: the six-feature summary space, the elbow-selected K*, and the assumption that balancing these features matters for model evaluation. No new physical or mathematical entities are introduced. The cohort is single-center and single-modality, so generality is untested.

free parameters (2)
  • K_max (upper bound of elbow scan) = 8
    User-chosen upper bound for the K-means cluster count scan; K* = 4 is then selected by the elbow criterion. The result can vary with this bound, though the paper reports stability across seeds.
  • K* (number of strata) = 4
    Determined via the elbow method with Kmax=8; not externally anchored, and the paper notes the optimal K varies with cohort.
assumptions (3)
  • domain assumption The six summary statistics (Imin, Imax, Imean, Istd, Δtmax, M) are sufficient to characterize cross-subject covariate heterogeneity.
    Invoked in Eq. 4; the paper's own limitation section admits rare sub-populations may be missed and demographic or clinical features may be needed.
  • domain assumption The elbow criterion identifies a meaningful number of phenotypic strata.
    K* selection in §3.1 relies on an inflection heuristic; alternative methods (gap statistic, silhouette) are acknowledged as future work.
  • domain assumption Intensity summary features after standardized preprocessing are stable and comparable across the cohort.
    §2.2 assumes z-scored or clamped intensities have consistent semantics; this is only empirically validated on one cohort.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Beyond Random Partitioning: Unsupervised Spatio-Temporal Stratification for Cohort Balancing in Longitudinal Medical Imaging." pith.science (2026). https://pith.science/paper/LD23LPKC

@misc{pith2026260800073,
  author       = {Pith},
  title        = {Pith review of: Beyond Random Partitioning: Unsupervised Spatio-Temporal Stratification for Cohort Balancing in Longitudinal Medical Imaging},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LD23LPKC}},
  note         = {Machine review of arXiv:2608.00073}
}
abstract

Rigorous dataset partitioning is a foundational, yet frequently overlooked, prerequisite for reliable deep learning in longitudinal medical imaging. Naively shuffling small clinical cohorts routinely introduces covariate shifts and temporal sampling imbalances across training, validation, and test subsets, exposing downstream models to out-of-distribution evaluation. We address this vulnerability with an auditable Tripartite Dataset Analytics Framework that systematically characterizes spatial grid integrity, multi-parametric intensity fingerprints, and longitudinal temporal trajectories, quantifying the heavy-tailed feature dispersion and irregular, episodic sampling intervals typical of real-world clinical cohorts. Building on this characterization, we formalize an unsupervised spatio-temporal cohort-balancing standard operating procedure (SOP) that combines elbow-optimized K-means clustering over a standardized, six-dimensional joint intensity-temporal feature space with intra-cluster proportionate stratified sampling. On a longitudinal, contrast-enhanced $T1$-weighted brain MRI cohort (N=149), the protocol reduces the maximum cross-subset intensity bias from 34.1% under conventional random shuffling to under 2.1%, while aligning longitudinal follow-up intervals closely around the population mean. Monte Carlo stress testing across ten random seeds and three split configurations confirms that this alignment remains tightly bounded, in clear contrast to the substantial variability of random partitioning. The resulting protocol offers a reproducible, generalizable procedure for cohort engineering in variable-length longitudinal clinical imaging workflows.

Figures

Figures reproduced from arXiv: 2608.00073 by the authors.

Figure 1
Figure 1. Clinical motivation: representative longitudinal tracking time￾lines illustrating spatial contrast and temporal sampling heterogene￾ity. Each row shows one patient tracked from the pre-contrast baseline scan (t = 0 h) through subsequent follow-up encounters, rendered under a fixed con￾trast window ([−0.337, 1]) so that intensity differences reflect genuine pathology rather than display normalization. Rows 2 and 3 ex… view at source ↗
Figure 2
Figure 2. Multi-parametric intensity fingerprint analysis of the baseline cohort. The multi-panel plot combines continuous kernel density estimations (violin contours), quartile boundaries (boxplots with median demarcations), and jittered individual subject points, highlighting the sharp distributional contrast between the convergent Imin values and the heavy-tailed Imax parameter. 2.2 Pillar 2: Multi-Parametric Statistical I… view at source ↗
Figure 3
Figure 3. Inter-metric bivariate correlation matrix of baseline imaging fea￾tures. Diagonal elements show marginal kernel density estimates. Off-diagonal panels present cross-parametric scatter plots with ordinary least-squares regres￾sion trajectories and 95% confidence intervals, revealing the strong coupling between Imax and image standard deviation (Istd) [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Longitudinal tracking and density mapping of patient follow-up trajectories. (a) Patient-wise scan trajectories sorted by total follow-up dura￾tion, colored continuously from baseline (t = 0 h) to the long-term observation limit (t = 48 h). (b) Empirical density distri…
Figure 5
Figure 5. Figure 5: Unsupervised spatio-temporal cohort stratification and pheno￾typic profiling (K∗ = 4). (a) Two-dimensional projection of the 6D joint fea￾ture space onto its leading principal components, overlaid with per-cluster kernel density estimation (KDE) contours. The resulting…
Figure 6
Figure 6. Figure 6: Robustness stress test A: Absolute alignment deviations across diverse holdout split ratios. Split-violin plots cross-examine the deviation distributions of the proposed spatio-temporal stratified protocol (deep cyan) against conventional naive random shuffling (orange…
Figure 7
Figure 7. Figure 7: Robustness stress test B: Absolute alignment deviations across iterative cross-validation folds. Split-violin plots compare the proposed stratified protocol (deep cyan) against conventional naive random 5-fold shuf￾fling (orange-red) across Folds 1 to 5. (a) Intensity …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 4 canonical work pages

  1. [1]

    Medical Image Analysis54, 280–296 (2019).https: //doi.org/10.1016/j.media.2019.03.009,https://doi.org/10.1016/ j.media.2019.03.009

    Cheplygina, V., de Bruijne, M., Pluim, J.P.W.: Not-so-supervised: A sur- vey of semi-supervised, multi-instance, and transfer learning in medi- cal image analysis. Medical Image Analysis54, 280–296 (2019).https: //doi.org/10.1016/j.media.2019.03.009,https://doi.org/10.1016/ j.media.2019.03.009

    work page doi:10.1016/j.media.2019.03.009 2019

  2. [2]

    arXiv preprint arXiv:2508.21580 (2025),https://arxiv.org/abs/2508.21580

    Disch, N.A., Kirchhoff, Y., Peretzke, R., Rokuss, M., Roy, S., Ulrich, C., Zimmerer, D., Maier-Hein, K.: Temporal flow matching for learning spatio- temporal trajectories in 4D longitudinal medical imaging. arXiv preprint arXiv:2508.21580 (2025),https://arxiv.org/abs/2508.21580

    arXiv 2025

  3. [3]

    IEEE Transactions on Medical Imaging38(9), 2059–2069 (2019).https://doi.org/10.1109/TMI.2019

    Gao,Y.,Liu,Y.,Wang,Y.,Shi,Z.,Yu,J.:Auniversalintensitystandardiza- tion method based on a many-to-one weak-paired cycle generative adversar- ial network for magnetic resonance images. IEEE Transactions on Medical Imaging38(9), 2059–2069 (2019).https://doi.org/10.1109/TMI.2019. 2894692

    work page doi:10.1109/tmi.2019

  4. [4]

    Medical Imaging with Deep Learning (PMLR)227, 1923–1940 (2024),https://proceedings.mlr.press/v227/ ghazvanchahi24a.html

    Ghazvanchahi, A., Maralani, P.J., Moody, A.R., Khademi, A.: Effect of intensity standardization on deep learning for WML segmentation in multi-centre FLAIR MRI. Medical Imaging with Deep Learning (PMLR)227, 1923–1940 (2024),https://proceedings.mlr.press/v227/ ghazvanchahi24a.html

    1923

  5. [5]

    Radiology278(2), 563–577 (2016).https://doi

    Gillies, R.J., Kinahan, P.E., Hricak, H.: Radiomics: Images are more than pictures, they are data. Radiology278(2), 563–577 (2016).https://doi. org/10.1148/radiol.2015151169

    doi:10.1148/radiol.2015151169 2016

  6. [6]

    IEEE Transactions on Biomedical Engineering69(3), 1173–1185 (2022)

    Guan, H., Liu, M.: Domain adaptation for medical image analysis: A survey. IEEE Transactions on Biomedical Engineering69(3), 1173–1185 (2022). https://doi.org/10.1109/TBME.2021.3117407

    arXiv 2022

  7. [7]

    arXiv preprint arXiv:2502.00851 (2025),https://arxiv.org/abs/2502.00851

    Herdiana, I., Kamal, M.A., Triyani, Estri, M.N., Renny: A more pre- cise elbow method for optimum k-means clustering. arXiv preprint arXiv:2502.00851 (2025),https://arxiv.org/abs/2502.00851

    arXiv 2025

  8. [8]

    Hu- man brain mapping40(17), 4952–4964 (2019)

    Isensee, F., Schell, M., Pflueger, I., Brugnara, G., Bonekamp, D., Neuberger, U., Wick, A., Schlemmer, H.P., Heiland, S., Wick, W., et al.: Automated brain extraction of multisequence mri using artificial neural networks. Hu- man brain mapping40(17), 4952–4964 (2019)

    2019

Show all 25 references
  1. [9]

    IEEE Transactions on Medical Imaging44(6), 2449–2462 (2025)

    Liu, Q., Fuster-Garcia, E., Hovden, I.T., MacIntosh, B.J., Grødem, E.O., Brandal, P., Lopez-Mateu, C., Sederevičius, D., Skogen, K., Schellhorn, T., et al.: Treatment-aware diffusion probabilistic model for longitudinal mri generation and diffuse glioma growth prediction. IEEE...

    2025

  2. [10]

    In: Proceedings of the Fifth Berkeley Symposium on Mathe- matical Statistics and Probability

    MacQueen, J.: Some methods for classification and analysis of multivariate observations. In: Proceedings of the Fifth Berkeley Symposium on Mathe- matical Statistics and Probability. vol. 1, pp. 281–297. Oakland, CA, USA (1967) Spatio-Temporal Stratification for Cohort Balancing 21

    1967

  3. [11]

    Computers in Biology and Medicine183, 109256 (2024).https://doi.org/10.1016/j.compbiomed.2024.109256

    Matta, S., Lamard, M., Zhang, P., Le Guilcher, A., Borderie, L., Cochener, B., Quellec, G.: A systematic review of generalization research in medi- cal image classification. Computers in Biology and Medicine183, 109256 (2024).https://doi.org/10.1016/j.compbiomed.2024.109256

    2024

  4. [12]

    Magnetic Resonance in Medicine42(6), 1072–1081 (1999)

    Nyül, L.G., Udupa, J.K.: On standardizing the MR image inten- sity scale. Magnetic Resonance in Medicine42(6), 1072–1081 (1999). https://doi.org/10.1002/(SICI)1522-2594(199912)42:6<1072:: AID-MRM11>3.0.CO;2-M

    1999 doi

  5. [13]

    MIT Press, Cam- bridge, MA (2009),https://mitpress.mit.edu/9780262170055/ dataset-shift-in-machine-learning/

    Quiñonero-Candela, J., Sugiyama, M., Schwaighofer, A., Lawrence, N.D.: Dataset Shift in Machine Learning. MIT Press, Cam- bridge, MA (2009),https://mitpress.mit.edu/9780262170055/ dataset-shift-in-machine-learning/

    2009

  6. [14]

    Advances in Neural Information Processing Systems (NeurIPS)32, 5320–5330 (2019)

    Rubanova, Y., Chen, R.T.Q., Duvenaud, D.K.: Latent ODEs for irregularly- sampled time series. Advances in Neural Information Processing Systems (NeurIPS)32, 5320–5330 (2019)

    2019

  7. [15]

    In: Proceedings of the 2011 31st International Conference on Distributed Computing Sys- tems Workshops

    Satopaa, V., Albrecht, J., Irwin, D., Raghavan, B.: Finding a "kneedle" in a haystack: Detecting knee points in system behavior. In: Proceedings of the 2011 31st International Conference on Distributed Computing Sys- tems Workshops. pp. 166–171. IEEE (2011).https://doi.org/10....

    2011

  8. [16]

    Statistical Methods in Medical Research23(5), 439–457 (2014).https://doi.org/ 10.1177/0962280214521452

    Shinohara, R.T., Sweeney, E.M., Goldsmith, J., Crainiceanu, C.M.: Volu- metric analysis of MRI outcomes in multi-center clinical trials. Statistical Methods in Medical Research23(5), 439–457 (2014).https://doi.org/ 10.1177/0962280214521452

    2014 doi

  9. [17]

    NeuroImage: Clinical6, 9–19 (2014).https://doi.org/10.1016/j.nicl.2014.08.008

    Shinohara,R.T.,Sweeney,E.M.,Goldsmith,J.,Shiee,N.,Mateen,F.J.,Cal- abresi, P.A., Jarso, S., Pham, D.L., Reich, D.S., Crainiceanu, C.M.: Statisti- cal normalization techniques for magnetic resonance imaging. NeuroImage: Clinical6, 9–19 (2014).https://doi.org/10.1016/j.nicl.2014.08.008

    2014 doi

  10. [18]

    Human Brain Map- ping17(3), 143–155 (2002).https://doi.org/10.1002/hbm.10062

    Smith, S.M.: Fast robust automated brain extraction. Human Brain Map- ping17(3), 143–155 (2002).https://doi.org/10.1002/hbm.10062

    2002 doi

  11. [19]

    NEJM AI2(5), e2400703 (2025).https://doi.org/10.1056/ aioa2400703,https://ai.nejm.org/doi/full/10.1056/AIoa2400703

    Tak, D., Garomsa, B.A., Zapaishchykova, A., Ye, Z., Vajapeyam, S., Ma- hootiha, M., Climent Pardo, J.C., Smith, C., Familiar, A.M., Chaunzwa, T., et al.: Longitudinal risk prediction for pediatric glioma with temporal deep learning. NEJM AI2(5), e2400703 (2025).https://doi.org...

    2025 doi

  12. [20]

    Journal of the Royal Statistical Society: Series B (Statistical Methodology)63(2), 411–423 (2001).https://doi

    Tibshirani, R., Walther, G., Hastie, T.: Estimating the number of clusters in a data set via the gap statistic. Journal of the Royal Statistical Society: Series B (Statistical Methodology)63(2), 411–423 (2001).https://doi. org/10.1111/1467-9868.00293

    2001

  13. [21]

    arXiv preprint arXiv:2511.04789 (2025),https://arxiv.org/abs/2511.04789

    Wang, X., Zhao, Y., Han, K., Luo, X., van Rooij, S., Stevens, J., He, L., Zhan, L., Sun, Y., Wang, W., Yang, C.: Conditional neural ODE for longitudinal Parkinson’s disease progression forecasting. arXiv preprint arXiv:2511.04789 (2025),https://arxiv.org/abs/2511.04789

    2025

  14. [22]

    Proceedings of 22 Q

    Yoon, J.S., Oh, K., Shin, Y., Mazurowski, M.A., Suk, H.I.: Domain generalization for medical image analysis: A review. Proceedings of 22 Q. Liu et al. the IEEE112(10), 1554–1591 (2024).https://doi.org/10.1109/jproc. 2024.3507831

    2024

  15. [23]

    arXiv preprint arXiv:1908.00473 (2019),https://arxiv.org/abs/1908.00473

    Zhang, P., Zhong, Y., Deng, Y., Tang, X., Li, X.: A survey on deep learning of small sample in biomedical image analysis. arXiv preprint arXiv:1908.00473 (2019),https://arxiv.org/abs/1908.00473

    1908 arXiv

  16. [24]

    Medical Image Analysis71, 102051 (2021).https://doi.org/10

    Zhao, Q., Liu, Z., Adeli, E., Pohl, K.M.: Longitudinal self-supervised learn- ing. Medical Image Analysis71, 102051 (2021).https://doi.org/10. 1016/j.media.2021.102051

    2021

  17. [25]

    Radiology295(2), 328–338 (2020)

    Zwanenburg, A., Vallères, M., Abdalah, M.A., Aerts, H.J.L.W., Andrea- rczyk, V., Apte, A., Ashrafinia, S., Bakas, S., Beukinga, R.J., Boellaard, R., et al.: The image biomarker standardization initiative: Standardized features for reproducible radiomics. Radiology295(2), 328–3...

    2020 doi

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.