Pith. sign in

REVIEW 3 major objections 5 minor 27 references

A graph model of DCE-MRI trajectories predicts breast cancer complete response better than vision and self-supervised baselines.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-11 11:32 UTC pith:3WCFM75X

load-bearing objection Solid empirical package on ISPY-2: temporal DAG + GraphSAGE plus three asymmetric trajectory losses that beat re-implemented baselines; absolute numbers stay modest and the asymmetry is intentional but under-tested. the 3 major comments →

arxiv 2607.04912 v1 pith:3WCFM75X submitted 2026-07-06 cs.CV cs.AIcs.LG

Graph Representation Learning of Longitudinal Medical Imaging Trajectories for Treatment Response Prediction

Johannes Kiechle , Richard Osuala , Daniel M. Lang , Stefan M. Fischer , Ivana Jan\'i\v{c}kov\'a , Karim Lekadir , Julia A. Schnabel , Jan C. Peeken This is my paper
classification cs.CV cs.AIcs.LG
keywords GNNsLongitudinalSelf-SupervisedBreast CancerpCR predictionDCE-MRItreatment trajectoriesGraphSAGE
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper argues that pathological complete response after neoadjuvant chemotherapy can be predicted more accurately by treating each patient's longitudinal DCE-MRI scans as a directed acyclic graph of time-ordered features and learning those trajectories with three complementary self-supervised objectives. On 585 patients from the public ISPY-2 cohort, each with four scans, the resulting GNN-pCR model improves balanced accuracy, AUC and Matthews correlation over CNN, LSTM and existing longitudinal self-supervised methods under identical 5-fold splits. Ablations show that population-level alignment of responders, decorrelation of successive time points, and additive temporal consistency each contribute, and that the same gains appear even when only early time points are available. The work therefore claims that explicit relational modeling of treatment trajectories, rather than simple concatenation or recurrence, better captures the imaging dynamics that distinguish responders from non-responders.

Core claim

A time-aware GraphSAGE network operating on a directed acyclic graph of ResNet embeddings, trained with three asymmetric trajectory losses (responder alignment, adjacent-timepoint decorrelation, and additive temporal consistency), yields higher 5-fold pCR prediction performance (bACC 0.6844, AUC 0.7203, MCC 0.3561) than vision and self-supervised baselines on 585 ISPY-2 patients, establishing that structured longitudinal graph modeling improves individualized treatment-response prediction.

What carries the argument

Asymmetric Treatment Response Loss: three complementary self-supervised terms (population-level alignment that attracts responders while repelling non-responders, patient-level decorrelation of successive embeddings, and temporal consistency that requires the final embedding to match the sum of intermediate latent differences) applied only to responders, together with a temporal DAG whose edges run strictly forward in time.

Load-bearing premise

That the three asymmetric losses learn clinically meaningful disease trajectories rather than artifacts of the homogeneous ISPY-2 scan schedule and the fixed ResNet18 features.

What would settle it

Retrain and re-evaluate the identical GNN-pCR pipeline on an external multi-center DCE-MRI cohort whose inter-scan intervals vary substantially; if balanced accuracy and AUC fall to the level of the CNN or LSTM baselines, the claim that the trajectory losses capture generalizable longitudinal dynamics is falsified.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Early-response models that use only the first two time points can still outperform concatenation and LSTM baselines, supporting earlier therapy adaptation.
  • The same graph-plus-asymmetric-loss recipe can be reused for other longitudinal imaging prediction tasks beyond breast-cancer pCR.
  • Public release of the curation library and code enables standardized multi-split benchmarks for future longitudinal pCR methods.
  • Including raw inter-scan time gaps yields only modest gains under ISPY-2's near-uniform schedule, suggesting the graph topology already encodes the dominant temporal structure.

Where Pith is reading between the lines

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

  • Because the alignment loss deliberately omits an attract term for non-responders, the method implicitly treats non-response as a heterogeneous residual class; this design choice may limit calibration when partial responders form a clinically important intermediate group.
  • The same DAG-plus-asymmetric-loss construction could be tested on longitudinal CT or PET trajectories in other solid tumors where complete-response surrogates exist.
  • If the additive temporal-consistency term is the main driver of early-prediction gains, replacing it with a learned ODE or continuous-time GNN layer would be a direct next experiment.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The manuscript proposes GNN-pCR, a 3D spatio-temporal framework for predicting pathological complete response (pCR) to neoadjuvant chemotherapy from longitudinal DCE-MRI. Image volumes at each timepoint are encoded by a shared ResNet18 into embeddings that form a directed acyclic graph (edges from every earlier to every later timepoint); GraphSAGE aggregates the graph into a patient-level embedding. Training is driven by an Asymmetric Treatment Response Loss whose three novel terms (population-level responder alignment, adjacent-timepoint decorrelation, and additive temporal consistency of latent differences) are applied only to responders, while non-responders receive only a repel term. On 585 ISPY-2 patients with four timepoints the method reports 5-fold means of bACC 0.6844, AUC 0.7203 and MCC 0.3561, outperforming CNN, CNN+LSTM, DINOv3 and two SSL baselines; ablations, early-timepoint experiments and inter-scan time-difference tests are also provided. Code and a PyPI curation library are promised.

Significance. If the comparative gains hold under external validation, the work supplies a concrete, reproducible imaging-only benchmark for longitudinal pCR prediction and a usable graph-plus-asymmetric-SSL recipe that improves on both supervised sequence models and recent SSL baselines on the same public cohort. The multi-split protocol, shared encoder, Wilcoxon tests, component ablations and planned open-source release are genuine strengths that raise the bar relative to many prior single-split ISPY-2 studies. The clinical claim that the learned representations capture 'clinically meaningful longitudinal' dynamics remains provisional until the asymmetry of the losses and the lack of multi-center data are addressed, but the engineering contribution is already useful to the community.

major comments (3)
  1. §2.2, Eqs. (1)–(3) and the final L_total formula: L_align, L_decorrelate and L_temporal are all gated by 1{i∈R}; non-responders (381/585) receive only the repel term. The manuscript never reports a symmetric counterpart that applies the same three trajectory objectives to both classes (or to non-responders alone). Without that control it is impossible to know whether the gains in Tables 1–2 arise from genuine trajectory modeling or from a one-sided regularizer that simply densifies the minority class. This is load-bearing for the claim of 'clinically meaningful longitudinal' representations for the full population.
  2. All results (Tables 1–2) are obtained on a single public multi-center trial (ISPY-2) with a near-homogeneous acquisition schedule. The Discussion itself notes that inter-scan time differences add little value for this reason. External validation on an independent cohort (or at least a leave-one-site-out analysis) is required before the method can be presented as a generalizable pCR benchmark; otherwise the reported margins may be dataset-specific.
  3. Table 1 ablation 'w/o GNN' replaces GraphSAGE by a linear head and still yields bACC 0.6610 (close to the full model). Combined with the fact that the three losses already operate on the same embeddings, it remains unclear how much of the lift is attributable to relational message passing versus the asymmetric SSL objectives alone. A clearer isolation (e.g., GNN trained only with supervised loss, or losses applied after mean-pooling) would strengthen the architectural claim.
minor comments (5)
  1. Abstract and p. 8: 'medical imagaging' is a typographical error; correct to 'imaging'.
  2. §3 Implementation Details: DINOv3 is applied to a single 2D axial slice while all other methods use full 3D volumes; the comparison is therefore not modality-matched and should be flagged more explicitly.
  3. Eq. (1): the random permutation π is described only in text; a short formal definition would improve reproducibility.
  4. Table 2 caption and body: the inter-scan time-difference experiment is interesting but the near-null result for most methods is under-discussed relative to the claim that temporal modeling is clinically meaningful.
  5. References [23] and [24] are arXiv preprints; if they remain unpublished at camera-ready, note the version and access date.

Circularity Check

0 steps flagged

No derivation circularity: GNN-pCR losses and topology are design choices evaluated by held-out 5-fold metrics and ablations, not algebraic restatements of the pCR label.

full rationale

The paper proposes a GraphSAGE head on a temporal DAG of ResNet18 embeddings plus three asymmetric self-supervised losses (population alignment among responders only, adjacent-timepoint decorrelation, additive temporal consistency of latent differences). These are applied only to the responder indicator in L_total and are never claimed to be derived from first principles; they are design choices whose utility is measured by independent 5-fold classification metrics (bACC/AUC/MCC) against vision and SSL baselines and by systematic ablations that remove each term or the GNN. The pCR labels enter only as the supervised evaluation target and as the R/N partition that gates the losses; the losses themselves do not algebraically restate or fit the binary label. Mild author-adjacent re-implementation of 3D-LART and ordinary self-citation of related longitudinal work exist but are not load-bearing for the central performance claim. The asymmetry noted by the skeptic (objectives applied only to responders) is a methodological limitation that may affect generalization claims, but it is not circularity under the definition used here: nothing reduces by construction to its own inputs. Score 1 reflects only the minor self-reference, not any forced prediction.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 2 invented entities

The central claim rests on standard GNN/CNN building blocks, the clinical definition of pCR, the ISPY-2 cohort construction, and three paper-specific loss design choices that treat responders and non-responders asymmetrically. No new physical entities are postulated. Free parameters are ordinary deep-learning hyperparameters and the choice of four fixed time points / ResNet18 embedding dimension.

free parameters (3)
  • SGD learning rate and schedule = 1e-2, 100 epochs
    Learning rate 1e-2, cosine annealing, 100 epochs, batch size 16 with accumulation; chosen by authors and affect final metrics.
  • ResNet18 embedding dimension D and GraphSAGE aggregation
    Feature dimension and neighborhood aggregation scheme are architectural free choices that define the latent space on which the three losses act.
  • Number and selection of DCE-MRI time points (T=4) = T=4
    Patients restricted to complete four-timepoint trajectories; early-prediction experiments vary T but the main claim uses the full set.
axioms (5)
  • domain assumption pCR is a clinically meaningful binary surrogate for long-term outcome after NACT
    Stated in Abstract and Introduction; labels the supervised evaluation target.
  • ad hoc to paper Directed acyclic graph with edges from every earlier to every later time point correctly encodes temporal treatment trajectories
    §2.1 Graph Topology Construction; design choice that enables message passing while forbidding reverse flow.
  • ad hoc to paper Non-responder embeddings should not be attracted to one another because their clinical outcomes are heterogeneous
    §2.2 Population-level Alignment; justifies the asymmetric loss that omits attract for non-responders.
  • domain assumption GraphSAGE message passing and ResNet18 global-average-pooled features are adequate backbones for 3D DCE-MRI volumes of shape (3,64,64,64)
    §2.1 and Implementation Details; standard components treated as given.
  • domain assumption Stratified 5-fold CV on the 585-patient ISPY-2 subset with complete four time points yields generalizable performance estimates
    §3 Experiments; no external test set is used.
invented entities (2)
  • Asymmetric Treatment Response Loss (L_align + L_decorrelate + L_temporal applied only to responders; L_repel to non-responders) no independent evidence
    purpose: Structure the latent space so responder trajectories are coherent, non-redundant, and temporally consistent while remaining separated from non-responders.
    Defined in §2.2; the three terms and their asymmetric application are the paper's main methodological invention. No independent evidence outside the reported ISPY-2 metrics.
  • Temporal DAG patient trajectory graph for DCE-MRI embeddings no independent evidence
    purpose: Provide a relational inductive bias for multi-timepoint aggregation via GNN message passing.
    §2.1; topology is a design choice specific to this work rather than a previously validated clinical construct.

pith-pipeline@v1.1.0-grok45 · 13820 in / 3458 out tokens · 24551 ms · 2026-07-11T11:32:59.036192+00:00 · methodology

0 comments
read the original abstract

In patients with breast cancer, pathological complete response (pCR) has been established as a clinically meaningful surrogate marker for long-term outcomes. While commonly treated with neoadjuvant chemotherapy (NACT), effective treatment decision-making remains challenging, as therapeutic response can vary substantially across patients, calling for predictive models capable of accurately estimating individualized treatment response. To address this, we propose an imaging-based 3D spatio-temporal framework for treatment response prediction that integrates a state-of-the-art graph neural network with relational modeling of temporal interactions across timepoints alongside three novel complementary self-supervised treatment trajectory representation learning objectives. Experiments across a cohort of 585 patients from the public ISPY-2 dataset demonstrate that our method substantially outperforms both vision and self-supervised learning baselines across several classification metrics. Alongside establishing a breast cancer pCR prediction benchmark, we include a principled ablation of our method and further introduce and empirically assess the impact of the available number of DCE-MRI timepoints per patient trajectory and the inclusion of inter-scan time-differences. Overall, our study substantiates the utility of clinically meaningful longitudinal medical imagaging modeling for predicting NACT-induced pCR. We will publicly share our code repository and a user-friendly PyPI library for dataset curation upon publication, effectively promoting reproducible open-source research.

Figures

Figures reproduced from arXiv: 2607.04912 by Daniel M. Lang, Ivana Jan\'i\v{c}kov\'a, Jan C. Peeken, Johannes Kiechle, Julia A. Schnabel, Karim Lekadir, Richard Osuala, Stefan M. Fischer.

Figure 1
Figure 1. Figure 1: Method overview. Longitudinal image-derived latent features are represented as a directed acyclic graph and aggregated through a graph neural network (GNN) projection head to obtain a compact patient-level embedding. Representation learning is guided by an asymmetric response-aware objective that combines (a) population-level alignment, (b) patient-level representation decorrelation, and (c) temporal consi… view at source ↗

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

27 extracted references · 3 linked inside Pith

  1. [1]

    Breast cancer: Global patterns of incidence, mortality, and survival

    World Health Organization. Breast cancer: Global patterns of incidence, mortality, and survival. https://www.who.int/news-room/fact-sheets/detail/breast-cancer,

  2. [2]

    Accessed: 2026-02-11

    2026

  3. [3]

    Predicting breast cancer re- sponse to neoadjuvant treatment using multi-feature MRI: results from the I-SPY 2 trail.NPJ Breast Cancer, 6(1):63, 2020

    Wen Li, David C Newitt, Jessica Gibbs, Lisa J Wilmes, Ella F Jones, Vignesh A Arasu, Fredrik Strand, Natsuko Onishi, Alex Anh-Tu Nguyen, John Kornak, Laura J Esserman, Don Berry, and Nola M Hylton. Predicting breast cancer re- sponse to neoadjuvant treatment using multi-feature MRI: results from the I-SPY 2 trail.NPJ Breast Cancer, 6(1):63, 2020

    2020

  4. [4]

    Laura M Spring, Geoffrey Fell, Andrea Arfe, Chandni Sharma, Rachel Greenup, Kerry L Reynolds, Barbara L Smith, Brian Alexander, Beverly Moy, Steven J Isakoff, Giovanni Parmigiani, Lorenzo Trippa, and Aditya Bardia. Pathologic complete response after neoadjuvant chemotherapy and impact on breast cancer recurrence and survival: a comprehensive meta-analysis...

    2020

  5. [5]

    Hylton and et al

    N.M. Hylton and et al. Locally advanced breast cancer: MR imaging for prediction of response to neoadjuvant chemotherapy-results from ACRIN 6657/ISPY trial. Radiology, 279(1):44–55, 2016

    2016

  6. [6]

    Yunsong Peng et al. Pretreatment DCE-MRI-based deep learning outperforms radiomics analysis in predicting pathologic complete response to neoadjuvant chemotherapy in breast cancer.Frontiers in Oncology, 12:846775, 2022

    2022

  7. [7]

    Deep learning prediction of pathologic complete response in breast cancer using MRI and other clinical data: a systematic review.Tomography, 8(6):2784– 2795, 2022

    Nabeeha Khan, Richard Adam, Pauline Huang, Takouhie Maldjian, and Tim Q Duong. Deep learning prediction of pathologic complete response in breast cancer using MRI and other clinical data: a systematic review.Tomography, 8(6):2784– 2795, 2022

    2022

  8. [8]

    Aaquib Syed, Richard Adam, Thomas Ren, Jinyu Lu, Takouhie Maldjian, and Tim Q Duong. Machine learning with textural analysis of longitudinal multi- parametric MRI and molecular subtypes accurately predicts pathologic complete response in patients with invasive breast cancer.PLOS One, 18(1):e0280320, 2023

    2023

  9. [9]

    Prediction of pathological complete response to chemotherapy for breast cancer using deep neural network with uncertainty quantification.Medical Physics, 51(12):9385–9393, 2024

    Bowen Jing, Kai Wang, Erich Schmitz, Shanshan Tang, Yunxiang Li, You Zhang, and Jing Wang. Prediction of pathological complete response to chemotherapy for breast cancer using deep neural network with uncertainty quantification.Medical Physics, 51(12):9385–9393, 2024

    2024

  10. [10]

    Breast multiparametric MRI for predic- tion of neoadjuvant chemotherapy response in breast cancer: the BMMR2 chal- lenge.Radiology: Imaging Cancer, 6(1):e230033, 2024

    Wen Li, Savannah C Partridge, David C Newitt, Jon Steingrimsson, Helga S Marques, Patrick J Bolan, Michael Hirano, Benjamin Aaron Bearce, Jayashree Kalpathy-Cramer, Michael A Boss, et al. Breast multiparametric MRI for predic- tion of neoadjuvant chemotherapy response in breast cancer: the BMMR2 chal- lenge.Radiology: Imaging Cancer, 6(1):e230033, 2024

    2024

  11. [11]

    Multimodal deep learning for predicting neoadjuvant treat- ment outcomes in breast cancer: a systematic review.Biology Direct, 20(1):72, 2025

    Eriseld Krasniqi, Lorena Filomeno, Teresa Arcuri, Gianluigi Ferretti, Simona Gas- parro, Alberto Fulvi, Arianna Roselli, Loretta D’Onofrio, Laura Pizzuti, Mad- dalena Barba, et al. Multimodal deep learning for predicting neoadjuvant treat- ment outcomes in breast cancer: a systematic review.Biology Direct, 20(1):72, 2025. 10 Kiechle et al

    2025

  12. [12]

    Temporal represen- tation learning of phenotype trajectories for pCR prediction in breast cancer

    Ivana Janíčková, Yen Y Tan, Thomas H Helbich, Konstantin Miloserdov, Zsuzsanna Bago-Horvath, Ulrike Heber, and Georg Langs. Temporal represen- tation learning of phenotype trajectories for pCR prediction in breast cancer. In International Conference on Medical Image Computing and Computer-Assisted In- tervention, pages 606–615. Springer, 2025

    2025

  13. [13]

    I-SPY 2 breast dynamic contrast enhanced MRI trial (ISPY2)(version 1).The Cancer Imaging Archive (TCIA), 2022

    Wen Li, DC Newitt, J Gibbs, LJ Wilmes, EF Jones, VA Arasu, F Strand, N Onishi, AAT Nguyen, J Kornak, et al. I-SPY 2 breast dynamic contrast enhanced MRI trial (ISPY2)(version 1).The Cancer Imaging Archive (TCIA), 2022

    2022

  14. [14]

    ACRIN 6698/I-SPY2 breast DWI dataset (2021).DOI: https://doi

    David C Newitt, Savannah C Partridge, Zheng Zhang, Jessica E Gibbs, Thomas L Chenevert, Mark A Rosen, Patrick J Bolan, Helga Marques, J Romanoff, L Cimino, and Nola M Hylton. ACRIN 6698/I-SPY2 breast DWI dataset (2021).DOI: https://doi. org/10.7937/tcia. kk02-6d95, 2025

    doi:10.7937/tcia 2021

  15. [15]

    Early prediction of neoadjuvant chemotherapy responsebyexploitingatransferlearningapproachonbreastDCE-MRIs.Scientific Reports, 11(1):14123, 2021

    Maria Colomba Comes, Annarita Fanizzi, Samantha Bove, Vittorio Didonna, Ser- gioDiotaiuti,DanieleLaForgia,AgneseLatorre,EugenioMartinelli,AriannaMen- cattini, Annalisa Nardone, et al. Early prediction of neoadjuvant chemotherapy responsebyexploitingatransferlearningapproachonbreastDCE-MRIs.Scientific Reports, 11(1):14123, 2021

    2021

  16. [16]

    M2fusion: Multi-time multimodal fusion for prediction of pathological complete response in breast cancer

    Song Zhang, Siyao Du, Caixia Sun, Bao Li, Lizhi Shao, Lina Zhang, Kun Wang, Zhenyu Liu, and Jie Tian. M2fusion: Multi-time multimodal fusion for prediction of pathological complete response in breast cancer. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 458–468. Springer, 2024

    2024

  17. [17]

    Longitudinal detec- tion of radiological abnormalities with time-modulated LSTM

    Ruggiero Santeramo, Samuel Withey, and Giovanni Montana. Longitudinal detec- tion of radiological abnormalities with time-modulated LSTM. InInternational Workshop on Deep Learning in Medical Image Analysis, pages 326–333. Springer, 2018

    2018

  18. [18]

    On tackling domain shift in breast MRI using only publicly-available data: Repro- ducible breast cancer segmentation and pCR prediction

    Jessica Kächele, Dimitrios Bounias, Alexandra Ertl, and Klaus Maier-Hein. On tackling domain shift in breast MRI using only publicly-available data: Repro- ducible breast cancer segmentation and pCR prediction. InDeep Breast Workshop on AI and Imaging for Diagnostic and Treatment Challenges in Breast Care, pages 310–319. Springer, 2025

    2025

  19. [19]

    DeepResidualLearning for Image Recognition

    KaimingHe,XiangyuZhang,ShaoqingRen,andJianSun. DeepResidualLearning for Image Recognition. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 770–778, 2016

    2016

  20. [20]

    Inductive Representation Learning on Large Graphs.Advances in Neural Information Processing Systems, 2017

    Will Hamilton, Zhitao Ying, and Jure Leskovec. Inductive Representation Learning on Large Graphs.Advances in Neural Information Processing Systems, 2017

    2017

  21. [21]

    Neural Message Passing for Quantum Chemistry

    Justin Gilmer, Samuel S Schoenholz, Patrick F Riley, Oriol Vinyals, and George E Dahl. Neural Message Passing for Quantum Chemistry. InInternational Confer- ence on Machine Learning, pages 1263–1272. PMLR, 2017

    2017

  22. [22]

    David C Newitt, Sheye O Aliu, Neil Witcomb, Gal Sela, John Kornak, Laura Esserman, and Nola M Hylton. Real-time measurement of functional tumor volume by MRI to assess treatment response in breast cancer neoadjuvant clinical trials: validation of the aegis SER software platform.Translational Oncology, 7(1):94–100, 2014

    2014

  23. [23]

    MONAI: An open-source framework for deep learning in healthcare.arXiv preprint arXiv:2211.02701, 2022

    M Jorge Cardoso, Wenqi Li, and Richard Nic Brown. MONAI: An open-source framework for deep learning in healthcare.arXiv preprint arXiv:2211.02701, 2022

    Pith/arXiv arXiv 2022

  24. [24]

    DINOv3.arXiv preprint arXiv:2508.10104, 2025

    Oriane Siméoni, Huy V Vo, Maximilian Seitzer, Federico Baldassarre, Maxime Oquab, Cijo Jose, Vasil Khalidov, Marc Szafraniec, Seungeun Yi, Michaël Rama- monjisoa, et al. DINOv3.arXiv preprint arXiv:2508.10104, 2025. Temporal Graph Representation Learning of Patient Trajectories 11

    Pith/arXiv arXiv 2025

  25. [25]

    SSL-AD: Spatiotemporal self-supervised learning for generalizability and adaptability across alzheimer’s prediction tasks and datasets.arXiv preprint arXiv:2509.10453, 2025

    Emily Kaczmarek, Justin Szeto, Brennan Nichyporuk, and Tal Arbel. SSL-AD: Spatiotemporal self-supervised learning for generalizability and adaptability across alzheimer’s prediction tasks and datasets.arXiv preprint arXiv:2509.10453, 2025

    Pith/arXiv arXiv 2025

  26. [26]

    Temporal representation learning of phenotype trajectories for pCR prediction in breast cancer

    cirmuw. Temporal representation learning of phenotype trajectories for pCR prediction in breast cancer. https://github.com/cirmuw/temporal-representation- learning/blob/a1c721872bb6c2caa4dbe7c352cd878e283e3dc8/README.md,

  27. [27]

    GitHub repository, commit a1c7218, accessed 25 Feb 2026

    2026