Pith. sign in

REVIEW 5 major objections 5 minor 27 references

CXR-TFT: Multi-Modal Temporal Fusion Transformer for Predicting Chest X-ray Trajectories

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Model forecasts abnormal chest X-ray findings up to 12 hours before they become radiographically evident.

desk verdict A genuinely interesting latent-space trajectory idea, but the training target leaks the future CXR, so the headline accuracy numbers are not trustworthy. read the letter →

arxiv 2507.14766 v1 pith:V52DBO7W submitted 2025-07-19 cs.LG cs.AIcs.CV

Mehak Arora , Ayman Ali , Kaiyuan Wu , Carolyn Davis , Takashi Shimazui , Mahmoud Alwakeel , Victor Moas , Philip Yang
show 2 more authors
Annette Esper Rishikesan Kamaleswaran
This is my paper · ORCID
classification cs.LGcs.AIcs.CV
keywords clinicaltrajectorypredictionchestX-raytemporalfusiontransformerlatentembeddinginterpolationICUmonitoringmulti-modalradiologyfindingsforecastingirregularlysampledtimeseries
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

CXR-TFT is a framework that tries to estimate a patient's most likely chest X-ray representation at every hour of an ICU stay, by fusing sporadic chest radiographs and radiology reports with hourly vitals, labs, and ventilator data. The paper's central claim is that this predicted representation detects abnormal findings—such as lung opacity, edema, consolidation, and pneumothorax—with about 95 percent accuracy 12 hours before the next actual chest X-ray shows them, and with higher AUROC than simply reusing the most recent CXR. If true, the model would add genuinely prospective, hour-level resolution to a test that is normally taken only on clinical demand, creating a window for earlier treatment of conditions like ARDS. The claim is a proof-of-principle for trajectory modelling in a continuous embedding space rather than in pixel space.

What carries the argument

The central object is the interpolated embedding trajectory defined by Equation 1: between two successive chest X-rays taken at hours t_k1 and t_k2, the target embedding for any intermediate hour is a straight-line interpolation of the two BioCLIP embeddings. This converts irregularly sampled imaging into hourly supervision and lets clinical measurements be aligned with imaging. The predictor is an encoder-decoder transformer (pre-norm, AdamW, cosine schedule) whose input at each hour concatenates an 82-dimensional clinical feature vector with the embedding of the last observed CXR; training minimizes a weighted MSE between predicted and interpolated target embeddings plus an auxiliary cross-entropy loss from an MLP classifier trained on CheXPert labels derived from radiology reports.

What would settle it

Take patients who happened to have a chest X-ray at an intermediate hour between the two CXRs used to build the interpolated target (for example, scans at hours 0, 6, and 12; use hours 0 and 12 to interpolate, keep the hour-6 scan held out). If the model's predicted embedding at hour 6 matches the held-out real X-ray's findings no better than a model that simply repeats the hour-0 findings, then the interpolation-based evaluation is the source of the reported accuracy. A second check: compute nearest-neighbour distances between interpolated embeddings and real CXR embeddings with the same reported findings; if interpolated points are far from any real embedding of the same findings, the latent space is not carrying the semantic content the paper assumes.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that the latent embedding trajectory between two real chest X-rays is learnable: given a patient's clinical measurements and the embedding of the most recent X-ray, an encoder-decoder transformer can predict the embedding at future hours with enough fidelity that a downstream classifier reads the same radiological findings from it as from a real X-ray. The quantitative claim is roughly 95 percent accuracy and AUROC up to 0.951 at 12 hours ahead and 94 percent accuracy at 24 hours ahead across ten finding classes, beating the previous-CXR baseline on most classes. The authors interpret this as effectively performing super-resolution in time on an infrequently sampled imaging modality.

Load-bearing premise

The framework rests on the assumption that a straight line through the latent embeddings of two real chest X-rays yields a meaningful chest X-ray representation at every hour in between, and that the vision encoder's embedding space is semantically meaningful for chest X-rays; the model is trained and evaluated against these interpolated points, so if the interpolation is invalid the reported 12-hour advantage is an artifact.

Editorial extensions

If this is right

  • Hour-level risk scores for findings like pneumonia, edema, and pneumothorax become possible between the sparse times when CXRs are actually taken.
  • A model that predicts in latent embedding space avoids the cost and hallucination risk of generating pixel-level images while still supporting classification of findings.
  • For ARDS, where delayed diagnosis is a known problem, a 12-hour advance warning from routinely collected ICU data could change when diagnostic imaging is ordered.
  • The same interpolation-and-predict recipe could be applied to other irregularly sampled imaging modalities or to any paired sensor stream with sparse high-dimensional observations.
  • Because the model uses only the most recent CXR plus clinical history, it is applicable at the bedside without waiting for a current radiology report.

Reading between the lines

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

  • The headline accuracy partly reflects evaluation against interpolated targets, which are by construction smooth and easy to predict; a fairer test would use real intermediate CXRs where they exist.
  • The interpolation assumption is the fragile link; a direct check would compare interpolated embeddings of normal-to-abnormal pairs against real radiograph embeddings to confirm the latent space carries the semantic content the paper assumes.
  • The method's value would be more convincing if compared against a purely clinical baseline that predicts future report labels directly from the same hourly vitals and labs, removing the imaging embeddings from the equation.
  • A natural extension is to decode predicted embeddings back into pixel-space CXRs, which would turn the trajectory model into a full forecasting system and allow visual inspection of why a finding was predicted.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes CXR-TFT, a transformer-based framework that combines hourly clinical measurements with latent embeddings of sparsely recorded chest X-rays to predict an estimated CXR embedding at every hour of an ICU stay. The predicted embeddings are passed through a classifier trained on radiology-report labels to estimate the likelihood of ten radiographic findings. On a single-center retrospective cohort, the authors report roughly 95% accuracy in detecting findings at predicted embeddings 12 hours before the next CXR and improvements over a baseline that uses the most recently recorded CXR. The central technical device is linear interpolation between successive CXR embeddings in the latent space of a pretrained vision-language encoder, used both to create dense hourly training targets and to evaluate 'Current Prediction' performance.

Significance. If the claims were valid, the model would be a clinically valuable early-warning tool for conditions such as ARDS, pneumonia, and pulmonary edema, and the idea of predicting CXR trajectories in a continuous embedding space would be a useful methodological direction. The paper also has strengths: it uses a large real ICU cohort, compares against a previous-CXR baseline, and includes a classifier-regularization objective tied to radiology labels. However, the significance is undermined by two load-bearing problems: the cited BioCLIP encoder is not a radiology model, and the 'Current Prediction' evaluation is circular because it compares predictions against interpolated targets that already include future CXR information. These issues directly affect the headline accuracy and AUROC numbers, so the contribution as written is not established. The manuscript does not provide machine-checked proofs or a working code link, so reproducibility rests on the unverifiable claim that the repository will be made available.

major comments (5)
  1. [Section 2.3, 'Image Encoding'; Ref. [23]] The manuscript states that 'BioCLIP [23], a vision language model trained to align radiology reports with corresponding image embeddings, was used to extract the latent space representation.' Reference [23] is Stevens et al., 'BioCLIP: A vision foundation model for the tree of life' (CVPR 2024), a model trained on images of organisms, not chest radiographs. There is no evidence that this encoder produces a semantically meaningful latent space for CXR findings. Since Eq. (1), the classifier regularization in Section 2.5, and every metric in Table 1 depend on this encoder, the central results are unsupported unless the authors identify and verify the actual CXR-aligned encoder used. This is a factual, checkable flaw that cannot be resolved by a citation correction alone.
  2. [Section 2.3, Eq. (1); Section 3, 'Current Prediction'] The training target I_T_t is defined by linear interpolation between the embeddings of two actual CXRs, so the target at every intermediate hour includes information from the future CXR. The 'Current Prediction' evaluation compares classifier outputs on predicted embeddings against labels derived from these same interpolated target trajectories. This evaluation is therefore circular: it rewards the model for reproducing a future-dependent interpolation rather than for genuinely predicting a future CXR. The reported 95% accuracy and the associated AUROC values do not, by themselves, support the abstract's claim of forecasting abnormal findings 12 hours before they become radiographically evident.
  3. [Section 3, Table 1] The 'Future Prediction' columns are more independent than 'Current Prediction' because they compare against ground-truth labels from the subsequent CXR report, but the time horizon is ambiguous. For a prediction made at hour t, is the comparison made against the CXR at t+12, t+24, or the next recorded CXR regardless of interval? The table also does not report confidence intervals or statistical significance for the differences between the model and the previous-CXR baseline. Given the large cohort size, such intervals and tests are necessary to support the headline claim of 'high accuracy' and 'significant improvement' over baseline.
  4. [Sections 2.2 and Abstract] The cohort size is stated as 17,690 in Section 2.2 but as '20,000 ICU patients' in the abstract. The manuscript should reconcile these numbers. In addition, no patient-level train/validation/test split is described, and no statement is made about whether multiple CXRs from the same patient can appear in both training and evaluation. Without such a split, the risk of data leakage across a patient's ICU stay is substantial, which would inflate the reported accuracies.
  5. [Section 2.5, Eq. (3)] The notation in Eq. (3) is inconsistent with the surrounding text. The text says the loss is between 'predicted labels of the decoder output and the target labels,' but Eq. (3) defines y_i,c,t = MLP(I_T_t) and p_i,c,t = MLP(I_p_t), with p used as the prediction and y as the target. If I_p_t is the previous-CXR embedding, the prediction source is unclear. Please clarify which embeddings are decoder outputs and which are targets, and align the equation with the text.
minor comments (5)
  1. [Section 2.3, 'Data Preprocessing'] Typo: 'All clinical measurements from the Electronic Medical Record (EMR) are were organized into hourly bins.' The word 'are' should be removed.
  2. [Figure 1 and Section 4] The model name is inconsistent: the title and abstract use 'CXR-TFT,' but Figure 1 and several places in Section 4 use 'CXR-TLT.' Please use one name throughout.
  3. [Section 2.4, 'Training CXR-TFT'] The text says 'the code to the complete data processing and training setup can be found at our Github Repository,' but no URL is provided. Please include the repository link.
  4. [Eq. (1)] The notation in Eq. (1) is confusing because both the timestamp indices k1 and k2 and the embedding subscripts I_T_k1 and I_T_k2 are used; the definition of k′ is also implicit. Please define all indices explicitly.
  5. [Figure 3] The caption says 'time prior to confirmation on subsequent chest X-rays,' but the axis labels and units of the horizontal axis are not described in the text. Please clarify how '12-hours in advance' and '24-hours in advance' map to the figure.

Circularity Check

2 steps flagged · score 6.0 of 10

Interpolation targets in Eq. (1) leak the future CXR into every hourly label, so the 12-hour 'forecast' and the Current Prediction metrics reduce to the paper's own training construction.

  1. self definitional [Section 2.1, Eq. (1); Abstract; Section 3 'Future Prediction']
    "The target output sequence ... used to train the model is obtained by linear interpolation in the embedding space between two recorded CXRs. Concretely, if a CXR scan was performed at time t_{k1}, and the next CXR scan was performed at t_{k2}, then ... I^T_{k'} = ((I^T_{k2}-I^T_{k1})/(k2-k1)) x (k' - k1) + I^T_{k1}, if k1 < k' < k2 (1). ... CXR-TFT demonstrated high accuracy in forecasting abnormal CXR findings up to 12 hours before they became radiographically evident."

    Every hourly training target between two CXRs is a linear blend that includes the embedding of the later, future CXR. At a time 12 hours before the next scan, roughly half of the target vector is the future scan's embedding (or more, depending on scan spacing). The transformer is therefore trained to output embeddings that already contain information from the CXR whose findings are later used as ground truth in the Future Prediction comparison. The claimed 12-hour lead time is thus an artifact of the interpolation schedule: the model is not forecasting from prior data alone but reproducing a target constructed with the future image. This is a definitional reduction of the central predictive claim, not an independently discovered early-detection capability.

  2. fitted input called prediction [Section 2.5, Eq. (4); Section 3 'Current Prediction']
    "The cross-entropy loss between predicted labels of the decoder output and the target labels was added to the training objective. ... The 'Current Prediction' results evaluate model performance by comparing predicted labels with labels derived from interpolated target CXR trajectories."

    The interpolated trajectories from Eq. (1) are used both as the supervision for the classifier-regularization loss L_BCE in Eq. (4) and as the source of labels for the Current Prediction evaluation. Thus the Current Prediction AUROC/AUPRC/accuracy values in Table 1 measure how well the model reproduces the same artificial interpolation labels on which it was trained. These numbers are a training-fit diagnostic, not an independent accuracy estimate against radiological truth. The evaluation is circular because the 'predicted labels' and the 'target labels' are both generated from the same interpolated embeddings by the same MLP.

full rationale

The paper's headline capability — forecasting abnormal CXR findings 12 hours before they become radiographically evident — is contaminated at the level of target construction. Equation (1) defines the supervision at every intermediate hour as a linear interpolation between the previous and next CXR embeddings, so the training target at hour t_k' already contains a fraction of the future CXR embedding. The Future Prediction evaluation then compares against labels from that same future CXR, making the reported lead time partially forced by the interpolation schedule rather than by genuinely predictive clinical signal. Separately, the Current Prediction metrics compare classifier outputs on predicted embeddings against labels derived from the same interpolated trajectories used in training, so those numbers are fit-to-training-construction statistics, not independent predictions. Both issues are exhibited directly from the paper's equations and text, so they meet the threshold for circularity. I assign 6 rather than 8-10 because the model still must map clinical features to the interpolated space and the Future Prediction comparison does use real radiology-report labels; the central claim is partially, not wholly, reducible to construction. I also note, without counting it as circularity, that the paper describes BioCLIP [23] as 'a vision language model trained to align radiology reports' while reference [23] is the tree-of-life BioCLIP model; that is a factual correctness risk about the encoder and the validity of the interpolation targets, but it is not a self-referential or definitional circularity. No load-bearing self-citation chain was found.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The paper introduces no physical entities, but it relies on two unsupported modeling constructs: a latent space whose semantic meaning for chest X-rays is assumed, and a linear-interpolation target that borrows future CXR content. The free parameters are hyperparameters and preprocessing choices whose influence on the headline result is not examined.

free parameters (3)
  • alpha (classifier regularization weight) = 0.5
    Hand-chosen in Eq. 4 to balance MSE and BCE losses; no sensitivity analysis or tuning curve is reported.
  • imputation value for missing clinical measurements = median of healthy reference range
    Section 2.3 uses this for all missing hourly values; no comparison to other imputation strategies or analysis of imputation rate.
  • hourly binning window = 1 hour
    Clinical measurements are aggregated into hourly bins; the choice of window affects alignment and is not varied.
assumptions (5)
  • ad hoc to paper Linear interpolation between two CXR embeddings in the encoder's latent space yields valid intermediate CXR representations (Eq. 1).
    No validation that interpolated embeddings correspond to plausible radiographs; the entire training target and part of the evaluation depend on this.
  • domain assumption The vision encoder cited as BioCLIP provides semantically meaningful, radiology-aligned embeddings for chest X-rays.
    Section 2.3 claims BioCLIP aligns radiology reports with image embeddings, but reference [23] is BioCLIP for the tree of life, trained on organism images, not CXRs.
  • domain assumption Class labels generated by CheXPert from radiology reports are accurate ground truth for the 10 findings.
    Section 2.3 relies on an automated labeler without clinician adjudication or error analysis.
  • domain assumption The classifier trained on MIMIC-CXR transfers to this single-center cohort and to model-predicted embeddings.
    No distribution analysis or manual review of predicted embeddings is provided; predicted embeddings may be out-of-distribution.
  • domain assumption Hourly forward-fill and median imputation preserve clinically meaningful temporal structure.
    Section 2.3 aggregates vitals and labs into hourly bins and imputes missing values; no comparison to other imputation or modeling strategies.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CXR-TFT: Multi-Modal Temporal Fusion Transformer for Predicting Chest X-ray Trajectories." pith.science (2026). https://pith.science/paper/V52DBO7W

@misc{pith2026250714766,
  author       = {Pith},
  title        = {Pith review of: CXR-TFT: Multi-Modal Temporal Fusion Transformer for Predicting Chest X-ray Trajectories},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/V52DBO7W}},
  note         = {Machine review of arXiv:2507.14766}
}
read the original abstract

In intensive care units (ICUs), patients with complex clinical conditions require vigilant monitoring and prompt interventions. Chest X-rays (CXRs) are a vital diagnostic tool, providing insights into clinical trajectories, but their irregular acquisition limits their utility. Existing tools for CXR interpretation are constrained by cross-sectional analysis, failing to capture temporal dynamics. To address this, we introduce CXR-TFT, a novel multi-modal framework that integrates temporally sparse CXR imaging and radiology reports with high-frequency clinical data, such as vital signs, laboratory values, and respiratory flow sheets, to predict the trajectory of CXR findings in critically ill patients. CXR-TFT leverages latent embeddings from a vision encoder that are temporally aligned with hourly clinical data through interpolation. A transformer model is then trained to predict CXR embeddings at each hour, conditioned on previous embeddings and clinical measurements. In a retrospective study of 20,000 ICU patients, CXR-TFT demonstrated high accuracy in forecasting abnormal CXR findings up to 12 hours before they became radiographically evident. This predictive capability in clinical data holds significant potential for enhancing the management of time-sensitive conditions like acute respiratory distress syndrome, where early intervention is crucial and diagnoses are often delayed. By providing distinctive temporal resolution in prognostic CXR analysis, CXR-TFT offers actionable 'whole patient' insights that can directly improve clinical outcomes.

Figures

Figures reproduced from arXiv: 2507.14766 by the authors.

Figure 1
Figure 1. Visual Depiction of the CXR-TLT Framework.(a) Sparsely recorded CXR [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Performance comparison of detecting radiographic findings on the em [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Temporal trends in AUROC for predicting radiological findings as a func [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 18 canonical work pages

  1. [23]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Stevens, S., Wu, J., Thompson, M.J., Campolongo, E.G., Song, C.H., Carlyn, D.E., Dong, L., Dahdul, W.M., Stewart, C., Berger-Wolf, T., et al.: Bioclip: A vision foundation model for the tree of life. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 19412–19424 (2024)

    2024

  2. [1]

    Diagnostics13(4), 743 (2023)

    Ahmad, H.K., Milne, M.R., Buchlak, Q.D., Ektas, N., Sanderson, G., Chamtie, H., Karunasena, S., Chiang, J., Holt, X., Tang, C.H., et al.: Machine learning augmented interpretation of chest x-rays: a systematic review. Diagnostics13(4), 743 (2023)

    work page 2023

  3. [2]

    Journal of Korean Medical Science40 (2024)

    Ahn, D.W., Seo, Y., Goo, T., Jeong, J.B., Park, T., Yoon, S.H.: Temporal radio- graphic trajectory and clinical outcomes in covid-19 pneumonia: A longitudinal study. Journal of Korean Medical Science40 (2024)

    work page 2024

  4. [3]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Bannur, S., Hyland, S., Liu, Q., Perez-Garcia, F., Ilse, M., Castro, D.C., Boecking, B., Sharma, H., Bouzid, K., Thieme, A., et al.: Learning to exploit temporal struc- ture for biomedical vision-language processing. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 15016–15027 (2023)

    2023

  5. [4]

    Intensive care medicine46, 1180–1183 (2020)

    Bellani, G., Pham, T., Laffey, J.G.: Missed or delayed diagnosis of ards: a common and serious problem. Intensive care medicine46, 1180–1183 (2020)

    work page 2020

  6. [5]

    Nature Biomedical Engineering pp

    Bluethgen, C., Chambon, P., Delbrouck, J.B., van der Sluijs, R., Połacin, M., Zambrano Chaves, J.M., Abraham, T.M., Purohit, S., Langlotz, C.P., Chaudhari, A.S.: A vision–language foundation model for the generation of realistic chest x-ray images. Nature Biomedical Engineering pp. 1–13 (2024)

    work page 2024

  7. [6]

    In: European conference on computer vision

    Boecking, B., Usuyama, N., Bannur, S., Castro, D.C., Schwaighofer, A., Hyland, S., Wetscherek, M., Naumann, T., Nori, A., Alvarez-Valle, J., et al.: Making the most of text semantics to improve biomedical vision–language processing. In: European conference on computer vision. pp. 1–21. Springer (2022)

    2022

  8. [7]

    Intensive care medicine 34, 264–270 (2008)

    Clec’h, C., Simon, P., Hamdi, A., Hamza, L., Karoubi, P., Fosse, J.P., Gonzalez, F., Vincent, F., Cohen, Y.: Are daily routine chest radiographs useful in critically ill, mechanically ventilated patients? a randomized study. Intensive care medicine 34, 264–270 (2008)

    work page 2008

Show all 27 references
  1. [8]

    Biomedical engi- neering online 21(1), 77 (2022)

    Duanmu, H., Ren, T., Li, H., Mehta, N., Singer, A.J., Levsky, J.M., Lipton, M.L., Duong, T.Q.: Deep learning of longitudinal chest x-ray and clinical variables pre- dicts duration on ventilator and mortality in covid-19 patients. Biomedical engi- neering online 21(1), 77 (2022)

    2022

  2. [9]

    Scientific reports 12(1), 5616 (2022)

    Gourdeau, D., Potvin, O., Archambault, P., Chartrand-Lefebvre, C., Dieumegarde, L., Forghani, R., Gagné, C., Hains, A., Hornstein, D., Le, H., et al.: Tracking and predicting covid-19 radiological trajectory on chest x-rays using deep learning. Scientific reports 12(1), 5616 (2022)

    2022

  3. [10]

    Intensive care medicine33, 639–644 (2007)

    Graat, M.E., Kröner, A., Spronk, P.E., Korevaar, J.C., Stoker, J., Vroom, M.B., Schultz, M.J.: Elimination of daily routine chest radiographs in a mixed medical– surgical intensive care unit. Intensive care medicine33, 639–644 (2007)

    2007

  4. [11]

    arXiv preprint arXiv:2312.00752 (2023)

    Gu, A., Dao, T.: Mamba: Linear-time sequence modeling with selective state spaces. arXiv preprint arXiv:2312.00752 (2023)

    2023 arXiv

  5. [12]

    Gu, Y., Yang, J., Usuyama, N., Li, C., Zhang, S., Lungren, M.P., Gao, J., Poon, H.: Biomedjourney: Counterfactual biomedical image generation by instruction- learning from multimodal patient journeys (2023),https://arxiv.org/abs/2310. 10765

    2023

  6. [13]

    Critical Care24, 1–9 (2020)

    Gutierrez, G.: Artificial intelligence in the intensive care unit. Critical Care24, 1–9 (2020)

    2020

  7. [14]

    In: Proceedings of the AAAI conference on artificial intelligence

    Irvin, J., Rajpurkar, P., Ko, M., Yu, Y., Ciurea-Ilcus, S., Chute, C., Marklund, H., Haghgoo, B., Ball, R., Shpanskaya, K., et al.: Chexpert: A large chest radiograph dataset with uncertainty labels and expert comparison. In: Proceedings of the AAAI conference on artificial in...

    2019

  8. [15]

    Scientific data 6(1), 317 (2019)

    Johnson, A.E., Pollard, T.J., Berkowitz, S.J., Greenbaum, N.R., Lungren, M.P., Deng, C.y., Mark, R.G., Horng, S.: Mimic-cxr, a de-identified publicly available database of chest radiographs with free-text reports. Scientific data 6(1), 317 (2019)

    2019

  9. [16]

    König, I.R., Fuchs, O., Hansen, G., von Mutius, E., Kopp, M.V.: What is precision medicine? European respiratory journal50(4) (2017)

    2017

  10. [17]

    Chest 123(5), 1607–1614 (2003)

    Krivopal, M., Shlobin, O.A., Schwartzstein, R.M.: Utility of daily routine portable chest radiographs in mechanically ventilated patients in the medical icu. Chest 123(5), 1607–1614 (2003)

    2003

  11. [18]

    Kyung, D., Kim, J., Kim, T., Choi, E.: Towards predicting temporal changes in a patient’s chest x-ray images based on electronic health records (2024)

    2024

  12. [19]

    Journal of the American College of Radiology18(5), S62–S72 (2021)

    Laroia, A.T., Donnelly, E.F., Henry, T.S., Berry, M.F., Boiselle, P.M., Colletti, P.M., Kuzniewski, C.T., Maldonado, F., Olsen, K.M., Raptis, C.A., et al.: Acr appropriateness criteria® intensive care unit patients. Journal of the American College of Radiology18(5), S62–S72 (2021)

    2021

  13. [20]

    arXiv preprint arXiv:1301.3781 (2013)

    Mikolov, T., Chen, K., Corrado, G., Dean, J.: Efficient estimation of word repre- sentations in vector space. arXiv preprint arXiv:1301.3781 (2013)

    2013 arXiv

  14. [21]

    arXiv preprint arXiv:2405.15442 (2024)

    Rasekh, A., Heidari, R., Rezaie, A.H.H.M., Sedeh, P.S., Ahmadi, Z., Mitra, P., Nejdl, W.: Towards precision healthcare: Robust fusion of time series and image data. arXiv preprint arXiv:2405.15442 (2024)

    2024 arXiv

  15. [22]

    Intensive care medicine47, 750–760 (2021)

    van de Sande, D., van Genderen, M.E., Huiskens, J., Gommers, D., van Bommel, J.: Moving from bytes to bedside: a systematic review on the use of artificial intelligence in the intensive care unit. Intensive care medicine47, 750–760 (2021)

    2021

  16. [24]

    In: Seminars in Respiratory and Critical Care Medicine

    Toy, D., Siegel, M.D., Rubinowitz, A.N.: Imaging in the intensive care unit. In: Seminars in Respiratory and Critical Care Medicine. vol. 43, pp. 899–923. Thieme Medical Publishers, Inc. (2022)

    2022

  17. [25]

    Tölle, M., Scharaf, M., Fischer, S., Reich, C., Zeid, S., Dieterich, C., Meder, B., Frey, N., Wild, P., Engelhardt, S.: Arbitrary data as images: Fusion of patient data across modalities and irregular intervals with vision transformers (2025),https: //arxiv.org/abs/2501.18237

    2025 arXiv

  18. [26]

    Advances in neural information pro- cessing systems 30 (2017)

    Vaswani,A.,Shazeer,N.,Parmar,N.,Uszkoreit,J.,Jones,L.,Gomez,A.N.,Kaiser, Ł., Polosukhin, I.: Attention is all you need. Advances in neural information pro- cessing systems 30 (2017)

    2017

  19. [27]

    In: arXiv preprint arXiv:2308.01317 (2023)

    Xu, S., Yang, L., Kelly, C., Sieniek, M., Kohlberger, T., Ma, M., Weng, W.H., Kiraly, A., Kazemzadeh, S., Melamed, Z., et al.: Elixr: Towards a general purpose x-ray artificial intelligence system through alignment of large language models and radiology vision encoders. In: ar...

    2023 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.