Pith. sign in

REVIEW 3 major objections 5 minor 2 references

A computationally frugal open-source foundation model for thoracic disease detection in lung cancer screening programs

T0 review · 3 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 This paper claims that TANGERINE, a lightweight open-source 3D masked autoencoder pretrained on 98,588 lung CT volumes, matches or beats heavier closed models on 14 thoracic disease classification tasks while needing far less labelled…

desk verdict A useful, frugal 3D MAE resource for LDCT, but the SOTA claim is overstated by a leaky CTRate seen-distribution benchmark and an unfair Gemini comparison. read the letter →

arxiv 2507.01881 v2 pith:BXP66UPW submitted 2025-07-02 eess.IV cs.CVcs.LG

Niccolò McConnell , Pardeep Vasudev , Daisuke Yamada , Daryl Cheng , Mehran Azimbagirad , John McCabe , Shahab Aslani , Ahmed H. Shahin
show 8 more authors
Yukun Zhou The SUMMIT Consortium Andre Altmann Yipeng Hu Paul Taylor Sam M. Janes Daniel C. Alexander Joseph Jacob
This is my paper
classification eess.IVcs.CVcs.LG
keywords 3Dmaskedautoencoderlow-doseCTlungcancerscreeningself-supervisedpretrainingthoracicdiseaseclassificationlabelefficiencydomaingeneralisationopen-sourcefoundationmodel
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

This paper introduces TANGERINE, an open-source 3D masked autoencoder pretrained on 98,588 low-dose CT volumes drawn from the SUMMIT lung screening study and 27 public datasets. The central claim is that this lightweight model reaches state-of-the-art accuracy on 14 thoracic disease classification tasks, including within-one-year lung cancer, interstitial lung abnormalities, bronchiectasis, and multi-label chest disease, while generalising to unseen clinical centres and needing far less labelled data and GPU time than comparators. If true, it would let lung cancer screening programmes pivot from cancer-only reading to comprehensive respiratory disease detection using models that run locally on modest hardware, avoiding cloud dependence. The paper also reports that frozen embeddings from the pretrained encoder already support competitive classification through a shallow network.

What carries the argument

The engine is a 3D masked autoencoder, a self-supervised learner that hides most of a volume and learns to reconstruct it. Volumes are resampled to $256^3$ and split into $16^3$ patches; 75% are masked, and a ViT-large encoder (24 blocks, embedding size 1,024, about 312 million parameters) processes only the visible patches, which cuts memory by roughly half and runtime by a factor of six relative to encoding all patches. A small 8-block decoder reconstructs the masked sub-volumes from the encoded visible patches and learned mask tokens, forcing the model to learn global thoracic anatomy. For downstream tasks the decoder is discarded and the encoder is fine-tuned end-to-end with a linear head, or a two-layer head for sparse cancer datasets, on a 2048-dimensional representation formed by concatenating the class token with global-average-pooled patch tokens. Frozen versions of this representation, classified by a shallow neural network, already perform competitively, and reconstructions show coherent recovery of bronchi, vessels, and pleura from heavily masked input.

What would settle it

Run the same 14-task benchmark with CT-Foundation-Gemini fine-tuned end-to-end on the same data splits; if it matches or exceeds TANGERINE's AUROCs, the state-of-the-art claim fails. Also, pretrain a TANGERINE variant without the 46,330 CTRate training volumes and retest on CTRate; a large drop would show the seen-distribution result depends on pretraining overlap rather than transfer.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a computationally frugal 3D masked autoencoder, in which the encoder sees only 25% of input patches during pretraining, can produce representations that match or beat much heavier and partly closed foundation models. Across pretrain-seen, pretrain-unseen, and domain-generalisation settings, TANGERINE achieves the highest mean AUROC on most tasks, with 0.894 for within-one-year cancer on SUMMIT and a statistical tie with CT-Foundation-Gemini on RadChest while surpassing it on the cancer class. With only 10% of fine-tuning data it can exceed comparators trained on 100%, and it converges in 5 to 10 epochs, cutting compute and carbon impact relative to training from scratch. The paper frames this as evidence that architectural simplicity, large-scale self-supervised pretraining, and open release are sufficient for state-of-the-art thoracic disease detection.

Load-bearing premise

The headline result assumes the comparison is fair: TANGERINE gets full end-to-end fine-tuning while its strongest closed competitor is only adapted from frozen cloud embeddings plus a small neural network, and on the CTRate benchmark the same 46,330 scans used for pretraining are also used for fine-tuning.

Editorial extensions

If this is right

  • Lung cancer screening programmes could run comprehensive multi-disease classification on-site with a single GPU, and even CPU inference for frozen embeddings, without sending scans to cloud APIs.
  • A centre with only 10 to 40 percent of the usual labelled data could fine-tune TANGERINE to match or beat models trained from scratch on full data, lowering annotation cost.
  • Models fine-tuned in one hospital should transfer to other centres and countries on tasks not seen during pretraining, reducing the need for per-site labelled retraining.
  • Fine-tuning reaches peak validation performance within 5 to 10 epochs, making iterative model updates for newly defined imaging phenotypes cheap in both energy and carbon.
  • The released weights and pipeline position the model for extension to segmentation, report generation, and quantitative scoring, not just classification.

Reading between the lines

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

  • The paper never end-to-end fine-tunes its strongest closed competitor; until that comparison is run, the magnitude of TANGERINE's advantage over a fully adapted cloud model is untested.
  • On the CTRate seen-distribution task, the 46,330 fine-tuning volumes are the same scans used in pretraining, so gains there may reflect scan-level exposure rather than transferable features; an ablation that pretrains without CTRate would isolate the true transfer effect.
  • The paper acknowledges that $256^3$ downsampling may lose fine detail and that pretraining demographics are narrow; these caveats mean its generalisation and clinical-readiness claims await prospective multi-centre validation.
  • Because the authors release the 3D MAE code and weights, the same pretraining recipe could be probed on other volumetric modalities, but the paper only demonstrates thoracic CT, so cross-modality transfer remains an open conjecture.
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

3 major / 5 minor

Summary. The paper introduces TANGERINE, a 3D masked-autoencoder vision transformer (ViT-large) pretrained on 98,588 LDCT volumes from the SUMMIT study and 27 public datasets. The authors fine-tune the encoder on a range of thoracic disease classification tasks and evaluate performance in three settings: pretrain-seen distributions, pretrain-unseen distributions, and domain generalisation. They compare against several baselines including CT-Foundation-Gemini, CT-Clip, Med3D, ResNet, ViT, and CT-Cancer-Foundation, report label-efficiency results at 100%/40%/10% fine-tuning data, and analyse training dynamics, computational cost, and interpretability. The central claim is that TANGERINE achieves state-of-the-art performance across 14 disease classification tasks while being open-source, computationally frugal, and label-efficient.

Significance. If the reported results hold, TANGERINE would be a valuable community resource: it provides open-source code and weights, requires only four A6000 GPUs for pretraining, and demonstrates strong performance across multiple thoracic diseases and clinical centres. The empirical methodology is generally rigorous, with five random seeds, 95% confidence intervals, and Bonferroni-corrected pairwise tests; the pretraining-size ablations and the frozen-embedding results are informative. However, the headline 'state-of-the-art' claim rests on two protocol choices that appear to favour TANGERINE relative to its closest comparators, and the significance of the contribution is therefore conditional on a fair re-evaluation.

major comments (3)
  1. [Methods, 'Datasets for developing TANGERINE' and 'Downstream task datasets'; Fig. 1] The pretrain-seen-distribution benchmark for CTRate is contaminated. The Methods state that the CTRate training set of 46,330 scans 'was utilised for pretraining TANGERINE', and the Downstream task datasets section then describes splitting this same 46,330-scan set 85:15 into training and validation for fine-tuning on Multi-Disease-CTRate. Because TANGERINE's masked-autoencoder pretraining reconstructs these exact volumes, the model has an unsupervised prior over the fine-tuning volumes; the from-scratch ViT, ResNet-50, and CT-Cancer-Foundation baselines do not share this prior (CT-Clip, which was also pretrained on CTRate, is a partial exception). This contradicts Fig. 1's assertion of 'ensuring subject exclusion' for pretrain-seen tasks and inflates the CTRate results in Fig. 2a and the label-efficiency results at 40% and 10% in Fig. 3a. Please re-run the CTRate experiments with a version of TANGERINE pretrained on the full pretraining set excluding all CTRate volumes, or equivalently fine-tune on a CTRate split whose subjects were never seen during pretraining, and report both the contaminated and clean numbers.
  2. [Methods, 'Comparison models'] The comparison against CT-Foundation-Gemini is asymmetric. Gemini is evaluated only through frozen cloud embeddings followed by an MLP, while TANGERINE is fine-tuned end-to-end. The Results text says 'All models were fine-tuned and evaluated on the same datasets', but the Methods specify that Gemini embeddings were 'fine-tuned using the equivalent multilayer perceptron (MLP) as used for frozen TANGERINE embeddings'. Since end-to-end fine-tuning generally improves over frozen embeddings—as the authors' own TANGERINE-Frozen results demonstrate—this protocol choice can only reduce Gemini's measured performance relative to TANGERINE. Please either obtain end-to-end fine-tuned Gemini results where the API permits, or restrict the headline comparison to the frozen-embedding setting (TANGERINE-Frozen versus Gemini-Frozen) and report both comparisons. In addition, Gemini was not evaluated on any SUMMIT task, so the reported SUMMIT results are not benchmarked against the closest external model.
  3. [Abstract and Results, 'Lung Disease Diagnosis'] The claim of 'state-of-the-art performance across 14 disease classification tasks' is stronger than the evidence supports. On the pretrain-unseen RadChest benchmark, TANGERINE 'matched CT-Foundation-Gemini mean AUROC across 17 classes (P=1.000)', and in the domain-generalisation setting CT-Foundation-Gemini exceeded TANGERINE on calcification (P=0.001). These are not instances of outperforming the state of the art. The abstract and conclusions should be reworded to say 'comparable or superior' performance, and the specific settings in which TANGERINE is SOTA versus comparable should be enumerated.
minor comments (5)
  1. [Extended Data Fig. 2 caption] The caption states 'Performance on CTRate classes in the pretrain-unseen-distribution setting', but the text describes evaluation on RadChest, a dataset not seen during pretraining. The caption should refer to RadChest, not CTRate.
  2. [Discussion, paragraph 4] The sentence 'TANGERINE's ability to detect early imaging phenotypes, such as interstitial lung abnormalities (ILAs), could help could support identification and enrichment of patient cohorts' contains a duplicated phrase 'could help could support'. It should read 'could help support'.
  3. [Data and Materials Availability] The statement 'model weights are available upon request during the review process' is in tension with the paper's 'open-source' claim. Please state the exact licence and provide a public, permanent URL for the weights, or clarify the conditions under which they will be released.
  4. [Methods, 'Downstream task datasets'] The description of the CTRate split ('we split the original training set into training and validation using an 85:15 split and kept the original provided test set') should explicitly state whether the split is subject-wise and should clarify that the training/validation volumes are the same as those used for pretraining. This is important for interpreting the pretrain-seen results.
  5. [Methods, 'Datasets for developing TANGERINE'] The sentence 'Since SUMMIT's distribution was seen during pretraining, it is considered in the pretrain-seen-distribution task dataset' is vague. Please specify whether any SUMMIT scans were used for pretraining and how the fine-tuning subsets are disjoint at the subject level from the pretraining scans.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: TANGERINE's claims are empirical benchmark results with no derivation chain that reduces to its own inputs.

full rationale

The paper is an empirical systems paper rather than a mathematical derivation, so the standard circularity failure modes do not arise. Its central claims—state-of-the-art AUROC across 14 tasks, label efficiency, and fast convergence—are supported by measurements on held-out test splits and by comparisons against external baseline models. I checked each enumerated pattern. (1) No quantity is defined in terms of the target result; AUROC and AUPRC are standard metrics applied to test sets that are explicitly held out (e.g., the CTRate test set of 3,040 scans is kept separate, and SUMMIT fine-tuning subsets are drawn from a partition unseen during pretraining). (2) No fitted parameter is renamed as a prediction; test predictions are made on withheld examples rather than on the fitted data. (3) The only author-overlapping citation is RetFound (ref. 25), used to motivate the MAE pretraining strategy; this is architectural inspiration, not a load-bearing theorem or uniqueness argument. (4) No uniqueness theorem from the authors' prior work is invoked to force a choice. (5) The MAE framework, 3D patch embedding, and 75% masking ratio are presented as explicit engineering choices, not smuggled in as externally validated necessities. (6) TANGERINE is a newly trained model, not a relabeling of a known empirical pattern. A protocol caveat exists: on the seen-distribution CTRate benchmark, the same 46,330 volumes used for SSL pretraining are also used for supervised fine-tuning, which may advantage TANGERINE over comparators that never saw those volumes; however, this is a benchmark-fairness and potential data-leakage concern, not circularity, because the reported test predictions are on the withheld 3,040-scan split, and independent unseen-distribution tasks (NLST-ACRIN, Duke, RadChest, COCA) provide external grounding. Therefore the circularity score is 0.

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

The framework's claims rest on standard self-supervised learning assumptions (SSL transfers to downstream tasks) and on domain choices: 256^3 downsampling preserves subtle disease features, and the comparison protocol is fair. The masking ratio and layer-wise decay are hand-set hyperparameters. No invented entities are introduced.

free parameters (2)
  • masking ratio = 0.75
    Chosen based on prior 2D MAE literature and small-scale preliminary experiments on a subset of data; the authors state a full ablation was deemed computationally prohibitive.
  • layer-wise learning rate decay = 0.75
    Hand-set during fine-tuning to preserve pretrained features; no ablation reported.
assumptions (3)
  • domain assumption Downsampling CT volumes to 256x256x256 and HU clipping to [-1200, 800] preserves clinically relevant diagnostic information.
    The paper acknowledges in Discussion that downsampling may lose fine-grained diagnostic detail; if this assumption fails, the reported performance may not translate to clinical practice.
  • domain assumption Self-supervised masked autoencoding on large unlabelled CT volumes produces representations that transfer to diverse downstream disease classification tasks.
    This is the foundational premise of the method; the paper's results support it empirically on held-out tasks.
  • ad hoc to paper The evaluation protocol, including the use of frozen Gemini embeddings and fine-tuning on CTRate volumes that were seen during pretraining, is a fair basis for the 'state-of-the-art' claim.
    The comparison to Gemini is not end-to-end fine-tuned, and CTRate fine-tuning shares training volumes with pretraining, which may inflate TANGERINE's relative performance.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A computationally frugal open-source foundation model for thoracic disease detection in lung cancer screening programs." pith.science (2026). https://pith.science/paper/BXP66UPW

@misc{pith2026250701881,
  author       = {Pith},
  title        = {Pith review of: A computationally frugal open-source foundation model for thoracic disease detection in lung cancer screening programs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BXP66UPW}},
  note         = {Machine review of arXiv:2507.01881}
}
read the original abstract

Low-dose computed tomography (LDCT) imaging employed in lung cancer screening (LCS) programs is increasing in uptake worldwide. LCS programs herald a generational opportunity to simultaneously detect cancer and non-cancer-related early-stage lung disease. Yet these efforts are hampered by a shortage of radiologists to interpret scans at scale. Here, we present TANGERINE, a computationally frugal, open-source vision foundation model for volumetric LDCT analysis. Designed for broad accessibility and rapid adaptation, TANGERINE can be fine-tuned off the shelf for a wide range of disease-specific tasks with limited computational resources and training data. Relative to models trained from scratch, TANGERINE demonstrates fast convergence during fine-tuning, thereby requiring significantly fewer GPU hours, and displays strong label efficiency, achieving comparable or superior performance with a fraction of fine-tuning data. Pretrained using self-supervised learning on over 98,000 thoracic LDCTs, including the UK's largest LCS initiative to date and 27 public datasets, TANGERINE achieves state-of-the-art performance across 14 disease classification tasks, including lung cancer and multiple respiratory diseases, while generalising robustly across diverse clinical centres. By extending a masked autoencoder framework to 3D imaging, TANGERINE offers a scalable solution for LDCT analysis, departing from recent closed, resource-intensive models by combining architectural simplicity, public availability, and modest computational requirements. Its accessible, open-source lightweight design lays the foundation for rapid integration into next-generation medical imaging tools that could transform LCS initiatives, allowing them to pivot from a singular focus on lung cancer detection to comprehensive respiratory disease management in high-risk populations.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

2 extracted references · 1 canonical work pages

  1. [5]

    Balata, H. et al. Targeted lung cancer screening selects individuals at high risk of cardiovascular disease. Lung Cancer 124, (2018). 6. Hinde, S. et al. The cost-effectiveness of the Manchester ‘lung health checks’, a community-based lung cancer low-dose CT screening pilot. Lung Cancer 126, (2018). 7. Field, J. K. et al. UK Lung Cancer RCT Pilot Screenin...

    work page 2018

  2. [34]

    Pai, S. et al. Foundation model for cancer imaging biomarkers. Nat Mach Intell 6, (2024). 35. Martinez, F. J. et al. Treatment Trials in Young Patients with COPD and Pre-COPD Patients: Time to Move Forward. Am J Respir Crit Care Med 205, (2021). 36. Yue, X. et al. Understanding Masked Autoencoders From a Local Contrastive Perspective. 37. Bolón-Canedo, V....

    arXiv 2024

Pith tools

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