INTERACT-CMIL: Multi-Task Shared Learning and Inter-Task Consistency for Conjunctival Melanocytic Intraepithelial Lesion Grading

Daniela S\"usskind; Julia M. Weller; Karin U. Loeffler; Leticia Ussem; Luna Toma; Martina C. Herwig-Carl; Mert Ikinci; Shadi Albarqouni; Yousef Yeganeh

REVIEW 4 major objections 4 minor 10 references

INTERACT-CMIL: Multi-Task Shared Learning and Inter-Task Consistency for Conjunctival Melanocytic Intraepithelial Lesion Grading

T0 review · 4 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read A shared five-task grading model that enforces the experts' known label co-occurrence lifts F1 by up to 55% over single-task baselines on a new 486-patch dataset.

desk verdict A plausible first benchmark for CMIL grading, but the component attribution and the temperature-scaling component are under-supported; worth a careful referee. read the letter →

arxiv 2512.22666 v1 pith:BNR4LTRX submitted 2025-12-27 cs.CV cs.LG

Mert Ikinci , Luna Toma , Karin U. Loeffler , Leticia Ussem , Daniela Süsskind , Julia M. Weller , Yousef Yeganeh , Martina C. Herwig-Carl

show 1 more author

Shadi Albarqouni

This is my paper · ORCID

classification cs.CVcs.LG

keywords conjunctivalmelanocyticintraepitheliallesionsmulti-tasklearninginter-dependencelossdigitalpathologyfoundationmodelselectivesupervisiongradingcriteriamacroF1

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 tries to establish that CMIL grading is best modeled as one multi-task problem with explicit inter-task consistency, not as five independent classifications. The proposed framework uses a shared feature representation, five task-specific heads, a training scheme that updates only three heads per iteration in rotation, and a regularizer that pushes the model's predicted joint distribution of labels toward the co-occurrence statistics observed in the data. On a newly collected, multi-center dataset of 486 expert-annotated biopsy patches, it improves macro F1 over a frozen pathology foundation-model baseline by double-digit relative margins on four of five axes, with the largest gain on the most subtle criteria. If these results hold, pathologists gain a reproducible, interpretable second reader for a rare but potentially lethal disease, and the field gains a benchmark for digital ocular pathology.

What carries the argument

The central mechanism is Shared Feature Learning with Combinatorial Partial Supervision (SFCS) plus an Inter-Dependence Loss. Five task heads share one encoder-derived representation; each training step updates only three of them, cycling through all three-head combinations. The Inter-Dependence Loss aligns the predicted joint distribution — the batch-average outer product of the active heads' probability vectors — with the empirical joint distribution of the expert one-hot labels, via KL divergence, added to the classification loss. This enforces the clinically known co-occurrence of criteria (e.g., severe atypia with vertical proliferation) while selective supervision stops any one task fr

What would settle it

One decisive check: within each training batch, shuffle the one-hot labels before computing the co-occurrence prior (preserving each task's marginal distribution but destroying genuine inter-task correlations); if the model's macro F1 stays at the reported level or the dependency loss shows no effect, the claim that cross-task dependency structure drives the gains is refuted.

Watch

Extended reading notes

Core claim

The central claim is that enforcing inter-task consistency across five interrelated grading criteria — WHO4, WHO5, horizontal spread, vertical spread, and cytologic atypia — yields substantially better predictions than predicting each criterion alone. The authors show that a multi-head architecture built on a frozen pathology encoder, trained with combinatorial partial supervision (cycling through all three-head subsets) and an inter-dependence loss that aligns the predicted joint distribution with the empirical joint distribution of expert labels, achieves mean macro F1 scores of 0.76, 0.88, 0.71, 0.76, and 0.70 across the five axes, with relative gains up to 55.1% over the strongest single

Load-bearing premise

The load-bearing premise is that batch-level co-occurrence statistics of one-hot labels in this 486-sample, class-imbalanced dataset are a stable, unbiased estimate of true inter-criterion relationships; if they are noisy or skewed by grade prevalence, the dependency loss will enforce spurious correlations and the claimed gains could vanish.

Editorial extensions

If this is right

If the reported gains hold, computational CMIL grading becomes feasible with a reproducible multi-center benchmark that other labs can extend.
The largest gains cluster on vertical spread and cytologic atypia, suggesting that explicit inter-criterion consistency is the key to detecting the most subtle, high-risk features.
The ablation numbers imply a practical recipe: a frozen pathology encoder plus selective multi-head supervision plus a dependency regularizer yields roughly 10% higher macro F1 than the strongest single-task baseline, with each component contributing additively.
The framework offers a template for other multi-criteria histopathology grading tasks where diagnostic axes are known to be correlated, not just for ocular lesions.
Because the model outputs coherent multi-criteria predictions, it could serve as a decision-support tool that flags internally inconsistent gradings for pathologist review.

Reading between the lines

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

The inter-dependence loss is computed from batch-level empirical co-occurrence, so its usefulness likely depends on the training distribution matching the target population; on datasets with different grade prevalence, the enforced dependencies could become misleading rather than helpful.
Selective supervision (three of five heads per step) may also function as implicit task-level ensembling or dropout; comparing it against task-dropout or gradient-blending baselines would clarify its mechanism beyond regularization.
If the co-occurrence prior reflects a property of the disease rather than one center's staining protocol, the same loss could serve as a domain-adaptation anchor when transferring across institutions with different class distributions.
Given the dataset's small size and high-grade predominance, the strongest reported gain (WHO4, +55%) should be validated on a larger external cohort before drawing clinical conclusions; the paper itself does not claim such validation.

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.

Desk Editor's Note

read the letter

Short version: this is a genuinely new application — first deep-learning framework for grading conjunctival melanocytic intraepithelial lesions, on a new multi-center dataset of 486 expert-annotated patches. The overall gain over frozen CHIEF and ResNet baselines is real on this dataset: macro F1 up 55% on WHO4 and 25% on vertical spread, consistent across all five axes, and patient-disjoint cross-validation. That is worth something, especially given the rarity of the disease.

What the paper does well: the problem is well-motivated, the multi-task framing maps directly onto clinical criteria, and the authors are honest about label imbalance and the small dataset. The shared-feature setup with selective supervision is a reasonable way to regularize with few samples. The qualitative examples are helpful.

Soft spots, in order of severity. First, the temperature-scaling component is never defined. Section 2.4 gives L_total = L_cls + λL_dep with no temperature parameter anywhere; the ablation '–temp.' tests something the reader cannot identify. That alone makes the 4% attribution unverifiable. Second, the ablations are cumulative, not leave-one-out. The reported contributions of each component depend on the removal order; the paper claims 'additive gains' but only shows one path. Third, the dependence loss uses the empirical co-occurrence of the training labels as the target for consistency. That's a summary of the labels, not an independent constraint; on 486 imbalanced samples, batch-level co-occurrence could encode sampling bias. The gains on held-out folds suggest it helps here, but there's no sensitivity analysis to batch size or class imbalance. Fourth, no code or data release, no significance tests for the relative gains, and the 'first DL framework' claim would need a more careful literature check (there may be prior work on PAM or C-MIN classification). These are addressable rather than fatal.

The stress-test note got the main issue right: the three-component attribution is not supported as stated because of the cumulative ablation design, and temperature scaling is undefined. The central claim of overall improvement over baselines is better supported.

Who is this for? Researchers working on computational pathology for rare ocular lesions, and anyone thinking about multi-task regularization in low-data regimes. It deserves a serious referee — the dataset and first benchmark are useful — but the authors should be asked to define temperature scaling, run proper leave-one-out ablations, and release code/data. I'd take it for review, but expect revision.

Referee Report

4 major / 4 minor

Summary. The paper presents INTERACT-CMIL, a multi-head classifier for five CMIL grading axes (WHO4, WHO5, horizontal spread, vertical spread, and cytologic atypia) built on frozen CHIEF embeddings and a shared 256-D projection. The training objective combines combinatorial partial supervision (three of five heads active per iteration), a KL-based inter-dependence loss that matches the predicted joint distribution over the active heads to the empirical joint distribution of the batch's one-hot labels, and, according to the ablation study, a temperature-scaling step that is not actually defined in the Methods. The model is evaluated with 5-fold cross-validation on a newly collected 486-patch dataset from three German university hospitals. Table 1 reports mean macro F1 of 0.76/0.88/0.71/0.76/0.70 for the five axes, corresponding to relative gains over a frozen-CHIEF single-task baseline of 55.1/24.7/15.2/25.0/12.5%. The paper claims that these gains arise additively from the three components.

Significance. If the reported results are reproducible, the contribution is a clinically aligned, multi-criteria benchmark for a rare ocular lesion and a sensible use of pathology foundation-model features. Strengths include patient-disjoint splits, multi-center annotation with expert adjudication, and head-to-head comparison against both CNN and foundation-model baselines. The full-model improvements over BaseCHIEF are large and consistent across all five axes, and the qualitative examples illustrate plausible predictions. The main significance is as a proof-of-concept for dependency-aware multi-task grading in a data-scarce pathology setting. The work is much less persuasive as a methodological contribution until the undefined temperature-scaling component is specified and the ablation evidence is made non-cumulative.

major comments (4)

[Sec. 2.4, Sec. 3.3.3] Temperature scaling is listed as one of the three key components and is removed in the ablation ('–dep.–temp.'), but it is never defined. Sec. 2.4 defines the objective as L_total = L_cls + λ L_dep and contains no temperature parameter; no equation, default value, or training-time description appears anywhere in the paper. Therefore the ~4% contribution attributed to 'temperature scaling' is not a reproducible finding. Please specify how temperature is applied (e.g., logit scaling before softmax, with initialization/scheduling) and report the value(s) used. If this component was inherited from a previous version, it needs to be fully integrated into the Methods and Figure 2.
[Sec. 3.3, Table 1] The ablation removes components in one cumulative order (full → –dep. → –dep.–temp. → –dep.–temp.–sel.), so the individual effect of each component is confounded with removal order. For example, the temperature-scaling effect is estimated by comparing 0.7397 (WHO4, –dep.) with 0.6924 (–dep.–temp.), after the dependency loss is already removed; the selective-supervision effect is estimated by comparing 0.6924 with 0.6120. These differences do not identify the contribution of each component in the full model and do not support the statement that 'each component provides additive gains across all five diagnostic tasks' (Sec. 3.3). Please run leave-one-out ablations (full model minus one component) and, if interactions are expected, a small factorial design.
[Table 1, Sec. 3.2] All performance claims rest on 5-fold cross-validation on 486 patches, but the paper reports no significance tests, confidence intervals for the differences, or paired comparisons. With only five folds, the mean/median and standard deviation in Table 1 are insufficient to establish that the gains over BaseCHIEF are statistically reliable, especially for horizontal spread (+0.0932 macro F1) where the reported standard deviations overlap. Please report per-fold paired differences with a Wilcoxon signed-rank test or bootstrap CIs, and state how many patients are involved in addition to the 486 patches, confirming that the patient-disjoint split is strictly enforced.
[Sec. 2.3, Eq. (2)] The dependence loss uses as its target P, the empirical joint distribution of the one-hot labels within each training batch. This is not an independent clinical constraint—it summarizes the training labels themselves, so the loss enforces internal consistency with dataset statistics rather than an external pathological prior. Given the small, class-imbalanced dataset, batch-level co-occurrence estimates may be noisy (e.g., the predominance of high-grade lesions). The paper does not analyze sensitivity to batch size or class imbalance, nor does it justify the KL direction used here. Please add such an analysis and clarify the clinical source of the target distribution, or temper the claim that the loss aligns predictions with 'established WHO and C-MIN standards.'

minor comments (4)

[Title/Abstract] There are two obvious typos: 'Conjunctiv al' in the title and 'Conjunctiv al' in the abstract. Also, the abstract says 'WHO4, WHO5, horizontal spread, vertical spread, and cytologic atypia' but the figure uses 'AP' for atypia; please unify the abbreviations and define them at first use.
[Fig. 3] ROC curves are shown without numeric AUC values either in the figure or the text. Since AUC is listed as a secondary metric, report the AUC values in Table 1 or in the figure caption.
[Sec. 3.1.1] The dataset description 'WHO5=2, WHO4=3' is unclear; it probably refers to the most frequent grade, but the sentence is ambiguous. Please specify the class distributions for all five tasks, including the number of classes per task and per-class counts.
[General] The paper calls the method 'reproducible' but provides no code, data, or model-checkpoint availability statement. Given the undefined temperature-scaling component and the lack of data release, the reproducibility claim is currently not supported.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the central performance claims are supported by held-out evaluation and the method's components are defined as supervised training objectives rather than predictions derived from the same labels.

full rationale

The paper's central claims are empirical: on patient-disjoint held-out folds, INTERACT-CMIL reports higher macro F1 than CNN and frozen-foundation-model baselines. The method is a standard multi-head classifier with a shared frozen CHIEF encoder, selective supervision over three randomly/cyclically active heads, and a KL-divergence regularizer that matches the predicted joint task distribution to the empirical co-occurrence of training labels. The dependence loss target P is computed from training labels, but it is used only as a supervised regularizer; the evaluation metric is per-task macro F1 on held-out data, so the reported gains are not forced by construction. No prediction is claimed that is identical to a fitted input. The 'inter-dependence' is a summary of training-label co-occurrence, but the paper does not claim to discover an external law from it; it claims improved classification, which is tested. The ablation attribution is weakened by cumulative removal and the temperature-scaling component is never defined in Sec. 2.4 (L_total = L_cls + λL_dep), which is a reproducibility gap rather than a circularity. There are no self-citations; CHIEF is an external pretrained model by other authors. Therefore no circularity is present.

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

The central claim relies on CHIEF's transferability, the reliability of batch co-occurrence as a consistency target, and label correctness. No new physical entities are introduced. The main free parameters are the dependency loss weight and the temperature scaling, neither of which is documented.

free parameters (2)

λ (dependency loss weight) = not reported
Controls the weight of the inter-dependence loss; no value or tuning procedure given.
temperature scaling parameter(s) = not reported
Mentioned in ablations as a performance component but not defined in Methods.

assumptions (3)

domain assumption CHIEF foundation model embeddings capture transferable histomorphologic features relevant to CMIL
The method freezes CHIEF features; if these features don't generalize to conjunctival patches, the reported performance would not hold.
domain assumption The empirical joint label distribution in a batch is a meaningful target for inter-task consistency
The dependence loss (Sec. 2.3) relies on batch-level co-occurrence statistics as a stable and informative regularizer.
domain assumption Expert-consensus labels are correct and consistent across centers
The evaluation treats labels as ground truth; inter-observer variability is mentioned as motivation but not quantified.

how reviews work

0 comments

Cite this review

Pith. "Pith review of INTERACT-CMIL: Multi-Task Shared Learning and Inter-Task Consistency for Conjunctival Melanocytic Intraepithelial Lesion Grading." pith.science (2026). https://pith.science/paper/BNR4LTRX

@misc{pith2026251222666,
  author       = {Pith},
  title        = {Pith review of: INTERACT-CMIL: Multi-Task Shared Learning and Inter-Task Consistency for Conjunctival Melanocytic Intraepithelial Lesion Grading},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BNR4LTRX}},
  note         = {Machine review of arXiv:2512.22666}
}

read the original abstract

Accurate grading of Conjunctival Melanocytic Intraepithelial Lesions (CMIL) is essential for treatment and melanoma prediction but remains difficult due to subtle morphological cues and interrelated diagnostic criteria. We introduce INTERACT-CMIL, a multi-head deep learning framework that jointly predicts five histopathological axes; WHO4, WHO5, horizontal spread, vertical spread, and cytologic atypia, through Shared Feature Learning with Combinatorial Partial Supervision and an Inter-Dependence Loss enforcing cross-task consistency. Trained and evaluated on a newly curated, multi-center dataset of 486 expert-annotated conjunctival biopsy patches from three university hospitals, INTERACT-CMIL achieves consistent improvements over CNN and foundation-model (FM) baselines, with relative macro F1 gains up to 55.1% (WHO4) and 25.0% (vertical spread). The framework provides coherent, interpretable multi-criteria predictions aligned with expert grading, offering a reproducible computational benchmark for CMIL diagnosis and a step toward standardized digital ocular pathology.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

10 extracted references

[1]

Conjunctival melanocytic lesions,

Scott C Bresler, Caroline Simon, Carol L Shields, Jonathan B McHugh, Anna M Stagner, and Rajiv M Patel, “Conjunctival melanocytic lesions,”Archives of Pathology & Laboratory Medicine, vol. 146, no. 5, pp. 632–646, 2022

2022
[2]

Tatyana Milman, Maya Eiger-Moscovich, Roger K Henry, Robert Folberg, Sarah E Coupland, Hans E Grossniklaus, Hardeep Singh Mudhar, Charles G Eberhart, Steffen Heegaard, Claudia Auw-H ¨adrich, et al., “Validation of the newly proposed world health organization classification system for conjunc- tival melanocytic intraepithelial lesions: a comparison with th...

2021
[3]

A multicenter study validates the who 2022 classification for conjuncti- val melanocytic intraepithelial lesions with clinical and prognostic relevance,

Hardeep Singh Mudhar, Yamini Krishna, Simon Cross, Claudia Auw-Haedrich, Raymond Barnhill, Svetlana Cherepanoff, Ralph Eagle, James Farmer, Robert Fol- berg, Hans Grossniklaus, et al., “A multicenter study validates the who 2022 classification for conjuncti- val melanocytic intraepithelial lesions with clinical and prognostic relevance,”Laboratory Investi...

2022
[4]

Management of conjunctival melanoma,

Bertil Damato and Sarah E Coupland, “Management of conjunctival melanoma,”Expert Review of Anticancer Therapy, vol. 9, no. 9, pp. 1227–1239, 2009

2009
[5]

Grossniklaus, C.G

H.E. Grossniklaus, C.G. Eberhart, T.T. Kivel ¨a, and In- ternational Agency for Research on Cancer,WHO Clas- sification of Tumours of the Eye: WHO Classification of Tumours, Volume 12, World Health Organization classi- fication of tumours. World Health Organization, 2018

2018
[6]

The 5th edition of the world health organization classifica- tion of tumours of the eye and orbit,

Tatyana Milman, Hans E Grossniklaus, Gabrielle Goldman-Levy, Tero T Kivel¨a, Sarah E Coupland, Va- lerie A White, Hardeep Singh Mudhar, Charles G Eber- hart, Robert M Verdijk, Steffen Heegaard, et al., “The 5th edition of the world health organization classifica- tion of tumours of the eye and orbit,”Ocular oncology and pathology, vol. 9, no. 3-4, pp. 71–95, 2023

2023
[7]

Deep learning in histopathology: the path to the clinic,

Jeroen Van der Laak, Geert Litjens, and Francesco Ciompi, “Deep learning in histopathology: the path to the clinic,”Nature medicine, vol. 27, no. 5, pp. 775–784, 2021

2021
[8]

Clinical-grade computational pathology using weakly supervised deep learning on whole slide images,

Gabriele Campanella, Matthew G Hanna, Luke Genes- law, Allen Miraflor, Vitor Werneck Krauss Silva, Klaus J Busam, Edi Brogi, Victor E Reuter, David S Klimstra, and Thomas J Fuchs, “Clinical-grade computational pathology using weakly supervised deep learning on whole slide images,”Nature medicine, vol. 25, no. 8, pp. 1301–1309, 2019

2019

Show all 10 references

[9]

Towards a general-purpose foundation model for computational pathology,

Richard J Chen, Tong Ding, Ming Y Lu, Drew FK Williamson, Guillaume Jaume, Andrew H Song, Bowen Chen, Andrew Zhang, Daniel Shao, Muhammad Sha- ban, et al., “Towards a general-purpose foundation model for computational pathology,”Nature Medicine, vol. 30, no. 3, pp. 850–862, 2024

2024
[10]

A pathology foundation model for cancer diagnosis and prognosis prediction,

Xiyue Wang, Junhan Zhao, Eliana Marostica, Wei Yuan, Jietian Jin, Jiayu Zhang, Ruijiang Li, Hongping Tang, Kanran Wang, Yu Li, et al., “A pathology foundation model for cancer diagnosis and prognosis prediction,” Nature, vol. 634, no. 8035, pp. 970–978, 2024

2024

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.

[1] [1]

Conjunctival melanocytic lesions,

Scott C Bresler, Caroline Simon, Carol L Shields, Jonathan B McHugh, Anna M Stagner, and Rajiv M Patel, “Conjunctival melanocytic lesions,”Archives of Pathology & Laboratory Medicine, vol. 146, no. 5, pp. 632–646, 2022

2022

[2] [2]

Tatyana Milman, Maya Eiger-Moscovich, Roger K Henry, Robert Folberg, Sarah E Coupland, Hans E Grossniklaus, Hardeep Singh Mudhar, Charles G Eberhart, Steffen Heegaard, Claudia Auw-H ¨adrich, et al., “Validation of the newly proposed world health organization classification system for conjunc- tival melanocytic intraepithelial lesions: a comparison with th...

2021

[3] [3]

A multicenter study validates the who 2022 classification for conjuncti- val melanocytic intraepithelial lesions with clinical and prognostic relevance,

Hardeep Singh Mudhar, Yamini Krishna, Simon Cross, Claudia Auw-Haedrich, Raymond Barnhill, Svetlana Cherepanoff, Ralph Eagle, James Farmer, Robert Fol- berg, Hans Grossniklaus, et al., “A multicenter study validates the who 2022 classification for conjuncti- val melanocytic intraepithelial lesions with clinical and prognostic relevance,”Laboratory Investi...

2022

[4] [4]

Management of conjunctival melanoma,

Bertil Damato and Sarah E Coupland, “Management of conjunctival melanoma,”Expert Review of Anticancer Therapy, vol. 9, no. 9, pp. 1227–1239, 2009

2009

[5] [5]

Grossniklaus, C.G

H.E. Grossniklaus, C.G. Eberhart, T.T. Kivel ¨a, and In- ternational Agency for Research on Cancer,WHO Clas- sification of Tumours of the Eye: WHO Classification of Tumours, Volume 12, World Health Organization classi- fication of tumours. World Health Organization, 2018

2018

[6] [6]

The 5th edition of the world health organization classifica- tion of tumours of the eye and orbit,

Tatyana Milman, Hans E Grossniklaus, Gabrielle Goldman-Levy, Tero T Kivel¨a, Sarah E Coupland, Va- lerie A White, Hardeep Singh Mudhar, Charles G Eber- hart, Robert M Verdijk, Steffen Heegaard, et al., “The 5th edition of the world health organization classifica- tion of tumours of the eye and orbit,”Ocular oncology and pathology, vol. 9, no. 3-4, pp. 71–95, 2023

2023

[7] [7]

Deep learning in histopathology: the path to the clinic,

Jeroen Van der Laak, Geert Litjens, and Francesco Ciompi, “Deep learning in histopathology: the path to the clinic,”Nature medicine, vol. 27, no. 5, pp. 775–784, 2021

2021

[8] [8]

Clinical-grade computational pathology using weakly supervised deep learning on whole slide images,

Gabriele Campanella, Matthew G Hanna, Luke Genes- law, Allen Miraflor, Vitor Werneck Krauss Silva, Klaus J Busam, Edi Brogi, Victor E Reuter, David S Klimstra, and Thomas J Fuchs, “Clinical-grade computational pathology using weakly supervised deep learning on whole slide images,”Nature medicine, vol. 25, no. 8, pp. 1301–1309, 2019

2019