Pith. sign in

REVIEW 3 major objections 6 minor 45 references

Label-free Concept Based Multiple Instance Learning for Gigapixel Histopathology

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read This paper proposes Concept MIL, a whole-slide image classifier that makes each prediction a weighted sum of human-understandable pathology concepts, removing the need for manual concept annotations.

desk verdict Useful first step: label-free concept bottleneck for WSI, but the 'inherently interpretable' claim is softer than the paper makes it sound. read the letter →

arxiv 2501.02922 v1 pith:B5URHADK submitted 2025-01-06 cs.CV cs.AI

Susu Sun , Leslie Tessier , Frédérique Meeuwsen , Clément Grisi , Dominique van Midden , Geert Litjens , Christian F. Baumgartner This is my paper · ORCID
classification cs.CVcs.AI
keywords multipleinstancelearningconceptbottleneckmodelswhole-slideimageshistopathologyinherentinterpretabilityvision-languagefoundationlabel-freeconceptsweaklysupervisedclassification
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 proposes Concept MIL, a whole-slide image classifier whose predictions can be read off as a sum of human-understandable pathology concepts. It claims to be the first concept-based, inherently interpretable model for gigapixel histopathology: instead of only highlighting a heatmap, it tells you that the slide is called tumor because of, say, 'dense nuclei' and 'disorganized cells' in the top 20 patches, and how much each concept contributed. The model avoids manual concept annotations by using a vision-language model trained on pathology image-text pairs to score each patch against a concept list, and the final prediction is a linear combination of concept scores, so contributions are exact by construction. On Camelyon16 and PANDA it reports accuracy and AUC above 0.9, on par with standard MIL models, and a pathologist user study indicates the concepts match clinical reasoning. If correct, it means an interpretable gigapixel classifier can be built with slide-level labels only, and every decision can be audited concept by concept.

What carries the argument

The load-bearing object is the linear concept decomposition. Concept activation vectors are built by cosine similarity between CONCH image patch embeddings and text embeddings of prompts like 'an H&E image of CONCEPT'; the concept MIL branch then applies a sparse, percentile-scaled attention vector $\beta$ over concepts and a logistic classifier over the sum $\sum_j \sum_c w_c f_{jc}\beta_c$ across the top-$K$ patches chosen by a differentiable Patch Attention-Guided Top-K module. This linear pre-activation is what makes the prediction an exact sum of per-concept contributions $\kappa_c$, so the explanations are not post-hoc approximations. The architecture also includes a jointly trained attention-based image MIL branch whose only job at inference is to supply the top-$K$ patch indices; the image branch's own prediction is discarded.

What would settle it

On a held-out set of WSIs with pixel-level annotations for individual concepts (for example, pathologist-outlined regions of 'dense nuclei' or 'cribriform pattern'), compute the correlation between each patch's concept activation score and whether a human labeled that concept as present; if several concepts show near-zero or negative correlation, the concept detection is not grounded in image content and the explanation mechanism fails. A complementary check is to re-run the trained model on a test slide with the largest single concept contribution removed from the linear sum, which should shift the logit by exactly the reported $\kappa_c$; if it does not, the linear decomposition does not reflect the actual decision process.

Watch

Extended reading notes

Core claim

The central claim is that Concept MIL is the first inherently interpretable WSI classifier whose decisions are expressed in pathology concepts rather than spatial heatmaps. The model operates by embedding all WSI patches with the CONCH image encoder, computing concept activation vectors as cosine similarities between patch embeddings and text embeddings of a predefined concept set, selecting the top 20 patches by attention from a jointly trained image MIL branch, and making the final prediction by a logistic function of a weighted sum over concepts and selected patches. Because the pre-activation is a linear combination, the contribution of each concept to the prediction is exactly $\kappa_c = w_c \sum_{j=1}^{K} f_{jc}\beta_c$, giving an inherent local explanation; dataset-level averages and t-SNE projections give global explanations. The authors report that this scheme reaches AUC and accuracy over 0.9 on Camelyon16 and PANDA, places 87.1% and 85.3% of the top-20 patches inside tumor regions, and produces top concepts that overlap with pathologists' choices in a three-pathologist user study.

Load-bearing premise

The method assumes that CONCH's shared image-text embedding space is aligned well enough that a patch's cosine similarity to a concept text reliably detects that concept; the paper itself shows this fails for 'fibrous tissue' and requires a manual pathologist filtering step, so if that alignment is unreliable for a chosen concept set, the concept activations and their explanations lose meaning.

Editorial extensions

If this is right

  • A clinician can trace each tumor or normal call to specific pathology concepts on specific top-K patches rather than to an opaque heatmap.
  • A WSI classifier with concept-level explanations can be trained using only slide-level labels, with no pixel or concept annotations.
  • Adapting the model to a new disease reduces to defining a new concept list and text prompts, as long as the vision-language embeddings are reliable for those concepts.
  • Dataset-level concept contribution plots allow developers to audit whether the model's decisions rest on clinically sensible features before deployment.
  • Because the prediction is a linear function of concept scores, removing a concept's contribution changes the logit by exactly its stated contribution, making the explanations faithful by construction.

Reading between the lines

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

  • The 'label-free' claim applies to training supervision, not the whole pipeline: the concept list is drafted with a general-purpose language model and then refined and verified by a pathologist, so expert effort moves from labeling concepts to curating them.
  • The linear logit makes counterfactual explanations easy to build — asking how the prediction would change if a concept were absent — although the paper does not develop this explicitly.
  • The method's interpretability is only as trustworthy as the embedding alignment of the vision-language model; for other stains, tissues, or concept sets, the manual filtering step the authors used would likely need to be repeated.
  • The binary design suggests a direct extension to Gleason grading or multi-class tasks by replacing the logistic head with a softmax over the same per-concept contributions, and multi-scale concepts could be handled by pooling concept scores across magnifications.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes Concept MIL, a dual-branch multiple-instance-learning model for whole-slide image classification in histopathology. An image MIL branch selects the top K salient patches via attention, and a concept MIL branch classifies the slide using concept activation vectors computed as cosine similarities between CONCH image embeddings and text embeddings of pathologist-refined concepts. The authors claim this is the first concept-based, inherently interpretable WSI model, with predictions obtained as a linear combination of human-understandable concepts, enabling faithful local and global explanations. Experiments on Camelyon16 and PANDA report AUC/accuracy above 0.9, localization scores around 0.85-0.87, and a small user study with three pathologists showing moderate agreement with the model's top concepts.

Significance. If the interpretability claim holds, the paper would be a valuable contribution to computational pathology, offering concept-level explanations without manual concept annotation and with classification performance on par with state-of-the-art MIL models. The work is timely given the rise of pathology foundation models and the need for trustworthy AI in medicine. The authors promise to release code and data, which would facilitate reproducibility. However, the central claim of inherent interpretability is weakened by the non-linear attention gating in the concept branch and by the limited validation of the underlying vision-language embedding alignment.

major comments (3)
  1. [Section III-D, Eqs. (5)-(8)] The prediction is not obtained through a purely linear combination of concepts, because beta_c in Eq. (5) is the output of a non-linear attention network Ac(·) operating on the entire transposed concept feature matrix of the top-K patches (Eqs. 3-4). Consequently, the contribution kappa_c in Eq. (7) is not an independent, counterfactual measure of concept c's influence: beta_c itself depends on all concepts and all selected patches, so changing one concept's activation can alter the apparent contribution of other concepts through the shared attention mechanism. The claim that the model enables 'tracing each concept's influence on the prediction' is therefore not fully justified. The authors should either modify the architecture to obtain a truly linear decomposition (e.g., removing the attention gating) or substantially temper the interpretability claim and discuss the faithfulness limitations introduced by the non-linear gating, in the same way the paper criticizes attention maps in ABMIL.
  2. [Section III-B, Fig. 2, Table I] The reliability of concept activations rests on the alignment between CONCH's image and text embedding spaces, and the paper itself demonstrates a misalignment: fat-cell patches receive high cosine similarity with 'fibrous tissue' (Fig. 2). The proposed filtering step—asking a pathologist to verify five high-activation patches per concept from 50K tumor-region patches in the training set—is too weak to establish semantic reliability across the test distribution. It samples only tumor patches, not normal tissue or test WSIs; it uses only five patches per concept, which is a very small sample for heterogeneous gigapixel slides; and it relies on a single pathologist's binary decision with no quantitative precision/recall or inter-observer agreement. Since all downstream explanations are built on these activations, the authors should provide stronger evidence that the refined concept set is reliably detected across the evaluated datasets, or explicitly scope the interpretability claim to the verified distribution.
  3. [Section IV-E, Table II] The concept MIL branch alone achieves only 0.688 AUC on Camelyon16, compared to 0.970 for the image MIL branch, indicating that the concept features carry limited discriminative signal on this dataset. The final model's good performance therefore depends heavily on the image branch's top-K patch selection, which is not concept-based. This raises a conceptual question: is the concept branch truly the decision-maker, or is it largely a post-hoc explainer of the image branch's selections? The paper should clarify the causal role of the two branches, for example by reporting how performance and explanation quality vary with K, or by ablating the image branch's influence at inference time. The current framing as an 'inherently interpretable' model is not fully supported if the core discriminative power comes from a branch whose selection mechanism is not concept-based.
minor comments (6)
  1. [Section IV-F2] The user study is small (10 samples per dataset, three pathologists) and the agreement scores range from 0.25 to 0.55; the claim that the model's concepts 'align with the concepts used by pathologists' should be reported with confidence intervals or a qualitative disclaimer, given the small sample.
  2. [Table III] The disease localization score for TransMIL is missing; the text explains why, but a placeholder (e.g., '—') in the table with a footnote would improve readability.
  3. [Section III-B] The concept filtering step would be more reproducible if the authors reported how many initial GPT4o concepts were proposed and how many were removed by the pathologist, along with the criteria used for removal.
  4. [Eq. (4)] The standard deviation in Eq. (4) is not clearly defined; please specify whether it is computed over the concept dimension, the patch dimension, or the full matrix, as this affects the scaling behavior.
  5. [Section IV-B] The final model is taken from the last epoch after 300 epochs without mention of early stopping or validation-based model selection; please justify this choice or describe the validation procedure used for hyperparameter tuning.
  6. [Abstract and Section III-B] The term 'label-free' is somewhat misleading because the method still requires a pathologist to refine the concept set and to verify high-activation patches; the authors should clarify this nuance in the abstract or introduction.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Concept MIL's classification and localization claims are validated on external benchmarks, and its explanation is an intended algebraic decomposition rather than a fitted input renamed as a prediction.

full rationale

Concept MIL does not exhibit derivation circularity. The final prediction in Eq. (5) is defined as a linear classifier over concept activation vectors, and the explanation in Eqs. (7)-(8) is the same expression regrouped into per-concept contributions; this is the model's construction, not a hidden equivalence between an input and an output. The paper's central quantitative claims are checked against held-out Camelyon16 and PANDA test data and against external baselines (Tables II and III), so the reported accuracy and localization scores do not reduce to the concept set or to any fitted parameter. The concept activations are cosine similarities in CONCH's shared embedding space, and the paper explicitly documents an alignment failure (Fig. 2) and filters concepts with a pathologist (Sec. III-B). That is a validity assumption about the foundation model, not a circular step: the classification labels and tumor masks are external to the concept projection. The user study measures agreement between model-chosen and pathologist-chosen concepts; even if the same pathologist had curated the concept list, this would affect the strength of the interpretability evaluation, not the derivation of the classification result. The only self-citation, Ref. [38], is used as an example of prior inherently interpretable medical models and is not load-bearing. No equation in the paper reduces to its own input except the intended identity between the linear prediction and its additive concept decomposition.

Assumptions & free parameters 8 free parameters · 4 assumptions · 0 invented entities

No new physical entities are introduced. The main load-bearing inputs are the pretrained CONCH embedding space, the human-curated concept list, and the assumption that cosine similarity equates to concept presence. Free parameters are mostly trained weights plus a few hyperparameters.

free parameters (8)
  • w_c and b_concept = learned
    Linear classifier weights and bias in the concept MIL branch (Eq. 5), trained on slide-level labels.
  • w'_d and b_img = learned
    Linear classifier weights and bias in the image MIL branch (Eq. 2), trained on slide-level labels.
  • Attention network parameters (image and concept branches) = learned
    Parameters of gated attention modules Ap and Ac (Eqs. 1 and 3) trained end-to-end.
  • K (number of selected patches) = 20
    Hyperparameter for both datasets; chosen by the authors.
  • gamma (attention percentile) = 0.75
    Percentile used to sparsify concept attention scores (Eq. 4).
  • t (attention temperature) = 3
    Sigmoid temperature for concept attention scores (Eq. 4).
  • lambda (L2 regularization weight) = 0.05
    Weight of L2 regularization on image-branch attention in the loss (Eq. 6).
  • Learning rates and weight decay = 0.001/0.0001 (lr), 0.001 (wd)
    Training hyperparameters: learning rate 0.001 for Camelyon16 and 0.0001 for PANDA, weight decay 0.001 for both.
assumptions (4)
  • domain assumption Cosine similarity in CONCH's shared embedding space is a valid measure of concept presence in a patch.
    The method computes concept activation vectors as cosine similarity between CONCH image and text embeddings (Section III.B). CONCH is a pretrained model; the paper itself reports misalignments (e.g., fat cells vs fibrous tissue), so this assumption is imperfect.
  • domain assumption The predefined concept set is exhaustive enough and each concept is visible at the chosen patch scale (256x256 at 0.5 um/pixel).
    Concepts are generated by GPT-4o and refined by a pathologist with reference patches to ensure visibility at this scale (Section III.B). The model's explanations are limited by this fixed concept vocabulary.
  • domain assumption Ground-truth tumor masks in Camelyon16 and PANDA are accurate enough for localization evaluation and for selecting tumor patches during concept filtering.
    Masks are used to compute the disease localization score (Section IV.D) and to sample 50K tumor patches for concept verification (Section III.B). Any mask noise directly affects these numbers.
  • standard math The differentiable PAG Top-K operator is a faithful approximation of hard Top-K selection.
    The paper adopts the perturbed maximum method from prior work [7, 39] to make top-K selection differentiable (Section III.C). This is a standard technique.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Label-free Concept Based Multiple Instance Learning for Gigapixel Histopathology." pith.science (2026). https://pith.science/paper/B5URHADK

@misc{pith2026250102922,
  author       = {Pith},
  title        = {Pith review of: Label-free Concept Based Multiple Instance Learning for Gigapixel Histopathology},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B5URHADK}},
  note         = {Machine review of arXiv:2501.02922}
}
read the original abstract

Multiple Instance Learning (MIL) methods allow for gigapixel Whole-Slide Image (WSI) analysis with only slide-level annotations. Interpretability is crucial for safely deploying such algorithms in high-stakes medical domains. Traditional MIL methods offer explanations by highlighting salient regions. However, such spatial heatmaps provide limited insights for end users. To address this, we propose a novel inherently interpretable WSI-classification approach that uses human-understandable pathology concepts to generate explanations. Our proposed Concept MIL model leverages recent advances in vision-language models to directly predict pathology concepts based on image features. The model's predictions are obtained through a linear combination of the concepts identified on the top-K patches of a WSI, enabling inherent explanations by tracing each concept's influence on the prediction. In contrast to traditional concept-based interpretable models, our approach eliminates the need for costly human annotations by leveraging the vision-language model. We validate our method on two widely used pathology datasets: Camelyon16 and PANDA. On both datasets, Concept MIL achieves AUC and accuracy scores over 0.9, putting it on par with state-of-the-art models. We further find that 87.1\% (Camelyon16) and 85.3\% (PANDA) of the top 20 patches fall within the tumor region. A user study shows that the concepts identified by our model align with the concepts used by pathologists, making it a promising strategy for human-interpretable WSI classification.

Figures

Figures reproduced from arXiv: 2501.02922 by the authors.

Figure 1
Figure 1. Training and obtaining prediction and explanation with Concept MIL. During training, we extract WSI image features as shown in (a) and project these image features to the predefined concept space to generate concept features as shown in (b), then jointly train the image MIL branch and concept MIL branch through a patch selection module as shown in (c). During inference, the concept MIL branch generates the final pre… view at source ↗
Figure 2
Figure 2. Patches containing fat cells get high cosine similarity scores with the concept of “fibrous tissue”, indicating a potential misalignment between the image and text spaces in CONCH [26]. image features in are then mapped to the concept space by calculating the cosine similarity between the normalized im￾age embeddings and text embeddings, producing the concept activation vectors fn for the N patches. In our prelimina… view at source ↗
Figure 3
Figure 3. Local explanations for predictions of tumor and normal cases from Camelyon16 dataset. A local explanation for a WSI prediction includes four components: (a) an attention map, (b) the top 20 patches selected based on attention scores, (c) concept features for each selected patch, and (d) the whole slide level concept contribution vector. IV. EXPERIMENTS AND RESULTS A. Datasets We present results on two widely-used co… view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: (a) Attention map from Concept MIL model on a Camelyon16 sample, with the top 20 patches highlighted by green dots. (b) Ground truth mask for the sample, showing tumor regions in yellow and normal tissue in dark green. Accuracy and the Area Under the ROC Curve (AUC) as…
Figure 4
Figure 4. Figure 4: Global explanations for the model trained on PANDA dataset. (a) and (b) are WSI-level mean concept contribution vectors for tumor and normal predictions. (c) and (d) show t-SNE plots of concept vectors from normal and tumor cases at the patch level and WSI level. (e) s…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 33 canonical work pages

  1. [1]

    Sanity checks for saliency maps

    Julius Adebayo, Justin Gilmer, Michael Muelly, Ian Goodfellow, Moritz Hardt, and Been Kim. Sanity checks for saliency maps. Advances in neural information processing systems , 31, 2018

    work page 2018

  2. [2]

    Diagnostic assessment of deep learning algorithms for detec- tion of lymph node metastases in women with breast cancer

    Babak Ehteshami Bejnordi, Mitko Veta, Paul Johannes Van Diest, Bram Van Ginneken, Nico Karssemeijer, Geert Litjens, Jeroen AWM Van Der Laak, Meyke Hermsen, Quirine F Manson, Maschenka Balkenhol, et al. Diagnostic assessment of deep learning algorithms for detec- tion of lymph node metastases in women with breast cancer. Jama, 318(22):2199–2210, 2017

    work page 2017

  3. [3]

    B-cos networks: Align- ment is all we need for interpretability

    Moritz B ¨ohle, Mario Fritz, and Bernt Schiele. B-cos networks: Align- ment is all we need for interpretability. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10329– 10338, 2022

    work page 2022

  4. [4]

    Steiner, Hester van Boven, Robert Vink, Christina Hulsbergen van de Kaa, Jeroen van der Laak, Mahul B

    Wouter Bulten, Kimmo Kartasalo, Po-Hsuan Cameron Chen, Peter Str¨om, Hans Pinckaers, Kunal Nagpal, Yuannan Cai, David F. Steiner, Hester van Boven, Robert Vink, Christina Hulsbergen van de Kaa, Jeroen van der Laak, Mahul B. Amin, Andrew J. Evans, Theodorus van der Kwast, Robert Allan, Peter A. Humphrey, Henrik Gr ¨onberg, Hemamali Samaratunga, Brett Delah...

    work page 2022

  5. [5]

    Clinical-grade com- putational pathology using weakly supervised deep learning on whole slide images

    Gabriele Campanella, Matthew G Hanna, Luke Geneslaw, Allen Mi- raflor, Vitor Werneck Krauss Silva, Klaus J Busam, Edi Brogi, Victor E Reuter, David S Klimstra, and Thomas J Fuchs. Clinical-grade com- putational pathology using weakly supervised deep learning on whole slide images. Nature medicine, 25(8):1301–1309, 2019

    work page 2019

  6. [6]

    This looks like that: deep learning for interpretable image recognition

    Chaofan Chen, Oscar Li, Daniel Tao, Alina Barnett, Cynthia Rudin, and Jonathan K Su. This looks like that: deep learning for interpretable image recognition. Advances in neural information processing systems , 32, 2019. 10

    work page 2019

  7. [7]

    Differ- entiable patch selection for image recognition

    Jean-Baptiste Cordonnier, Aravindh Mahendran, Alexey Dosovitskiy, Dirk Weissenborn, Jakob Uszkoreit, and Thomas Unterthiner. Differ- entiable patch selection for image recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 2351–2360, 2021

    work page 2021

  8. [8]

    Clinically applicable deep learning for diagnosis and referral in retinal disease

    Jeffrey De Fauw, Joseph R Ledsam, Bernardino Romera-Paredes, Stanislav Nikolov, Nenad Tomasev, Sam Blackwell, Harry Askham, Xavier Glorot, Brendan O’Donoghue, Daniel Visentin, et al. Clinically applicable deep learning for diagnosis and referral in retinal disease. Nature medicine, 24(9):1342–1350, 2018

    work page 2018

Show all 45 references
  1. [9]

    Inherently interpretable position-aware convolutional motif kernel networks for biological se- quencing data

    Jonas C Ditz, Bernhard Reuter, and Nico Pfeifer. Inherently interpretable position-aware convolutional motif kernel networks for biological se- quencing data. Scientific Reports, 13(1):17216, 2023

    2023

  2. [10]

    Sparse activations for interpretable disease grading

    Kerol R Djoumessi Donteu, Indu Ilanchezian, Laura K ¨uhlewein, Hanna Faber, Christian F Baumgartner, Bubacarr Bah, Philipp Berens, and Lisa M Koch. Sparse activations for interpretable disease grading. In Medical Imaging with Deep Learning , 2024

    2024

  3. [11]

    Gigapixel end-to-end training using streaming and attention

    Stephan Dooper, Hans Pinckaers, Witali Aswolinskiy, Konnie Hebeda, Sofia Jarkman, Jeroen van der Laak, Geert Litjens, BIGPICTURE Consortium, et al. Gigapixel end-to-end training using streaming and attention. Medical Image Analysis , 88:102881, 2023

    2023

  4. [12]

    An update of the gleason grading system

    Jonathan I Epstein. An update of the gleason grading system. The Journal of urology, 183(2):433–440, 2010

    2010

  5. [13]

    Sarah Fremond, Sonali Andani, Jurriaan Barkey Wolf, Jouke Dijkstra, Sin´ead Melsbach, Jan J Jobsen, Mariel Brinkhuis, Suzan Roothaan, Ina Jurgenliemk-Schulz, Ludy CHW Lutgens, et al. Interpretable deep learning model to predict the molecular classification of endometrial cance...

    2023

  6. [14]

    Pannuke: an open pan-cancer histology dataset for nuclei instance segmentation and classification

    Jevgenij Gamper, Navid Alemi Koohbanani, Ksenija Benet, Ali Khuram, and Nasir Rajpoot. Pannuke: an open pan-cancer histology dataset for nuclei instance segmentation and classification. In Digital Pathology: 15th European Congress, ECDP 2019, Warwick, UK, April 10–13, 2019, Pr...

    2019

  7. [15]

    Towards automatic concept-based explanations

    Amirata Ghorbani, James Wexler, James Y Zou, and Been Kim. Towards automatic concept-based explanations. Advances in neural information processing systems, 32, 2019

    2019

  8. [16]

    Hover-net: Simulta- neous segmentation and classification of nuclei in multi-tissue histology images

    Simon Graham, Quoc Dang Vu, Shan E Ahmed Raza, Ayesha Azam, Yee Wah Tsang, Jin Tae Kwak, and Nasir Rajpoot. Hover-net: Simulta- neous segmentation and classification of nuclei in multi-tissue histology images. Medical image analysis , 58:101563, 2019

    2019

  9. [17]

    Regression concept vectors for bidirectional explanations in histopathology

    Mara Graziani, Vincent Andrearczyk, and Henning M ¨uller. Regression concept vectors for bidirectional explanations in histopathology. In Understanding and Interpreting Machine Learning in Medical Image Computing Applications: First International Workshops, MLCN 2018, DLF 2018...

    2018

  10. [18]

    Explainable discovery of disease biomarkers: The case of ovarian cancer to illustrate the best practice in machine learning and shapley analysis

    Weitong Huang, Hanna Suominen, Tommy Liu, Gregory Rice, Carlos Salomon, and Amanda S Barnard. Explainable discovery of disease biomarkers: The case of ovarian cancer to illustrate the best practice in machine learning and shapley analysis. Journal of Biomedical Informatics, 14...

    2023

  11. [19]

    A visual–language foundation model for pathology image analysis using medical twitter

    Zhi Huang, Federico Bianchi, Mert Yuksekgonul, Thomas J Montine, and James Zou. A visual–language foundation model for pathology image analysis using medical twitter. Nature Medicine , pages 1–10, 2023

    2023

  12. [20]

    Attention-based deep multiple instance learning

    Maximilian Ilse, Jakub Tomczak, and Max Welling. Attention-based deep multiple instance learning. In International conference on machine learning, pages 2127–2136. PMLR, 2018

    2018

  13. [21]

    Additive mil: Intrinsically interpretable multiple instance learning for pathology

    Syed Ashar Javed, Dinkar Juyal, Harshith Padigela, Amaro Taylor- Weiner, Limin Yu, and Aaditya Prakash. Additive mil: Intrinsically interpretable multiple instance learning for pathology. Advances in Neural Information Processing Systems , 35:20689–20702, 2022

    2022

  14. [22]

    Si-mil: Taming deep mil for self-interpretability in gigapixel histopathology

    Saarthak Kapse, Pushpak Pati, Srijan Das, Jingwei Zhang, Chao Chen, Maria Vakalopoulou, Joel Saltz, Dimitris Samaras, Rajarsi R Gupta, and Prateek Prasanna. Si-mil: Taming deep mil for self-interpretability in gigapixel histopathology. In Proceedings of the IEEE/CVF Conference...

    2024

  15. [23]

    Interpretability beyond feature attribution: Quan- titative testing with concept activation vectors (tcav)

    Been Kim, Martin Wattenberg, Justin Gilmer, Carrie Cai, James Wexler, Fernanda Viegas, et al. Interpretability beyond feature attribution: Quan- titative testing with concept activation vectors (tcav). In International conference on machine learning , pages 2668–2677. PMLR, 2018

    2018

  16. [24]

    Concept bottleneck models

    Pang Wei Koh, Thao Nguyen, Yew Siang Tang, Stephen Mussmann, Emma Pierson, Been Kim, and Percy Liang. Concept bottleneck models. In International conference on machine learning , pages 5338–5348. PMLR, 2020

    2020

  17. [25]

    A multi-resolution model for histopathology image classification and localization with multiple instance learning

    Jiayun Li, Wenyuan Li, Anthony Sisk, Huihui Ye, W Dean Wallace, William Speier, and Corey W Arnold. A multi-resolution model for histopathology image classification and localization with multiple instance learning. Computers in biology and medicine , 131:104253, 2021

    2021

  18. [26]

    A visual-language foundation model for computational pathology

    Ming Y Lu, Bowen Chen, Drew FK Williamson, Richard J Chen, Ivy Liang, Tong Ding, Guillaume Jaume, Igor Odintsov, Long Phi Le, Georg Gerber, et al. A visual-language foundation model for computational pathology. Nature Medicine, 30:863–874, 2024

    2024

  19. [27]

    Data-efficient and weakly supervised computational pathology on whole-slide images

    Ming Y Lu, Drew FK Williamson, Tiffany Y Chen, Richard J Chen, Matteo Barbieri, and Faisal Mahmood. Data-efficient and weakly supervised computational pathology on whole-slide images. Nature biomedical engineering, 5(6):555–570, 2021

    2021

  20. [28]

    Label-free concept bottleneck models

    Tuomas Oikarinen, Subhro Das, Lam M Nguyen, and Tsui-Wei Weng. Label-free concept bottleneck models. arXiv preprint arXiv:2304.06129, 2023

    2023 arXiv

  21. [29]

    Chatgpt: Optimizing language models for dialogue

    OpenAI. Chatgpt: Optimizing language models for dialogue. https: //openai.com/research/chatgpt, 2023. Accessed: 2024-10- 07

    2023

  22. [30]

    Predicting biochemical recurrence of prostate cancer with artificial intelligence

    Hans Pinckaers, Jolique van Ipenburg, Jonathan Melamed, Angelo De Marzo, Elizabeth A Platz, Bram van Ginneken, Jeroen van der Laak, and Geert Litjens. Predicting biochemical recurrence of prostate cancer with artificial intelligence. Communications Medicine, 2(1):64, 2022

    2022

  23. [31]

    Improving interpretability for computer-aided diag- nosis tools on whole slide imaging with multiple instance learning and gradient-based explanations

    Antoine Pirovano, Hippolyte Heuberger, Sylvain Berlemont, Sa ¨ıd Ladjal, and Isabelle Bloch. Improving interpretability for computer-aided diag- nosis tools on whole slide imaging with multiple instance learning and gradient-based explanations. In Interpretable and Annotation-...

    2020

  24. [32]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning , p...

    2021

  25. [33]

    Discover- then-name: Task-agnostic concept bottlenecks via automated concept discovery

    Sukrut Rao, Sweta Mahajan, Moritz B ¨ohle, and Bernt Schiele. Discover- then-name: Task-agnostic concept bottlenecks via automated concept discovery. arXiv preprint arXiv:2407.14499 , 2024

    2024 arXiv

  26. [34]

    Silhouettes: a graphical aid to the interpretation and validation of cluster analysis

    Peter J Rousseeuw. Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. Journal of computational and applied mathematics, 20:53–65, 1987

    1987

  27. [35]

    Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead

    Cynthia Rudin. Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. nat mach intell 1 (5): 206–215, 2019

    2019

  28. [36]

    Transmil: Transformer based correlated multiple instance learning for whole slide image classification

    Zhuchen Shao, Hao Bian, Yang Chen, Yifeng Wang, Jian Zhang, Xiangyang Ji, et al. Transmil: Transformer based correlated multiple instance learning for whole slide image classification. Advances in neural information processing systems , 34:2136–2147, 2021

    2021

  29. [37]

    Deep neural network models for computational histopathology: A survey

    Chetan L Srinidhi, Ozan Ciga, and Anne L Martel. Deep neural network models for computational histopathology: A survey. Medical image analysis, 67:101813, 2021

    2021

  30. [38]

    Inherently interpretable multi-label classification using class-specific counterfactuals

    Susu Sun, Stefano Woerner, Andreas Maier, Lisa M Koch, and Chris- tian F Baumgartner. Inherently interpretable multi-label classification using class-specific counterfactuals. arXiv preprint arXiv:2303.00500 , 2023

    2023 arXiv

  31. [39]

    Differentiable zooming for multiple instance learning on whole-slide images

    Kevin Thandiackal, Boqi Chen, Pushpak Pati, Guillaume Jaume, Drew FK Williamson, Maria Gabrani, and Orcun Goksel. Differentiable zooming for multiple instance learning on whole-slide images. In European Conference on Computer Vision , pages 699–715. Springer, 2022

    2022

  32. [40]

    Visualizing data using t-sne

    Laurens Van der Maaten and Geoffrey Hinton. Visualizing data using t-sne. Journal of machine learning research , 9(11), 2008

    2008

  33. [41]

    Machine learning models for multiparametric glioma grading with quantitative result interpretations

    Xiuying Wang, Dingqian Wang, Zhigang Yao, Bowen Xin, Bao Wang, Chuanjin Lan, Yejun Qin, Shangchen Xu, Dazhong He, and Yingchao Liu. Machine learning models for multiparametric glioma grading with quantitative result interpretations. Frontiers in neuroscience , 12:1046, 2019

    2019

  34. [42]

    Mprotonet: A case- based interpretable model for brain tumor classification with 3d multi- parametric magnetic resonance imaging

    Yuanyuan Wei, Roger Tam, and Xiaoying Tang. Mprotonet: A case- based interpretable model for brain tumor classification with 3d multi- parametric magnetic resonance imaging. In Medical Imaging with Deep Learning, pages 1798–1812. PMLR, 2024

    2024

  35. [43]

    Whole slide images based cancer survival predic- 11 tion using attention guided deep multiple instance learning networks

    Jiawen Yao, Xinliang Zhu, Jitendra Jonnagaddala, Nicholas Hawkins, and Junzhou Huang. Whole slide images based cancer survival predic- 11 tion using attention guided deep multiple instance learning networks. Medical Image Analysis , 65:101789, 2020

    2020

  36. [44]

    Post-hoc concept bottleneck models

    Mert Yuksekgonul, Maggie Wang, and James Zou. Post-hoc concept bottleneck models. arXiv preprint arXiv:2205.15480 , 2022

    2022 arXiv

  37. [45]

    Top-down neural attention by excitation backprop

    Jianming Zhang, Sarah Adel Bargal, Zhe Lin, Jonathan Brandt, Xiaohui Shen, and Stan Sclaroff. Top-down neural attention by excitation backprop. International Journal of Computer Vision , 126(10):1084– 1102, 2018

    2018

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.