Pith. sign in

REVIEW 5 major objections 6 minor 31 references

Hierarchical Vision Transformer with Prototypes for Interpretable Medical Image Classification

T0 review · 5 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read HierViT is a Vision Transformer that classifies medical images by first scoring human-defined attributes, then showing prototype images and attention heatmaps, claiming both high accuracy and clinically aligned reasoning.

desk verdict A well-specified ViT+prototype hybrid with a real LIDC accuracy gain, but the interpretability claim needs a faithfulness test before it holds up. read the letter →

arxiv 2502.08997 v1 pith:ZA7AEYA6 submitted 2025-02-13 cs.CV

Luisa Gallée , Catharina Silvia Lisson , Meinrad Beer , Michael Götz This is my paper · ORCID
classification cs.CV
keywords ExplainableAIHierarchicalPredictionPrototypeLearningVisionTransformerMedicalimageclassificationLIDC-IDRIderm7ptAttentionheatmaps
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 introduces HierViT, a Vision Transformer that produces its final medical diagnosis only after scoring a set of human-defined visual attributes, showing a prototype image for each score, and highlighting the image region behind each attribute. On LIDC-IDRI, it reports 94.8% Within-1-Accuracy for lung-nodule malignancy, above the listed comparison methods, and on derm7pt it reaches a target accuracy comparable to the strongest baselines. The aim is to show that interpretability can be built into the architecture itself, so clinicians can check the model's reasoning in the same feature language they use, rather than relying on post-hoc attention maps. If this works, high-accuracy vision transformers become auditable tools for high-risk medical decisions.

What carries the argument

The load-bearing mechanism is the hierarchical attribute-token classifier: after the ViT encoder produces a token vector, a separate transformer layer per attribute scores that attribute, and these attribute token vectors are stacked and passed through a target transformer that makes the final prediction. Each attribute layer is paired with a prototype layer holding 16 learnable vectors per attribute class, trained with a Euclidean-distance loss; a push operation then replaces each prototype with the closest real training sample, so prototypes are actual images rather than synthetic abstractions. The optional ViT-based decoder adds a segmentation mask, and the training loss is the sum of target, attribute, and segmentation losses, with a warm-up phase before prototype learning begins. What makes the design 'inherently interpretable' is that the target head consumes only attribute-level representations, not raw image tokens, so the final decision is formally a function of human-defined feature scores.

What would settle it

A concrete intervention experiment would settle the causal claim: take a correctly classified LIDC-IDRI nodule, replace the attribute token for a high-stakes feature such as spiculation with the prototype vector of the opposite rating while leaving all image tokens fixed, and check whether the malignancy prediction shifts toward the clinically expected direction. If large attribute interventions leave target predictions essentially unchanged across many samples, the claim that the model reasons through human-defined attributes is falsified. A complementary check is to retrain without attribute supervision entirely and see whether target accuracy stays high; if it does, the attribute hierarchy is not carrying the diagnostic signal.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that a Vision Transformer can be restructured so that the final classification is computed from attribute-level token vectors, each produced by a dedicated transformer layer and tied to learnable prototype vectors. The attribute branch scores clinically defined features, the prototype layer associates each score with real training images via a push operation, and the target branch consumes the stacked attribute tokens to make the final call. The paper reports that this design reaches 94.8% Within-1-Accuracy on LIDC-IDRI malignancy and 73.8% target accuracy on derm7pt, and that a prototype-inference variant, where the target is predicted from the closest prototypes instead of the attribute heads, retains 94.4% on LIDC-IDRI. This is presented as evidence that interpretability need not be added after the fact: the model's intermediate decisions can be expressed in the same features radiologists and dermatologists use.

Load-bearing premise

The load-bearing premise is that the radiologist- and dermatologist-defined attribute labels are complete and reliable enough that scoring these attributes first is the right causal path to the diagnosis; if those labels are noisy or miss the true decision boundary, the target branch may learn to bypass the attribute semantics and the promised human-aligned reasoning does not hold.

Editorial extensions

If this is right

  • A clinician could validate a prediction by checking each attribute score, comparing the prototype image with the inference image, and inspecting the attribute-specific attention heatmap; mismatches would flag an unreliable call before it reaches the diagnosis.
  • The prototype-inference variant shows that even when the target head is forced to use prototype vectors rather than learned attribute heads, LIDC-IDRI accuracy only drops from 94.8% to 94.4%, suggesting that explanations and accuracy do not have to trade off.
  • Because the architecture is built on a standard ViT encoder, the same hierarchical attribute design could be transferred to other medical imaging tasks where discrete clinical attributes exist, such as breast imaging or chest radiograph reporting.
  • On derm7pt, treating lesion features as a hierarchy rather than as independent multi-label outputs yields accuracy comparable to the best listed baselines, indicating that structuring the task by clinical attributes does not cost prediction performance.
  • The attention heatmaps are not generic saliency maps but are attached to individual attributes, giving a potential new way to localize pathological structures: the spiculation heatmap focuses on nodule edges, while the sphericity and lobulation heatmaps focus on the interior.

Reading between the lines

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

  • A direct intervention experiment—swapping the attribute token for 'spiculation' to the opposite prototype or score while keeping the image tokens fixed—would test whether the target prediction actually moves in the clinically expected direction; the paper does not report such a test, but the architecture makes it straightforward.
  • If attribute labels are noisy or subjective, the target branch might learn to exploit correlations among attributes rather than their clinical meaning; one way to check is to corrupt or hold out attribute labels during training and measure how much target accuracy and explanation fidelity degrade.
  • The prototype push operation stores real training images, so the explanations inherit any dataset bias; on underrepresented nodule or lesion subtypes, the nearest prototypes may be misleading even if the attribute score is correct.
  • The reported expert consultation on LIDC-IDRI is qualitative; a formal evaluation with multiple radiologists, measuring how often they change their diagnosis after seeing the attributes and prototypes, would quantify whether the explanations truly align with human reasoning.
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

5 major / 6 minor

Summary. The paper proposes HierViT, a Vision Transformer with a hierarchical classification branch that predicts human-defined attribute scores (e.g., spiculation, sphericity for lung nodules; dermatoscopic features for skin lesions), uses those attribute token vectors to predict the final target, and offers interpretability through attribute scores, prototype images, and attention heatmaps. An optional decoder produces segmentation masks when available. The method is evaluated on LIDC-IDRI for lung nodule malignancy, where it reports a Within-1-Accuracy of 94.8% for the target, and on derm7pt for skin lesion classification, where it reports accuracy comparable to existing multi-task methods. The paper claims that HierViT is 'inherently interpretable' and that its reasoning is aligned with human reasoning because the target prediction is derived from human-defined attributes. The central scientific claims are that the hierarchical attribute branch improves accuracy over previous prototype-based methods and that the resulting explanations are causally tied to the model's decision process.

Significance. If the central interpretability claim were established, HierViT would be a useful contribution to explainable medical image classification: it combines ViT-based feature extraction with hierarchical attribute prediction, prototype learning, and attention visualization, and it reports results on two public benchmarks with confidence intervals. The architecture and losses are clearly specified, and the use of external benchmarks with radiologist-defined attributes is a strength. However, the paper does not currently provide evidence that the target prediction actually depends on the attribute semantics in the way the explanation claims. The only causal-looking experiment, the 'w proto. inference' variant in Table 1, is an oracle experiment that uses ground-truth attribute labels, and no perturbation or ablation test is reported. In addition, parts of the accuracy comparison mix incompatible metrics (Within-1-Accuracy versus binary accuracy) and the derm7pt differences are within statistical noise. These issues bear directly on the paper's two headline claims, so the manuscript needs substantial revision before the claims can be accepted.

major comments (5)
  1. [Section 2 (Eqs. (1)–(3), Fig. 1)] The target branch consumes the stacked attribute token vectors, not the scalar attribute scores produced by Eq. (1). Nothing in the architecture or the losses forces the target prediction to depend on the human-defined attribute semantics; the token vectors are high-dimensional and can carry shortcut information unrelated to the attributes. The claim in Section 4 that 'these human-defined attributes ... feed into the target prediction branch' is therefore not established. Please add an intervention test, such as shuffling or masking the predicted attribute scores at inference and measuring the change in target accuracy, or training a variant whose target branch receives only the scalar attribute scores. This is necessary to support the central 'inherently interpretable' claim.
  2. [Section 3.3 (Table 1, 'HierViT w proto. inference')] The 'w proto. inference' variant uses the ground-truth attribute value to select prototypes, as the text states ('The prototype's groundtruth attribute value is used for prediction, ignoring attribute heads'). This is an oracle experiment: it shows that the target branch can classify when given correct attribute information, but it does not show that the deployed model's target prediction follows its own predicted attributes. Please report a non-oracle variant that uses the model's predicted attribute scores or prototype selections, and explicitly discuss the oracle limitation in the text.
  3. [Section 3.3 (Table 1)] The state-of-the-art comparison mixes incompatible metrics: TransUnet, Res-trans, and TransPND report binary classification accuracy (marked with an asterisk), while HierViT reports Within-1-Accuracy on the five-level malignancy rating. These numbers are not directly comparable, so the statement that HierViT 'outperforms SOTA methods' is not supported for those rows. Additionally, the LIDC-IDRI preprocessing (excluding nodules detected by fewer than three radiologists or smaller than 3 mm) and the cross-validation split may differ from the protocols used by the compared methods; please provide a same-protocol comparison or explicitly state the comparability limits.
  4. [Section 3.3 (Table 2)] On derm7pt, the reported differences between HierViT and FusionM4Net are within overlapping 95% binomial confidence intervals (e.g., target 73.8 [69.5, 78.1] versus 74.3 [70.0, 78.6]). The abstract's phrase 'superior and comparable prediction accuracy' and the text's statement that HierViT 'attain[s] the best accuracy in target prediction' overstate the evidence. Please report a paired significance test or a non-overlapping-interval criterion, or revise the claim to say that performance is statistically indistinguishable from the best baseline.
  5. [Section 3 (all experiments)] No ablation study is reported for the proposed components: the hierarchical attribute branch, the prototype loss, the decoder branch, the warm-up phase, and the choice of target-branch input. Since the paper's novelty is the combination of these components, ablations are needed to show which components contribute to target accuracy and to justify hyperparameters such as lambda_proto = 0.01 and push_step = 2. Without ablations, the performance and interpretability claims cannot be attributed to the proposed architecture.
minor comments (6)
  1. [Section 1] The code availability statement contains the placeholder 'https://github.com/XXX'; please provide the actual repository URL or remove the claim that the code is publicly available.
  2. [Section 2 (Eq. (4))] The abbreviation 'CSE' for cross-entropy loss is nonstandard; the usual abbreviation is 'CE'. Please correct this throughout.
  3. [Section 3.1] For LIDC-IDRI, please state whether the reported 95% confidence interval is computed over pooled predictions or over the five cross-validation folds, and clarify how many test samples contribute to the interval.
  4. [Section 3.2] The qualitative evaluation relies on a single expert's consultation; this should be described as an anecdotal check rather than as a systematic evaluation of explanation quality.
  5. [References] Reference [27] (Pathak et al.) appears in the bibliography but is not cited in the text; please cite it where relevant or remove it.
  6. [Table 2] The average column 'Ø' is the mean over attributes and target; please justify this equal-weight aggregation, since attributes and target have different numbers of classes and different clinical importance.

Circularity Check

2 steps flagged · score 4.0 of 10

External accuracy results are not circular, but the human-aligned interpretability claim reduces to training on human-defined attribute labels plus an oracle prototype variant.

  1. self definitional [Abstract; Section 2, Eq. (1)]
    "It is interpretable by design, as it derives the target output with human-defined features that are visualized by exemplary images (prototypes). ... The loss function term for attribute learning Lattr minimizes the mean value of the classification error Lclass over all attributes. In the following yattr_a is the ground truth label of attribute a=1...A, and yhat_attr_a is the respective prediction: Lattr = 1/A sum_a Lclass(yattr_a, yhat_attr_a). (1)"

    The 'human-defined features' that ground the interpretability claim are exactly the attribute labels yattr_a in Eq. (1), and the model is trained to predict those labels. Saying the model 'adapts its reasoning to that of humans' because it uses human-defined features is therefore equivalent to saying it fits the human labels it was optimized on. Moreover, Section 2 states the target is predicted from stacked attribute token vectors, not from the scalar human-defined ratings, so the architecture does not force the target output to depend on the human-defined semantics; the attribute heads may be epiphenomenal to the target prediction.

  2. fitted input called prediction [Section 3.3, LIDC-IDRI results]
    "The w. proto. inference variant extends the method by replacing attribute token vectors with the closest prototype vectors during inference for target prediction, similar to Proto-Caps [8]. The prototype’s groundtruth attribute value is used for prediction, ignoring attribute heads. While this slightly reduces prediction performance, it makes prototypes directly causal for target prediction, enhancing explainability credibility."

    This variant feeds the ground-truth attribute values into the target branch, so the claim that prototypes are 'directly causal' is tested only when the correct human labels are provided as input. The reported target accuracy under this oracle variant shows the target branch can classify given true attributes; it does not show that the deployed model's predicted attributes or prototypes cause the target prediction. The causal role is thus an artifact of injecting ground-truth labels, i.e., the input contains the target information by construction.

full rationale

The accuracy claims on LIDC-IDRI and derm7pt are evaluated against external benchmarks and are not circular: the target and attribute accuracies are measured on held-out data. The circularity burden is confined to the interpretability narrative. The paper's core claim that HierViT is 'inherently interpretable' and 'adapts its reasoning to that of humans' rests on the fact that the model is trained to predict human-defined attribute labels (Eq. 1) and then the target is predicted from attribute-related token vectors. Since the attribute labels are the training targets of Lattr, reporting that the model predicts them well is a statement about fitting, not an independent confirmation of human-like reasoning. Further, the only experiment intended to show that prototypes are causal for the target, the 'w proto. inference' variant, substitutes the ground-truth attribute value into the target branch, making the target prediction an oracle evaluation rather than a test of the deployed model's reliance on its inferred attributes. The self-citations to Proto-Caps [8] and the user study [10] are present, but the central benchmark comparisons do not depend on them; hence the overall circularity is partial, concentrated in the interpretability and causal-credibility claims.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The ledger shows that the method rests on standard deep learning assumptions plus two domain-specific bets: that the human-defined attributes are reliable and causally relevant, and that prototype and attention visualizations reflect the model's true reasoning. There are no invented physical entities. The main free parameters are hyperparameters, not fitted constants in a derivation, and no sensitivity analysis is reported.

free parameters (6)
  • lambda_proto = 0.01
    Weight of the prototype loss in the final training phase (Eq. 7); chosen by hand to keep focus on the primary task.
  • number_of_prototypes_per_attribute_value = 16
    Eq. 2 sums p = 1 to 16 for each attribute value; fixed without reported tuning.
  • push_step = 2
    Frequency of the prototype push operation, set as a hyperparameter.
  • learning_rates = lr=0.001 (LIDC), lr=0.00001 with prototype lr=0.01 (derm7pt)
    Chosen per dataset; no search or sensitivity analysis is reported.
  • warm_up_epochs = 2 (LIDC), 20 (derm7pt)
    Warm-up duration before the prototype loss is enabled; tuned per dataset.
  • preprocessing_thresholds = nodules >= 3 mm and detected by >= 3 radiologists; derm7pt center crop 450x450
    Data selection rules that affect the LIDC dataset composition and the derm7pt input.
assumptions (5)
  • domain assumption ImageNet-1K pretrained ViT features transfer to CT and dermoscopic images.
    The encoder starts from ImageNet-1K weights and is fine-tuned on medical images; no analysis is given for the transfer gap.
  • domain assumption Human-defined attributes are sufficient and reliably labeled to serve as intermediate targets.
    The hierarchical branch is trained against LIDC appearance ratings and derm7pt features (Eq. 1), assuming these labels are accurate and causally relevant.
  • domain assumption Euclidean distance in attribute feature space corresponds to semantic similarity.
    Prototype loss (Eq. 2) and the push operation choose nearest samples using Euclidean distance.
  • domain assumption Attention heatmaps accurately localize the regions used for each attribute.
    Attention visualizations are presented as evidence of reasoning without specifying or validating the heatmap method.
  • domain assumption The target prediction genuinely depends on attribute semantics rather than bypassing them.
    The target branch stacks attribute token vectors, but no intervention or perturbation test confirms the final prediction uses the attributes as intended.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hierarchical Vision Transformer with Prototypes for Interpretable Medical Image Classification." pith.science (2026). https://pith.science/paper/ZA7AEYA6

@misc{pith2026250208997,
  author       = {Pith},
  title        = {Pith review of: Hierarchical Vision Transformer with Prototypes for Interpretable Medical Image Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZA7AEYA6}},
  note         = {Machine review of arXiv:2502.08997}
}
read the original abstract

Explainability is a highly demanded requirement for applications in high-risk areas such as medicine. Vision Transformers have mainly been limited to attention extraction to provide insight into the model's reasoning. Our approach combines the high performance of Vision Transformers with the introduction of new explainability capabilities. We present HierViT, a Vision Transformer that is inherently interpretable and adapts its reasoning to that of humans. A hierarchical structure is used to process domain-specific features for prediction. It is interpretable by design, as it derives the target output with human-defined features that are visualized by exemplary images (prototypes). By incorporating domain knowledge about these decisive features, the reasoning is semantically similar to human reasoning and therefore intuitive. Moreover, attention heatmaps visualize the crucial regions for identifying each feature, thereby providing HierViT with a versatile tool for validating predictions. Evaluated on two medical benchmark datasets, LIDC-IDRI for lung nodule assessment and derm7pt for skin lesion classification, HierViT achieves superior and comparable prediction accuracy, respectively, while offering explanations that align with human reasoning.

Figures

Figures reproduced from arXiv: 2502.08997 by the authors.

Figure 1
Figure 1. Proposed model The patchified image is linearly projected and pro￾cessed by a transformer encoder, producing a token vector that serves as the input for both a hierarchical classifier and a decoder. The hierarchical classifier processes the token vector through multiple transformer layers, one for each at￾tribute, with individual heads providing attribute ratings. For target prediction, the token vectors from the at… view at source ↗
Figure 2
Figure 2. Reasoning process Three sample cases are illustrated, (a) correctly predicted, (b) and (c) incorrectly predicted. For three of the eight attributes (spiculation, sphericity, lobulation), the score, attention heatmap, and prototype image of the respective attribute are displayed. attribute ratings reflect the visual characteristics of the sample nodule, and the closest attribute prototypes exhibit similar traits. The… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 23 canonical work pages

  1. [1]

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T. et al. : An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. InProc. ICLR(2021)

    work page 2021

  2. [2]

    et al.: Segment Anything

    Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L. et al.: Segment Anything. InProc. ICCV,pp. 4015–4026 (2023)

    work page 2023

  3. [3]

    Xie, E., Wang, W., Yu, Z., Anandkumar, A., Alvarez, J.M., Luo, P.: SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers. InProc. NeurIPS, vol. 34, pp. 12077–12090 (2021)

    work page 2021

  4. [4]

    Caron, M., Touvron, H., Misra, I., Jégou, H., Mairal, J., Bojanowski, P.et al.: Emerging Properties in Self-Supervised Vision Transformers. InProc. ICCV,pp. 9650–9660 (2021)

    work page 2021

  5. [5]

    He, K., Chen, X., Xie, S., Li, Y., Dollár, P., Girshick, R.: Masked Autoencoders Are Scalable Vision Learners. InProc. CVPR,pp. 16000–16009 (2022)

    work page 2022

  6. [6]

    In Nat Mach Intell, vol

    Rudin, C.: Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead. In Nat Mach Intell, vol. 1, pp. 206–215 (2019) https://doi.org/10.1038/s42256-019-0048-x

    doi:10.1038/s42256-019-0048-x 2019

  7. [7]

    Chen, C., Li, O., Tao, D., Barnett, A., Rudin, C., Su, J.K.: This Looks Like That: Deep Learning for Interpretable Image Recognition. In Proc. NeurIPS, vol. 32 (2019)

    work page 2019

  8. [8]

    Gallée, L., Beer, M., Götz, M.: Interpretable Medical Image Classification Using Prototype Learning and Privileged Information. InProc. MICCAI, pp. 435–445 (2023) https://doi.org/10.1007/978-3-031-43895-0_41

    work page doi:10.1007/978-3-031-43895-0_41 2023

Show all 31 references
  1. [9]

    et al.: On the interpretability of artificial intelligence in radiology: challenges and opportunities

    Reyes, M., Meier, R., Pereira, S., Silva, C.A., Dahlweid, F., von Tengg-Kobligk, H. et al.: On the interpretability of artificial intelligence in radiology: challenges and opportunities. InRadiology: artificial intelligence, vol. 2, no. 3 (2020) https: //doi.org/10.1148/ryai.2...

    2020 doi

  2. [10]

    Gallée, L., Lisson, C.S., Lisson, C.G., Drees, D., Weig, F., Vogele, D.et al.: Evalu- ating the Explainability of Attributes and Prototypes for a Medical Classification Model. InProc. xAI, (2024) https://doi.org/10.1007/978-3-031-63787-2_3

    2024 doi

  3. [11]

    (2022) https://doi.org/10.48550/arXiv.2208.10431

    Xue, M., Huang, Q., Zhang, H., Cheng, L., Song, J., Wu, M.et al.: ProtoPFormer: Concentrating on Prototypical Parts in Vision Transformers for Interpretable Im- age Recognition. (2022) https://doi.org/10.48550/arXiv.2208.10431

    2022 doi

  4. [12]

    In IET Image Processing (2024) https://doi.org/10.1049/ ipr2.13074

    Xu, Y., Meng, Z.: Interpretable vision transformer based on prototype parts for COVID-19 detection. In IET Image Processing (2024) https://doi.org/10.1049/ ipr2.13074

    2024

  5. [13]

    (2024) https://doi.org/ 10.48550/arXiv.2403.06961

    Demir, U., Jha, D., Zhang, Z., Keles, E., Allen, B., Katsaggelos, A.K.et al.: Ex- plainable Transformer Prototypes for Medical Diagnoses. (2024) https://doi.org/ 10.48550/arXiv.2403.06961

    2024 doi arXiv

  6. [14]

    Rigotti, M., Miksovic, C., Giurgiu, I., Gschwind, T., Scotton, P.: Attention-based Interpretability with Concept Transformers. InProc. ICLR(2022)

    2022

  7. [15]

    InTCIA (2015) https://doi.org/10

    Armato III, S.G., McLennan, G., Bidaut, L., McNitt-Gray, M.F., Meyer, C.R., Reeves, A.P.et al..: Data From LIDC-IDRI. InTCIA (2015) https://doi.org/10. 7937/K9/TCIA.2015.LO9QL9SX

    2015

  8. [16]

    et al.: The lung image database consortium (LIDC) and image database resource initiative (IDRI): a completed reference database of lung nod- ules on CT scans

    Armato III, S.G., McLennan, G., Bidaut, L., McNitt-Gray, M.F., Meyer, C.R., Reeves, A.P. et al.: The lung image database consortium (LIDC) and image database resource initiative (IDRI): a completed reference database of lung nod- ules on CT scans. In Med. Phys., vol. 38, no. 2...

    2011 doi

  9. [17]

    Hancock, M.C., Magnan, J.F.: Lung nodule malignancy classification using only radiologist-quantified image features as inputs to statistical learning algorithms: probing the Lung Image Database Consortium dataset with two statistical learning methods. InJ. Med. Imaging,vol. 3,...

    2016 doi

  10. [18]

    InJBHI, vol

    Kawahara, J., Daneshvar, S., Argenziano, G., Hamarneh, G.: Seven-Point Checklist and Skin Lesion Classification Using Multitask Multimodal Neural Nets. InJBHI, vol. 23, pp. 538–546 (2019) https://doi.org/10.1109/JBHI.2018.2824327

    2019

  11. [19]

    InArch Dermatol., vol

    Argenziano, G., Fabbrocini, G., Carli, P., De Giorgi, V., Sammarco, E., Delfino, M.: Epiluminescence microscopy for the diagnosis of doubtful melanocytic skin lesions: comparison of the ABCD rule of dermatoscopy and a new 7-point checklist based on pattern analysis. InArch Der...

    1998 doi

  12. [20]

    Hussein, S., Cao, K., Song, Q., Bagci, U.: Risk stratification of lung nodules using 3D CNN-based multi-task learning. InProc. IPMI,pp. 249–260 (2017) https:// doi.org/10.1007/978-3-319-59050-9_20

    2017 doi

  13. [21]

    Hussein, S., Gillies, R., Cao, K., Song, Q., Bagci, U.: Tumornet: Lung nodule char- acterization using multi-view convolutional neural network with gaussian process. In Proc. ISBI,pp. 1007–1010 (2017) https://doi.org/10.1109/ISBI.2017.7950686

    2017

  14. [22]

    LaLonde, R., Torigian, D., Bagci, U.: Encoding visual attributes in capsules for explainable medical diagnoses. InProc. MICCAI,pp. 294–304 (2020) https://doi. org/10.1007/978-3-030-59710-8_29

    2020 doi

  15. [23]

    Wang, H., Zhu, H., Ding, L.: Accurate classification of lung nodules on CT images using the TransUnet. InFront. Public Health,vol. 10 (2022) https://doi.org/10. 3389/fpubh.2022.1060798

    2022

  16. [24]

    InInt J CARS,vol

    Liu, D., Liu, F., Tie, Y., Qi, L., Wang, F.: Res-trans networks for lung nodule classification. InInt J CARS,vol. 17, no. 6, pp. 1059–1068 (2022) https://doi.org/ 10.1007/s11548-022-02576-5

    2022 doi

  17. [25]

    Wang, R., Zhang, Y., Yang, J.: TransPND: A Transformer Based Pulmonary Nodule Diagnosis Method on CT Image. In Proc. PRCV, pp. 348–360 (2022) https://doi.org/10.1007/978-3-031-18910-4_29

    2022 doi

  18. [26]

    InMIA, vol

    Wang, Y., Feng, Y., Zhang, L., Zhou, J.T., Liu, Y., Goh, R.S.M.et al.: Adversarial multimodal fusion with attention mechanism for skin lesion classification using clinical and dermoscopic images. InMIA, vol. 81, pp. 102535 (2022) https://doi. org/10.1016/j.media.2022.102535

    2022

  19. [27]

    Pathak, S., Schlötterer, J., Veltman, J., Geerdink, J., van Keulen, M., Seifert, C.: Prototype-Based Interpretable Breast Cancer Prediction Models: Analysis and Challenges. InProc. xAI, (2024) https://doi.org/10.1007/978-3-031-63787-2_2

    2024 doi

  20. [28]

    In Psy- chological bulletin, vol

    Koehler, D.J.: Explanation, imagination, and confidence in judgment. In Psy- chological bulletin, vol. 110, nr. 3, pp. 499-–519, (1991) https://doi.org/10.1037/ 0033-2909.110.3.499

    1991

  21. [29]

    InPat- tern Recognition, vol

    Bi, L., Feng, D.D., Fulham, M., Kim, J.: Multi-Label classification of multi- modality skin lesion via hyper-connected convolutional neural network. InPat- tern Recognition, vol. 107, pp. 107502 (2020) https://doi.org/10.1016/j.patcog. 2020.107502

    2020

  22. [30]

    In MIA, vol

    Tang, P., Yan, X., Nan, Y., Xiang, S., Krammer, S., Lasser, T.: FusionM4Net: A multi-stagemulti-modallearningalgorithmformulti-labelskinlesionclassification. In MIA, vol. 76, pp. 102307 (2022) https://doi.org/10.1016/j.media.2021.102307

    2022

  23. [31]

    Coppola, D., Lee, H.K., Guan, C.: Interpreting mechanisms of prediction for skin cancer diagnosis using multi-task learning. InProc. CVPR Workshops, pp. 734–735 (2020)

    2020

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.