REVIEW 4 major objections 5 minor 46 references
Retrieval-Augmented Vision Foundation Models for Robust Leukemia Cell Classification across Multiple Microscopy Datasets
T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash
Pith's one-line read A two-stage leukemia classification pipeline reaches 94.23% held-out accuracy for detection, but ALL/AML subtyping collapses to the majority class under the same protocol, which the paper attributes to dataset-specific background shortcuts.
desk verdict The Stage 2 ALL/AML collapse under held-out evaluation is a real and useful negative result, but the paper overstates its causal mechanism and its 'robust' framing; still worth refereeing. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing mechanism is the held-out dataset evaluation protocol: each stage trains on four (Stage 1) or two (Stage 2) datasets and is tested on a dataset never seen in training or validation, so that any surviving accuracy must transfer across acquisition, staining, and background differences. Within that protocol, retrieval-augmented classification (RAC) is the mechanism that attempts to ground predictions in cytomorphology: the query image's embedding is compared by cosine similarity to a bank of labeled training embeddings, the top-$k=20$ neighbors cast a similarity-weighted vote $p_{\mathrm{ret}}$, and the final logits are $\operatorname{logits}_{\mathrm{final}} = \log p_{\mathrm{probe}} + \alpha \log p_{\mathrm{ret}}$, with $\alpha$ selected on an internal validation split for linear-probe configurations and fixed at $0.5$ for LoRA configurations. Low-Rank Adaptation (LoRA) is the cost-effective adaptation mechanism: it freezes the pretrained encoder and trains two low-rank matrices $B$ and $A$ whose product $BA$ is added to the attention projection weights, so only about 0.3 to 0.5 million parameters are trained instead of roughly 86 million. Label harmonization, collapsing five heterogeneous datasets into normal/ALL/AML classes under a uniform inclusion criterion, is what makes cross-dataset training possible, and the synthetic-background replacement of C-NMC 2019 images is the preprocessing step that the paper identifies as the source of the Stage 2 shortcut.
What would settle it
Replace the backgrounds of the 130 held-out ALL-IDB2 cells with the same synthetic healthy-smear background used for C-NMC 2019 training images and retest the Stage 2 classifiers: if ALL recall stays near zero the background-mismatch explanation is unsupported, whereas a large jump would confirm it; alternatively, a held-out set containing ALL and AML cells imaged on the same platform would settle whether subtype classification can transfer once the background confound is removed.
Extended reading notes
Core claim
The paper's central claim is that under a held-out dataset protocol, the value of domain-specialized pretraining is large but mostly replaceable, and the protocol itself exposes a failure that random splits hide. With frozen encoders and linear probing, the hematology-pretrained encoder beats the general-purpose encoder by 0.2577 accuracy on Stage 1 (0.8923 versus 0.6346), but LoRA adaptation shrinks that gap to 0.0193 and lifts the general-purpose encoder to 0.9115, above the specialized encoder's frozen 0.8923; adding retrieval to the specialized encoder gives the best Stage 1 result, 0.9423. In Stage 2, the held-out protocol reveals complete collapse: on a test set of 3,424 cells (130 ALL, 3,294 AML), ALL recall is 0.0000 for the two biomedical encoders and at most 0.0231 for the general-purpose encoder, despite a within-domain random split yielding ALL recall of 0.997 to 1.000 and training loss converging to 0.0000. The paper attributes this to shortcut learning on the synthetic background of the training ALL source, and reports that 96.5% of retrieved neighbors for query ALL images belong to the AML class, consistent with the encoder grouping cells by acquisition source rather than morphology.
Load-bearing premise
The explanation of the Stage 2 collapse assumes that the main difference between the ALL cells used for training (with a synthetic replaced background) and the ALL cells used for testing (with a natural smear background) is the background itself, rather than differences in staining, magnification, cell selection, or the fact that the test set has 130 ALL cells against 3,294 AML cells.
Editorial extensions
If this is right
- A general-purpose vision encoder fine-tuned with LoRA can substitute for a domain-specialized encoder in leukemia detection, so expensive hematology-specific pretraining is not strictly required when a modest fine-tuning budget exists.
- Retrieval augmentation is only useful when the encoder's embedding space already reflects cytomorphology; on general-purpose encoders it can reduce accuracy, so RAC should be enabled based on retrieval neighbor purity rather than by default.
- Within-domain accuracy in subtype classification is not a reliable measure of clinical readiness: the same classifiers that score 99.7–100% ALL recall on random splits collapse to 0–2.3% on a held-out dataset.
- Label harmonization across heterogeneous single-cell datasets enables cross-dataset training, but when each subtype comes from a single imaging source, the model can still learn source-specific shortcuts instead of morphology.
- A practical consequence for screening is that Stage 1 can operate as a first-line triage with 94.23% held-out accuracy, while Stage 2 should not be deployed until datasets with both subtypes under one imaging protocol are available.
Reading between the lines
- The paper does not test this, but the same held-out logic implies that any classifier whose classes are confounded with acquisition source is vulnerable to collapse, so the protocol could serve as a general shortcut detector across medical imaging tasks.
- A decisive experiment would be to apply the synthetic-background normalization to the held-out ALL-IDB2 cells: if ALL recall recovers substantially, the background is confirmed as the shortcut; if not, the bottleneck lies in genuine cytomorphological similarity between ALL and AML blasts.
- Because RAC's benefit tracked encoder specialization, retrieval neighbor purity on a validation set could be used as a cheap predictor of whether to fuse retrieval at inference time, avoiding the observed accuracy drops on general-purpose encoders.
- The paper's framework suggests that multi-center datasets that include both subtypes under shared acquisition protocols would be the next enabling resource; until such data exist, ALL/AML single-cell classification should be reported with a domain-shift evaluation rather than only a random split.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. This manuscript benchmarks three vision encoders (DinoBloom, BiomedCLIP, CLIP) under linear probing, LoRA, and a retrieval-augmented classification (RAC) module for two sequential tasks: Stage 1 binary leukemia detection and Stage 2 ALL/AML subtyping. Five public single-cell microscopy datasets are harmonized into three classes, and a held-out dataset protocol (ALL-IDB2 for Stage 1; ALL-IDB2 plus AML-Cytomorphology for Stage 2) is used to measure domain-shift generalization. Stage 1 reports best accuracy of 0.9423 for DinoBloom with RAC and shows LoRA narrowing the pretraining gap between CLIP and DinoBloom. Stage 2 collapses to the AML class under the held-out protocol (ALL recall near zero in Table 8) while a within-domain random 80/20 split achieves near-perfect ALL recall (Table 9), a contrast the paper attributes to dataset-specific background shortcuts learned during training. The paper concludes that cost-effective adaptation can compensate for most of the advantage of domain-specialized pretraining, and that the held-out protocol is a diagnostic tool for exposure of dataset-specific artifacts.
Significance. The most valuable element of this paper is the Stage 2 negative result: Table 8 versus Table 9 shows that near-perfect in-domain performance (ALL recall up to 1.0000 for frozen CLIP) coexists with near-zero held-out ALL recall across all three encoders, a clean and honestly reported demonstration that within-split evaluation overstates robustness. The label harmonization protocol of Section 4.3 with its explicit inclusion criteria is a concrete reusable contribution, and the paper ships a sensible control design (the random-split experiment) and a candid limitations statement (Section 5.6). If the Stage 1 claims were statistically supported, the paper would provide a useful controlled comparison of pretraining specialization, PEFT, and retrieval augmentation in hematology imaging. As it stands, the Stage 1 comparisons rest on a 260-image test set without confidence intervals, and the Stage 2 causal mechanism (synthetic-background shortcut) is not isolated from other source-level confounds; these two load-bearing points need work before the robustness framing can be accepted.
major comments (4)
- [Section 5.2-5.3, Tables 4, 6, 7] All Stage 1 conclusions about relative gains rest on the n=260 ALL-IDB2 held-out set, and no confidence intervals or significance tests are reported. At this sample size, the 0.0077 difference between DinoBloom LoRA and DinoBloom LoRA+RAC is about 2 images, the claimed closing of the pretraining gap from 0.2577 to 0.0193 is about 5 images, and the headline RAC gain of 0.0500 for DinoBloom linear probing is about 13 images; with binomial standard errors of roughly 0.016-0.028 at this n, most paired comparisons in Table 7 are indistinguishable from sampling noise. The paper should report bootstrap or exact confidence intervals for Tables 4 and 6 and restrict the conclusions in Section 6 ("surpassed", "viable alternative") to contrasts that survive interval comparison.
- [Section 5.4 and Section 6] The causal attribution of the Stage 2 collapse to a synthetic-background shortcut is not isolated from other confounds that vary with the subtype label. In training, ALL images come exclusively from D1 (black background replaced synthetically, Reinhard-normalized) and AML images exclusively from D4 (natural background, patient-level labels, four WHO 2022 genetic subtypes, and acknowledged instance-level label noise); in testing, ALL (D5) and AML (D2) differ in acquisition, stain, magnification, and cell-selection pipeline as well as background. No experiment separates background from these confounds, so Table 8 and Figure 12 establish that performance does not transfer and that source identity predicts the label, but not that background is the mechanism. The conclusive wording in Section 6 ("This is attributable to each subtype class coming from a single dataset...") exceeds the hedged "a possible explanation" in Section 5.4. A testable fix within the paper's scope would be to apply the Section 4.2 background-replacement procedure to the D5 ALL images at inference and check whether ALL recall recovers.
- [Section 5.4] The 96.5% AML-neighbor retrieval statistic is reported without correcting for the Stage 2 retrieval-bank composition: the bank contains 60,909 AML and 8,491 ALL training images (87.8% AML), so random retrieval alone would yield roughly 87.8% AML neighbors. The observed 96.5% is therefore a modest excess over the base rate, no per-query distribution or confidence bound is reported, and the statistic cannot discriminate the background-shortcut hypothesis from other dataset-specific cues that organize the embedding space. The alpha=8 sweep is likewise a consistency check rather than a mechanism test, since a top-20 retrieval set dominated by AML under many embedding geometries would behave identically.
- [Sections 4.7, 5.1, 5.6] The held-out protocol is only partially held out for the domain-specialized encoder. DinoBloom's pretraining corpus is reported to include D2, D3, and D4, so the normal-class and AML appearance distributions in both training and evaluation have been seen during pretraining; only the ALL class (D1 excluded from pretraining, D5 held out) is genuinely unseen. The authors acknowledge this in Section 5.6, but it means the headline pretraining-specialization gap of 0.2577 in Table 5 conflates domain-specialized pretraining with direct exposure to the evaluation domains, and the abstract's "robust... across multiple datasets" framing should be tempered accordingly.
minor comments (5)
- [Table 4, DinoBloom linear probe + RAC row] The reported triple (accuracy 0.9423, recall 0.9423, F1 0.9307) is not jointly realizable on the balanced 130/130 held-out test under the macro-averaged definitions in Equations (3)-(5); please verify the confusion matrix or correct the rounding.
- [Section 4.6] The retrieval top-k is fixed at k=20 with no sensitivity analysis, and alpha is tuned on an internal validation split for linear probing but fixed at 0.5 for LoRA; the paper acknowledges the alpha inconsistency in Section 5.3, but Table 7 should mark the linear-probe and LoRA comparisons as not directly comparable.
- [Sections 1 and 3.2] References [6] and [32] are traffic-sign classification papers cited to support statements about domain shift and VFM reusability in medical imaging; a domain-shift survey or a medical-imaging-specific citation would be more appropriate.
- [Section 4.7] Although the paper describes a two-stage pipeline, the reported metrics are per-stage; no end-to-end evaluation propagates Stage 1 errors into Stage 2, so the performance claims should be labeled as per-stage to avoid implying system-level evaluation.
- [Section 5.4] The statement that "training loss reached 0.0000" should report the actual final loss values and their precision; an exactly zero cross-entropy value is unusual and may be a rounding artifact.
Circularity Check
No significant circularity: the benchmark, held-out protocol, and diagnostic experiments are self-contained and do not reduce to their own inputs.
full rationale
The paper's derivation chain is empirical rather than definitional. Stage 1 and Stage 2 classifiers are trained on datasets that exclude the held-out evaluation sets (D5 for Stage 1; D5 and D2 for Stage 2), and the RAC fusion weight alpha is selected on an internal validation split, never on held-out data; no fitted parameter is renamed as a prediction. The Stage 2 collapse (ALL recall 0.0000-0.0231 under held-out evaluation versus 0.9971-1.0000 under a within-domain random split) is an observed empirical contrast, not an identity derived from the training setup. The retrieval-neighbor statistic (96.5% AML neighbors) is not forced by construction either: although the Stage 2 retrieval bank is 87.8% AML, the observed value is an excess over random retrieval and is presented as diagnostic evidence rather than as a mathematical consequence. The paper's attribution of the collapse to smear-background shortcuts is an interpretation with competing explanations (staining, magnification, cell-selection pipeline, label noise, and extreme test imbalance), but a causal claim that is under-supported by ablations is a correctness risk, not circularity. The self-citations in the reference list ([6] and [32]) concern traffic-sign detection and are not load-bearing for the leukemia classification claims. No equation in the paper is equivalent to its own input, and no load-bearing premise rests solely on a self-citation. Therefore the paper receives a score of 0 on the circularity scale.
Assumptions & free parameters
free parameters (4)
- RAC fusion weight alpha (linear probe) =
grid {0, 0.25, 0.5, 0.75, 1}, selected on internal validation split
- RAC fusion weight alpha (LoRA) =
0.5 (fixed)
- Retrieval top-k =
20
- LoRA rank r =
8
assumptions (5)
- domain assumption The three encoders are comparable enough that pretraining domain is the main source of performance variation.
- domain assumption ALL-IDB2 with n=260 is a representative held-out domain for Stage 1.
- domain assumption The Stage 2 held-out pair (ALL-IDB2 plus AML-Cytomorphology) is a valid unseen-domain test for subtyping.
- domain assumption The 96.5% AML nearest-neighbor rate indicates encoder-induced embedding structure beyond the bank's class imbalance.
- domain assumption Patient-level diagnostic labels in D4 correctly transfer to every single-cell crop.
Cite this review
Pith. "Pith review of Retrieval-Augmented Vision Foundation Models for Robust Leukemia Cell Classification across Multiple Microscopy Datasets." pith.science (2026). https://pith.science/paper/4U4SU3NM
@misc{pith2026260810657,
author = {Pith},
title = {Pith review of: Retrieval-Augmented Vision Foundation Models for Robust Leukemia Cell Classification across Multiple Microscopy Datasets},
year = {2026},
howpublished = {\url{https://pith.science/paper/4U4SU3NM}},
note = {Machine review of arXiv:2608.10657}
}
read the original abstract
Leukemia cell image classification is challenged by real-world domain shifts from acquisition, staining, illumination, and site protocols, causing single-dataset models to generalize poorly in real clinical scenarios. This work presents a robust framework for leukemia classification across multiple heterogeneous datasets using a two-stage pipeline with a pretrained vision foundation model. Stage 1 performs binary classification (leukemia vs. non-leukemia) and is trained using 122,167 single-cell images. Stage 2 is conditionally applied to Stage 1 positives to perform subtype classification into Acute Lymphoblastic Leukemia (ALL) and Acute Myeloid Leukemia (AML), trained using 69,400 single-cell images. Labels are harmonized across five heterogeneous datasets to enable cross-dataset training, and performance is evaluated on a held-out dataset protocol to assess domain-shift generalization. Within this pipeline, three encoders are benchmarked (DinoBloom, pretrained on single-cell images; BiomedCLIP, pretrained on biomedical data; and CLIP as a general-purpose model) under linear probing, Low-Rank Adaptation (LoRA), and a Retrieval-Augmented Classification (RAC) module that retrieves the top-k most similar cell images to provide cytomorphological grounding. The objective is to quantify how much domain-specific pretraining contributes to performance under domain shift, and whether cost-effective adaptation and retrieval can be a viable alternative to expensive domain-specialized pretraining. The held-out protocol additionally serves as a diagnostic tool, revealing when classification performance is attributable to dataset-specific artifacts rather than to cytomorphological features.
Figures
Figures from the paper (9 more)
Reference graph
Works this paper leans on
-
[1]
Global, regional, and national burden of childhood leukemia from 1990 to 2021,
Liu, W., Fang, J., Zhu, M., Zhou, J., and Yuan, C., “Global, regional, and national burden of childhood leukemia from 1990 to 2021,”BMC Pediatrics25(1), 571 (2025)
work page 2025
-
[2]
Diagnosis and typing of leukemia using a single peripheral blood cell through deep learning,
Yan, G., Gao, M., Shi, W., Liang, H., Qin, L., Liu, A., Kong, X., Zhao, H., Zhao, J., and Yan, Q., “Diagnosis and typing of leukemia using a single peripheral blood cell through deep learning,”Cancer Science116(2), 533–543 (2025)
work page 2025
-
[3]
A review of deep learning techniques for leukemia cancer classification based on blood smear images,
Oybek Kizi, R. F., Theodore Armand, T. P., and Kim, H.-C., “A review of deep learning techniques for leukemia cancer classification based on blood smear images,”Applied Biosciences4(1), 9 (2025)
work page 2025
-
[4]
Reliable leukemia detection via transfer-enhanced bayesian cnns,
Hita, X., Javed, F., and Lodi, S., “Reliable leukemia detection via transfer-enhanced bayesian cnns,”Com- puters in Biology and Medicine202, 111419 (2026)
work page 2026
-
[5]
Nunna, H. K., Altable, A., Gundala, P., and Rangarajan, P. K., “Vista: vision transformer-attention enhanced cnn ensemble for optimized classification of acute lymphoblastic leukemia benign and progressive malignant stages,”International Journal of Information Technology(2024)
work page 2024
-
[6]
Traffic sign classification using real-time gpu-embedded systems,
Lopez-Montiel, M., Orozco-Rosas, U., S´ anchez-Adame, M., Montiel, O., Picos, K., and Tapia, J. J., “Traffic sign classification using real-time gpu-embedded systems,”SN Computer Science7(1), 12 (2025)
work page 2025
-
[7]
Dinobloom: A foundation model for generalizable cell embeddings in hematology,
Koch, V., Wagner, S. J., Kazeminia, S., Sancar, E., Hehr, M., Schnabel, J. A., Peng, T., and Marr, C., “Dinobloom: A foundation model for generalizable cell embeddings in hematology,” in [Medical Image Computing and Computer Assisted Intervention – MICCAI 2024],Lecture Notes in Computer Science 15012, 520–530, Springer, Cham (2024)
work page 2024
-
[8]
A multimodal biomedical foundation model trained from fifteen million image–text pairs,
Zhang, S., Xu, Y., Usuyama, N., Xu, H., Bagga, J., Tinn, R., Preston, S., Rao, R., Wei, M., Valluri, N., Wong, C., Tupini, A., Wang, Y., Mazzola, M., Shukla, S., Liden, L., Gao, J., Crabtree, A., Piening, B., Bifulco, C., Lungren, M. P., Naumann, T., Wang, S., and Poon, H., “A multimodal biomedical foundation model trained from fifteen million image–text ...
work page 2025
Show all 46 references
-
[9]
Learning transferable visual models from natural language super- vision,
Radford, A., Kim, J. W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., Krueger, G., and Sutskever, I., “Learning transferable visual models from natural language super- vision,” in [Proceedings of the 38th International Conferen...
2021
-
[10]
Gradient-based learning applied to document recogni- tion,
LeCun, Y., Bottou, L., Bengio, Y., and Haffner, P., “Gradient-based learning applied to document recogni- tion,”Proceedings of the IEEE86(11), 2278–2324 (1998)
1998
-
[11]
An image is worth 16x16 words: Transformers for image recognition at scale,
Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., and Houlsby, N., “An image is worth 16x16 words: Transformers for image recognition at scale,” in [International Conferenc...
2021
-
[12]
Acute lymphoblastic leukemia subtypes detection using vision transformer model,
Prasad, P. and Anbarasi L., J., “Acute lymphoblastic leukemia subtypes detection using vision transformer model,” in [5th International Conference on Data Intelligence and Cognitive Informatics (ICDICI)], 1413– 1418 (2024)
2024
-
[13]
All-vit: A novel approach for detection of acute lym- phoblastic leukemia,
Pandey, P., RG, R. T., Pati, R. P., and Singh, S., “All-vit: A novel approach for detection of acute lym- phoblastic leukemia,” in [12th International Conference on Emerging Trends in Engineering & Technology – Signal and Information Processing (ICETET–SIP)], 1–6 (2025)
2025
-
[14]
Leukovit-unet: A u-net and vision transformer hybrid approach for leukemia cell segmentation and classification,
Revathi, B. and Kaliappan, M., “Leukovit-unet: A u-net and vision transformer hybrid approach for leukemia cell segmentation and classification,”International Journal of Computational Intelligence Sys- tems19(1), 257 (2026)
2026
-
[15]
Hybrid vision transformer and cnn for leukemia detection in blood smear images,
Preethika, P. and Ananthajothi, K., “Hybrid vision transformer and cnn for leukemia detection in blood smear images,” in [11th International Conference on Communication and Signal Processing (ICCSP)], 1344– 1349 (2025)
2025
-
[16]
Imbalanced domain generaliza- tion for robust single cell classification in hematological cytomorphology,
Umer, R. M., Gruber, A., Boushehri, S. S., Metak, C., and Marr, C., “Imbalanced domain generaliza- tion for robust single cell classification in hematological cytomorphology,” in [ICLR Workshop on Domain Generalization], (2023). arXiv:2303.07771
2023 arXiv
-
[17]
Cross-platform deep learning enables automated cytomorphologic subtyping of acute leukemia from bone marrow smears,
Wang, G., Chen, W., Zhao, H., Chen, X., Wu, M., Yu, Q., Liu, Q., Liu, H., Song, E., Xue, M., Sun, C., Chu, Z., and Hu, Y., “Cross-platform deep learning enables automated cytomorphologic subtyping of acute leukemia from bone marrow smears,”Virchows Archiv(2026)
2026
-
[18]
Microscopic hematological image classification with captions using few-shot learning in data-scarce environments,
Patel, T., El-Sayed, H., and Sarker, M. K., “Microscopic hematological image classification with captions using few-shot learning in data-scarce environments,” in [IEEE International Conference on Internet of Things and Intelligence Systems (IoTaIS)], 184–190 (2024)
2024
-
[19]
Evaluating vision-language models for hematology image classification: Performance analysis of clip and its biomedical ai variants,
Patel, T., El-Sayed, H., and Sarker, M. K., “Evaluating vision-language models for hematology image classification: Performance analysis of clip and its biomedical ai variants,” in [36th Conference of Open Innovations Association (FRUCT)], 578–584 (2024)
2024
-
[20]
Multimodal knowledge graph–guided rag-llm for clinical decision support in pediatric leukemia,
Song, J. K., Youk, D. B., Kim, H., and Hwang, S.-H., “Multimodal knowledge graph–guided rag-llm for clinical decision support in pediatric leukemia,”Cancer Research and Treatment(2026). Epub ahead of print
2026
-
[21]
Images retrieval and classifica- tion for acute myeloid leukemia blood cell using deep metric learning,
Naing, K. M., Kittichai, V., Tongloy, T., Chuwongin, S., and Boonsang, S., “Images retrieval and classifica- tion for acute myeloid leukemia blood cell using deep metric learning,” in [Recent Challenges in Intelligent Information and Database Systems (ACIIDS)],Communications i...
2023
-
[22]
AI-based hematological malignancy prediction from peripheral blood smears in a large diagnostic laboratory cohort,
Dasdelen, M. F., Kukuljan, I., Lienemann, P., Ozlugedik, F., Sadafi, A., Hehr, M., Spiekermann, K., Pohlkamp, C., and Marr, C., “AI-based hematological malignancy prediction from peripheral blood smears in a large diagnostic laboratory cohort,”Leukemia40(6), 1318–1322 (2026)
2026
-
[23]
A hierarchical ensemble inference pipeline for robust white blood cell classification under domain shifts,
Ng, T., Dai, R., and Chen, H., “A hierarchical ensemble inference pipeline for robust white blood cell classification under domain shifts,” in [IEEE 23rd International Symposium on Biomedical Imaging (ISBI)], 1–4 (2026)
2026
-
[24]
Do foundation models truly outperform domain-specific models? evidence from digital pathology,
Ben Rabah, C. and Serag, A., “Do foundation models truly outperform domain-specific models? evidence from digital pathology,”Machine Learning and Knowledge Extraction8(6), 164 (2026)
2026
-
[25]
Advancements in leukemia management: Bridging diagnosis, prognosis and nanotechnology (review),
Li, J., Wang, Y., Dong, C., and Luo, L., “Advancements in leukemia management: Bridging diagnosis, prognosis and nanotechnology (review),”International Journal of Oncology65(6), 112 (2024)
2024
-
[26]
Normal and leukemic hematopoiesis: Are leukemias a stem cell disorder or a reacquisition of stem cell characteristics?,
Passegu´ e, E., Jamieson, C. H., Ailles, L. E., and Weissman, I. L., “Normal and leukemic hematopoiesis: Are leukemias a stem cell disorder or a reacquisition of stem cell characteristics?,”Proceedings of the National Academy of Sciences of the United States of America100(Supp...
2003
-
[27]
Morphology of leukaemias,
Ladines-Castro, W., Barrag´ an-Iba˜ nez, G., Luna-P´ erez, M. A., Santoyo-S´ anchez, A., Collazo-Jaloma, J., Mendoza-Garc ´ ıa, E., and Ramos-Pe˜ nafiel, C. O., “Morphology of leukaemias,”Revista M´ edica del Hospital General de M´ exico79(2), 107–113 (2016)
2016
-
[28]
Lymphoid and myeloid lineage commitment in multipotent hematopoietic progenitors,
Kondo, M., “Lymphoid and myeloid lineage commitment in multipotent hematopoietic progenitors,”Im- munological Reviews238(1), 37–46 (2010)
2010
-
[29]
Role of morphology in the diagnosis of acute leukemias: Systematic review,
Sekar, M. D., Raj, M., and Manivannan, P., “Role of morphology in the diagnosis of acute leukemias: Systematic review,”Indian Journal of Medical and Paediatric Oncology44(5), 464–473 (2023)
2023
-
[30]
Domain-incremental white blood cell classification with privacy-aware continual learning,
Kumari, P., Bozorgpour, A., Reisenb¨ uchler, D., Jost, E., Crysandt, M., Matek, C., and Merhof, D., “Domain-incremental white blood cell classification with privacy-aware continual learning,”Scientific Re- ports15, 25468 (2025)
2025
-
[31]
All-idb: The acute lymphoblastic leukemia image database for image processing,
Donida Labati, R., Piuri, V., and Scotti, F., “All-idb: The acute lymphoblastic leukemia image database for image processing,” in [Proceedings of the IEEE International Conference on Image Processing (ICIP)], 2045–2048 (2011)
2011
-
[32]
Evaluation method of deep learning-based embedded systems for traffic sign detection,
Lopez-Montiel, M., Orozco-Rosas, U., S´ anchez-Adame, M., Picos, K., and Ross, O. H. M., “Evaluation method of deep learning-based embedded systems for traffic sign detection,”IEEE Access9, 101217–101238 (2021)
2021
-
[33]
Dinov2: Learning robust visual features without supervision,
Oquab, M., Darcet, T., Moutakanni, T., Vo, H. V., Szafraniec, M., Khalidov, V., Fernandez, P., Haziza, D., Massa, F., El-Nouby, A., Assran, M., Ballas, N., Galuba, W., Howes, R., Huang, P.-Y., Li, S.-W., Misra, I., Rabbat, M., Sharma, V., Synnaeve, G., Xu, H., Jegou, H., Maira...
2024
-
[34]
Application of convolutional neural networks in medical images: a bibliometric analysis,
Jia, H., Zhang, J., Ma, K., Qiao, X., Ren, L., and Shi, X., “Application of convolutional neural networks in medical images: a bibliometric analysis,”Quantitative Imaging in Medicine and Surgery14(5), 3501–3518 (2024)
2024
-
[35]
Parameter-efficient fine-tuning for large models: A comprehensive survey,
Han, Z., Gao, C., Liu, J., Zhang, J., and Zhang, S. Q., “Parameter-efficient fine-tuning for large models: A comprehensive survey,”arXiv preprint arXiv:2403.14608(2024)
2024 arXiv
-
[36]
LoRA: Low-rank adaptation of large language models,
Hu, E. J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., and Chen, W., “LoRA: Low-rank adaptation of large language models,”arXiv preprint arXiv:2106.09685(2021)
2021 arXiv
-
[37]
Low-rank few-shot adaptation of vision-language models,
Zanella, M. and Ben Ayed, I., “Low-rank few-shot adaptation of vision-language models,” in [Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPR W)], 1593–1603 (2024)
2024
-
[38]
To- wards better understanding of domain shift on linear-probed visual foundation models,
Heim, E., Zhao, A., Tanner, C., Perkonigg, M. G., Ramesh, V. K., Angel, S., and Sampson, A., “To- wards better understanding of domain shift on linear-probed visual foundation models,” in [NeurIPS 2023 Workshop on Distribution Shifts: Connecting Methods and Applications], (2023)
2023
-
[39]
Retrieval-augmented generation for knowledge-intensive nlp tasks,
Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., K¨ uttler, H., Lewis, M., Yih, W., Rockt¨ aschel, T., Riedel, S., and Kiela, D., “Retrieval-augmented generation for knowledge-intensive nlp tasks,”arXiv preprint arXiv:2005.11401(2020)
2020 arXiv
-
[40]
Retrieval augmented classification for long-tail visual recognition,
Long, A., Yin, W., Ajanthan, T., Nguyen, V., Purkait, P., Garg, R., Blair, A., Shen, C., and van den Hengel, A., “Retrieval augmented classification for long-tail visual recognition,”arXiv preprint arXiv:2202.11233 (2022)
2022 arXiv
-
[41]
Class imbalance on medical image classification: towards better evaluation practices for discrimination and calibration performance,
Mosquera, C., Ferrer, L., Milone, D. H., Luna, D., and Ferrante, E., “Class imbalance on medical image classification: towards better evaluation practices for discrimination and calibration performance,”European Radiology34(12), 7895–7903 (2024)
2024
-
[42]
All challenge dataset of ISBI 2019 (C-NMC 2019),
Mourya, S., Kant, S., Kumar, P., Gupta, A., and Gupta, R., “All challenge dataset of ISBI 2019 (C-NMC 2019),” (2019)
2019
-
[43]
A single-cell morphological dataset of leukocytes from aml patients and non-malignant controls,
Matek, C., Schwarz, S., Marr, C., and Spiekermann, K., “A single-cell morphological dataset of leukocytes from aml patients and non-malignant controls,” (2019)
2019
-
[44]
A dataset for microscopic peripheral blood cell images for development of automatic recognition systems,
Acevedo, A., Merino, A., Alf´ erez, S., Molina, ´A., Bold´ u, L., and Rodellar, J., “A dataset for microscopic peripheral blood cell images for development of automatic recognition systems,” (2020)
2020
-
[45]
A morphological dataset of white blood cells from patients with four different genetic aml entities and non-malignant controls (aml-cytomorphology mll helmholtz),
Hehr, M., Sadafi, A., Matek, C., Lienemann, P., Pohlkamp, C., Haferlach, T., Spiekermann, K., and Marr, C., “A morphological dataset of white blood cells from patients with four different genetic aml entities and non-malignant controls (aml-cytomorphology mll helmholtz),” (2023)
2023
-
[46]
Lam, S. T. T. and Wang, W., [Normal Histology of Blood], BCcampus Open Publishing (2021)
2021
Reviewed August 12, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.