Chest radiography, like many diagnostic medical exams, produces a long-tailed distribution of clinical findings; while a small subset of diseases is routinely observed, the vast majority of diseases are relatively rare [1]. This poses a challenge for standard deep learning methods, which exhibit bias toward the most common classes at the expense of the important but rare "tail" classes [2]. Many existing methods [3] have been proposed to tackle this specific type of imbalance, though only recently has attention been given to long-tailed medical image recognition problems [4-6]. Diagnosis on chest X-rays (CXRs) is also a multi-label problem, as patients often present with multiple disease findings simultaneously; however, only a select few studies incorporate knowledge of label co-occurrence into the learning process [7-9, 12]. Since most large-scale image classification benchmarks contain single-label images with a mostly balanced distribution of labels, many standard deep learning methods fail to accommodate the class imbalance and co-occurrence problems posed by the long-tailed, multi-label nature of tasks like disease diagnosis on CXRs [2].

The first event, CXR-LT 2023 [13], aimed to achieve these goals by providing high-quality benchmark CXR data for model development and conducting comprehensive evaluations to identify ongoing issues impacting lung disease classification performance. Building on the success of CXR-LT 2023, the CXR-LT 2024 [14] expands the dataset to 377,110 chest X-rays (CXRs) and 45 disease labels, including 19 new rare disease findings. It also introduces a new focus on zero-shot learning to address limitations identified in the previous event.

For this year's version of CXR-LT 2026, we continue and extend the CXR-LT benchmark with two restructured tasks focused on long-tailed multi-label classification of chest X-rays. These tasks are directly motivated by real-world diagnostic scenarios, where disease distributions are highly imbalanced, and ground truth labels may vary in quality and coverage. By addressing these challenges, CXR-LT 2026 encourages the development of methods can better handle data imbalance and label uncertainty in clinical practice.

Dataset

This challenge is based on a curated set of over 160,000 chest X-ray images selected from the PadChest[15], supplemented with PadChest-GR[16], where radiologists provided manual annotations. We curated a final set of 36 long-tailed classes, which span a broad range of chest findings and serve as the prediction targets for this year’s classification task.

Task

Given a CXR, our challenge includes two tasks, to be held as independent tasks:

• Task 1: Given a chest X-ray, the goal is to detect all clinical findings present in the image. This is a multi-label classification task, meaning that a single image may contain multiple findings. You can train your models on the 30 classes listed in Train_CLASS.txt, and they will be evaluated on the same set. You may use external data sources for training, provided you do not use any data reserved for testing.
• Task 2: Given a chest X-ray, the goal is to detect Out-of-distribution(OOD) clinical findings that are not present in the training set. This is also a multi-label classification task, and a single image may contain multiple findings. For Task 2, you can still train your models on the same 30 classes listed in Train_CLASS.txt, but the test set includes images containing six additional findings that do not appear in the training data(listed in Eval_Class_Task2.txt). Therefore, your model must produce predictions for these unseen OOD classes, even though they are not part of the training labels.

For all tasks, participants will be provided with a large, automatically labeled training set of >160,000 CXR images with 30/6 binary disease labels. While last year's CXR-LT was a success, we hope that CXR-LT 2026 can provide even further meaningful methodological advances toward clinically realistic multi-label, long-tailed, and zero-shot disease classification on CXR.

Access to Images

This competition uses all images from PADCHEST. By successfully registering for this competition, you have demonstrated that you have proper access to this dataset. Now you may follow the instructions at https://bimcv.cipf.es/bimcv-projects/padchest/ to download the dataset.

Access to Labels

Unlike the original set of PADCHEST labels, this competition includes labels for 36 disease findings. You can access and download the labelset using this link : https://github.com/CXR-LT/CXR-LT-2026. When uncompressed, it contains:

• CXRLT_2026_training.csv: This file contains the training images, metadata, and labels for both Task1 and Task2. Reminder: You may use external data for this competition, but you may only train your models on the 30 labels listed above.. You may not share these labels under any circumstances, and you are required to delete them once this competition concludes (04/10/2026).
• CXRLT_2026_TEST_ALL_TASK1.csv: This file contains the development set images and metadata for Task 1. You should use your trained models to generate predictions on the provided development set images and upload your submission according to the instructions below.This file will be available before the development phase begins (11/24/2025).
• CXRLT_2026_TEST_ALL_TASK2.csv: This file contains the development set images and metadata for Task 2. You should use your trained models to generate predictions on the provided development set images and upload your submission.This file will be available before the development phase begins (11/24/2025).
• development_sample_submission.zip: This file provides an example submission containing randomly generated predictions in the required format.This file will be available before the development phase begins (11/24/2025).
• Train_CLASS.txt: A newline-delimited text file with training set classes
• Eval_Class_Task2.txt: newline-delimited text file with classes used for evaluation on Task 2.
• Removal.csv: This file lists images that cannot be used during the training phase. These images may be reserved for testing or development, and therefore must be excluded from all training procedures.

Evaluation

For each task, models will be evaluated on the provided testing set using “macro-averaged” mean Average Precision (mAP).There will be two phases of the competition:

• Phase 1: Development Phase (starts 11/24/2025; ends 01/13/2026). We will provide you with labeled training data and unlabeled validation data for model development. Train your models on the training data and upload predictions on this development set. Model performance on the development set data will be displayed on the public leaderboard for your feedback.
• Phase 2: Test Phase (starts 01/13/2026; ends 01/27/2026). We will provide you with unlabeled testing data for final evaluation. Using your models trained on the provided training data, upload predictions on this newly released test set. The test set includes images held out from PadChest, along with images from additional sources, to evaluate the generalizability of the models. Final competition results will only be based on model performance during the Test Phase on the provided test set. For this phase, the leaderboard will be kept hidden until the CXR-LT event at ISBI 2026, and you will be given up to five successful submissions. To participate in the testing phase, teams must have made at least one valid submission during the development phase. Only teams that meet this requirement will be eligible to enter the testing phase.

Online Evaluation

The competition will be conducted through the CodaLab platform.

Awards

This challenge is hosted in conjunction with the ISBI 2026 challenge. For each task, the top three teams will be invited to submit their work to the ISBI 2026 challenge track proceedings, receive an official certificate, and present their methods in an oral session at the ISBI 2026 CXR-LT Challenge in London. After the conference, we will prepare a summary paper to be submitted to a top medical imaging journal (e.g., MedIA, TMI). A selected group of top-performing teams from each task (final number to be announced) will be invited to contribute and describe their methods in this journal publication, similar to our previous challenge report [12].

This competition uses data from PADCHEST [15].

To participate in this competition, you must follow these steps:

• Step 1: Access and download all the images and metadata on PadChest.
• Step 2: Fill out this Google Form providing (i) your CodaLab email address, (ii).confirmation that you agree to follow all competition rules. Participants who violate the rules may have their results disqualified by the organizers.
• Step 3: Register for the competition on CodaLab via the "Participate" tab and await our review. If you have completed these steps correctly, you will be admitted to the competition, and we will email you links to download the necessary data. You are not permitted to share these labels under any circumstances.

Rules

If you have completed these steps correctly, you will be admitted to the competition and we will provide links to download the necessary data by email! You are not permitted to share these labels whatsoever.

• You cannot train on images from this competition's development and test sets. If we find any evidence that you have trained on images listed in Removal.csv, you will be disqualified and barred from participating in the ISBI 2026 CXR-LT challenge.
• You are permitted to train on external data and use pretrained models, so long as they were not trained on images from this competition's development and test sets. For example, you can use models pretrained on ImageNet, but you should avoid using models pretrained on PADChest since those models were likely trained on images from this competition's development and/or test sets, which is not permitted.
• For Task 2 only, you are not allowed to use any images containing the six OOD labels for supervised training, nor may you use any pretrained models obtained using these images or similar label supervision. These OOD-related labels must remain unseen during training to preserve the OOD setting and ensure a fair evaluation.
• All submissions must contain (i) a predictions .csv file and (ii) a code/ directory with all code used for model development and inference.
• Please see detailed submission instructions under "Learn the Details" → "Submission Format". You can find a properly formatted sample submission for the development set under "Participate" → "Files" once registered for the competition.
• Your .csv must contain the dicom_id column denoting each unique image in the given evaluation set and must contain a column for each of the 30/6 evaluation classes. All values in these 30/6 class-specific columns must be values in the range [0, 1] representing the probability that the given image (row) contains the given finding (column).
• All of your training and inference code must be placed in a code/ directory. You will only be allowed to submit to and participate in the ISBI 2026 CXR-LT challenge if your code is transparent and reproducible.
• Your code/ directory and the prediction .csv file should be compressed together into a single .zip file, which you will upload for submission.
• You may, under no circumstances, share any of the provided labeled training data in accordance with the PadChest.
• If you wish to participate in a team, you must use a single shared CodaLab account.
• After completion of the challenge (04/10/2026), you must delete all provided labels from your computer(s).
• You may not manually or automatically extract labels from text reports associated with any PadChest (or PadChest_GR) cases to get the label directly.
• After the testing phase, the top three performing teams for each task will be required to share their code and provide a runnable Docker image to verify that their results can be reproduced.
• To participate in the testing phase, teams must have made at least one valid submission during the development phase. Only teams that meet this requirement will be eligible to enter the testing phase.

1. Zhou SK, Greenspan H, Davatzikos C, Duncan JS, Van Ginneken B, Madabhushi A, Prince JL, Rueckert D, Summers RM. A review of deep learning in medical imaging: Imaging traits, technology trends, case studies with progress highlights, and future promises. Proceedings of the IEEE. 2021 Feb 26;109(5):820-38.
2. Holste G, Wang S, Jiang Z, Shen TC, Shih G, Summers RM, Peng Y, Wang Z. Long-Tailed Classification of Thorax Diseases on Chest X-Ray: A New Benchmark Study. In Data Augmentation, Labelling, and Imperfections: Second MICCAI Workshop, DALI 2022, Held in Conjunction with MICCAI 2022, Singapore, September 22, 2022, Proceedings 2022 Sep 16 (pp. 22-32). Cham: Springer Nature Switzerland.
3. Zhang Y, Kang B, Hooi B, Yan S, Feng J. Deep long-tailed learning: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2023 Apr 19.
4. Zhang R, Haihong E, Yuan L, He J, Zhang H, Zhang S, Wang Y, Song M, Wang L. MBNM: multi-branch network based on memory features for long-tailed medical image recognition. Computer Methods and Programs in Biomedicine. 2021 Nov 1;212:106448.
5. Ju L, Wang X, Wang L, Liu T, Zhao X, Drummond T, Mahapatra D, Ge Z. Relational subsets knowledge distillation for long-tailed retinal diseases recognition. In Medical Image Computing and Computer Assisted Intervention-MICCAI 2021: 24th International Conference, Strasbourg, France, September 27-October 1, 2021, Proceedings, Part VIII 24 2021 (pp. 3-12). Springer International Publishing.
6. Yang Z, Pan J, Yang Y, Shi X, Zhou HY, Zhang Z, Bian C. ProCo: Prototype-Aware Contrastive Learning for Long-Tailed Medical Image Classification. In Medical Image Computing and Computer Assisted Intervention-MICCAI 2022: 25th International Conference, Singapore, September 18-22, 2022, Proceedings, Part VIII 2022 Sep 16 (pp. 173-182). Cham: Springer Nature Switzerland.
7. Chen H, Miao S, Xu D, Hager GD, Harrison AP. Deep hierarchical multi-label classification of chest X-ray images. In International Conference on Medical Imaging with Deep Learning 2019 May 24 (pp. 109-120). PMLR.
8. Wang G, Wang P, Cong J, Liu K, Wei B. BB-GCN: A Bi-modal Bridged Graph Convolutional Network for Multi-label Chest X-Ray Recognition. arXiv preprint arXiv:2302.11082. 2023 Feb 22.
9. Chen B, Li J, Lu G, Yu H, Zhang D. Label co-occurrence learning with graph convolutional networks for multi-label chest x-ray image classification. IEEE Journal of Biomedical and Health Informatics. 2020 Jan 16;24(8):2292-302.
10. Johnson AE, Pollard TJ, Greenbaum NR, Lungren MP, Deng CY, Peng Y, Lu Z, Mark RG, Berkowitz SJ, Horng S. MIMIC-CXR-JPG, a large publicly available database of labeled chest radiographs. arXiv preprint arXiv:1901.07042. 2019 Jan 21.
11. PhysioNet. MIMIC-CXR-JPG - chest radiographs with structured labels [Internet]. Available from: https://physionet.org/content/mimic-cxr-jpg/2.0.0/.
12. Moukheiber D, Mahindre S, Moukheiber L, Moukheiber M, Wang S, Ma C, Shih G, Peng Y, Gao M. Few-Shot Learning Geometric Ensemble for Multi-label Classification of Chest X-Rays. In Data Augmentation, Labelling, and Imperfections: Second MICCAI Workshop, DALI 2022, Held in Conjunction with MICCAI 2022, Singapore, September 22, 2022, Proceedings 2022 Sep 16 (pp. 112-122). Cham: Springer Nature Switzerland.
13. CodaLab. CXR-LT: Multi-Label Long-Tailed Classification on Chest X-Rays [Internet]. Available from: https://codalab.lisn.upsaclay.fr/competitions/12599.
14. Lin M, Holste G, Wang S, Zhou Y, Wei Y, Banerjee I, Chen P, Dai T, Du Y, Dvornek NC, Ge Y, Guo Z, Hanaoka S, Kim D, Messina P, Lu Y, Parra D, Son D, Soto Á, Urooj A, Vidal R, Yamagishi Y, Yan P, Yang Z, Zhang R, Zhou Y, Celi LA, Summers RM, Lu Z, Chen H, Flanders A, Shih G, Wang Z, Peng Y. CXR-LT 2024: A MICCAI challenge on long-tailed, multi-label, and zero-shot disease classification from chest X-ray. Med Image Anal. 2025 Dec;106:103739.
15. Bustos, A., Pertusa, A., Salinas, J. M., & De La Iglesia-Vaya, M. (2020). Padchest: A large chest x-ray image dataset with multi-label annotated reports. Medical image analysis, 66, 101797.
16. de Castro, D. C., Bustos, A., Bannur, S., Hyland, S. L., Bouzid, K., Wetscherek, M. T., ... & Pertusa, A. (2025). Padchest-gr: A bilingual chest X-ray dataset for grounded radiology report generation. NEJM AI, 2(7), AIdbp2401120.