-
Notifications
You must be signed in to change notification settings - Fork 3
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Showing
2 changed files
with
18 additions
and
20 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,8 +1,6 @@ | ||
# LongTailCXR | ||
|
||
## **[WORK IN PROGRESS]** | ||
|
||
Code repository for **"Long-Tailed Classification of Thorax Diseases on Chest X-Ray: A New Benchmark Study"** by Gregory Holste, Song Wang, Ziyu Jiang, Tommy C. Shen, Ronald D. Summers, Yifan Peng, and Zhangyang Wang. To be presented at [DALI 2022](https://dali-miccai.github.io/), a MICCAI workshop. | ||
Code repository for [**"Long-Tailed Classification of Thorax Diseases on Chest X-Ray: A New Benchmark Study"**](https://arxiv.org/abs/2208.13365) by Gregory Holste, Song Wang, Ziyu Jiang, Tommy C. Shen, Ronald D. Summers, Yifan Peng, and Zhangyang Wang. To be presented at [DALI 2022](https://dali-miccai.github.io/), a MICCAI workshop. | ||
|
||
----- | ||
|
||
|
@@ -20,23 +18,23 @@ Imaging exams, such as chest radiography, will yield a small set of common findi | |
|
||
All trained model weights are available below. In the following table, best results are **bolded** and second-best results are <u>underlined</u>. See paper for full results (bAcc = [balanced accuracy](https://scikit-learn.org/stable/modules/generated/sklearn.metrics.balanced_accuracy_score.html#sklearn.metrics.balanced_accuracy_score)). | ||
|
||
| Method | NIH-LT bAcc | MIMIC-CXR-LT bAcc | NIH-LT Weights | MIMIC-CXR-LT Weights | | ||
| Method | NIH-CXR-LT bAcc | MIMIC-CXR-LT bAcc | NIH-CXR-LT Weights | MIMIC-CXR-LT Weights | | ||
| :--- | :---: | :---: | :---: | :---: | | ||
| Softmax | 0.115 | 0.169 | link | link | | ||
| CB Softmax | 0.269 | 0.227 | link | link | | ||
| RW Softmax | 0.260 | 0.211 | link | link | | ||
| Focal Loss | 0.122 | 0.172 | link | link | | ||
| CB Focal Loss | 0.232 | 0.191 | link | link | | ||
| RW Focal Loss | 0.197 | 0.239 | link | link | | ||
| LDAM | 0.178 | 0.165 | link | link | | ||
| CB LDAM | 0.235 | 0.225 | link | link | | ||
| CB LDAM-DRW | 0.281 | 0.267 | link | link | | ||
| RW LDAM | 0.279 | 0.243 | link | link | | ||
| RW LDAM-DRW | <u>0.289</u> | <u>0.275</u> | link | link | | ||
| MixUp | 0.118 | 0.176 | link | link | | ||
| Balanced-MixUp | 0.155 | 0.168 | link | link | | ||
| Decoupling (cRT) | **0.294** | **0.296** | link | link | | ||
| Decoupling ($\tau$-norm) | 0.214 | 0.230 | link | link | | ||
| Softmax | 0.115 | 0.169 | [link](https://drive.google.com/file/d/1lzDBDwRRcKmYHaypyLc59MsbdSbOuq75/view?usp=sharing) | [link](https://drive.google.com/file/d/1iKMqNX_KvczyuJZAibmRSWU5cqKQ0psn/view?usp=sharing) | | ||
| CB Softmax | 0.269 | 0.227 | [link](https://drive.google.com/file/d/1m0Xt_COF8SY5ZxKo3qrpwBDBjTnZIfd7/view?usp=sharing) | [link](https://drive.google.com/file/d/1GDCWZ0J1GhGdEqcEP9Ubp42p4tz_w58b/view?usp=sharing) | | ||
| RW Softmax | 0.260 | 0.211 | [link](https://drive.google.com/file/d/1rvl_W3ZP6-059hevrP8FiRC3CM41DN65/view?usp=sharing) | [link](https://drive.google.com/file/d/1li4zP5-hCtazWVzC8Cp99o3hUU6t_PVv/view?usp=sharing) | | ||
| Focal Loss | 0.122 | 0.172 | [link](https://drive.google.com/file/d/1YuMcxv9d8H1rH-nP-ccMmrT3MXJb0SGQ/view?usp=sharing) | [link](https://drive.google.com/file/d/1OxnUQxjAfsrydXcaJ6Xy2-WlA7jNNkRG/view?usp=sharing) | | ||
| CB Focal Loss | 0.232 | 0.191 | [link](https://drive.google.com/file/d/1wOk9NlDrp4c52WjvJsqetxlEVfFJndBr/view?usp=sharing) | [link](https://drive.google.com/file/d/1ZzPJTA-OBLYphkzO5yF8r_VZgpoa6tXT/view?usp=sharing) | | ||
| RW Focal Loss | 0.197 | 0.239 | [link](https://drive.google.com/file/d/1wMa6hd8J3jxlled7B66iDV43C3zdtL8l/view?usp=sharing) | [link](https://drive.google.com/file/d/1eTZ5K8HeDHzu3y_Nj0K2_MxPrK-E9MJg/view?usp=sharing) | | ||
| LDAM | 0.178 | 0.165 | [link](https://drive.google.com/file/d/1i_kXKI4IXbWyABk6ChsqkSAaRu_LkmCi/view?usp=sharing) | [link](https://drive.google.com/file/d/1eT16iWKrpxJNIghLdaSq9Hr4aAt99CAL/view?usp=sharing) | | ||
| CB LDAM | 0.235 | 0.225 | [link](https://drive.google.com/file/d/1p8uYrJH539Q9DgsEg7Ru_wOyRbZ1_taF/view?usp=sharing) | [link](https://drive.google.com/file/d/1mlOcyTuAN5SVlBXw-qyON7jXk7dHnhho/view?usp=sharing) | | ||
| CB LDAM-DRW | 0.281 | 0.267 | [link](https://drive.google.com/file/d/17HMaldk6pwHEHZ-c3SJwPw3JWeYjCtI6/view?usp=sharing) | [link](https://drive.google.com/file/d/1YUtJq5iPgbd4O_p77EhhXJoA_CfvR8Ct/view?usp=sharing) | | ||
| RW LDAM | 0.279 | 0.243 | [link](https://drive.google.com/file/d/1TZikaKB2sAqBA4o6bp9zVly463UAAftH/view?usp=sharing) | [link](https://drive.google.com/file/d/1X6p12_79o46OIBvSnnwERurv9x7eMf7t/view?usp=sharing) | | ||
| RW LDAM-DRW | <u>0.289</u> | <u>0.275</u> | [link](https://drive.google.com/file/d/1hVe7y4sWE0o90UsZSRraQAU0UEmcu73c/view?usp=sharing) | [link](https://drive.google.com/file/d/1OVHRGfQVia3SU5UTBoQ2FtcRkiaYK63E/view?usp=sharing) | | ||
| MixUp | 0.118 | 0.176 | [link](https://drive.google.com/file/d/1gP1LTgBQsrgCqzu3lyFK7TkcaPnSI-q7/view?usp=sharing) | [link](https://drive.google.com/file/d/1OjlkBsuumdvTtrUfBSGnCbONhXEk_cYf/view?usp=sharing) | | ||
| Balanced-MixUp | 0.155 | 0.168 | [link](https://drive.google.com/file/d/1_GQXraEbGVMVu5WpAN8k1YB74M5yTNcV/view?usp=sharing) | [link](https://drive.google.com/file/d/16xA335kGktjH-O8iu8821LJc279bKjMh/view?usp=sharing) | | ||
| Decoupling (cRT) | **0.294** | **0.296** | [link](https://drive.google.com/file/d/1nOqVEeZBmzyMM8fm46ziY6dqQHHcsAHm/view?usp=sharing) | [link](https://drive.google.com/file/d/1rbpyKxQsIGZbclMW0Fauxbt2TrxXoK8H/view?usp=sharing) | | ||
| Decoupling (tau-norm) | 0.214 | 0.230 | -- | -- | | ||
|
||
----- | ||
|
||
|
@@ -46,7 +44,7 @@ To reproduce the results presented in this paper... | |
1. Register to download the MIMIC-CXR dataset from https://physionet.org/content/mimic-cxr/2.0.0/, and download the NIH ChestXRay14 dataset from https://nihcc.app.box.com/v/ChestXray-NIHCC/. | ||
2. Install prerequisite packages with Anaconda: `conda env create -f lt_cxr.yml` and `conda activate lt_cxr`. | ||
3. Run all MIMIC-CXR-LT experiments: `bash run_mimic-cxr-lt_experiments.sh` (changing the `--data_dir` argument to your MIMIC-CXR path). | ||
4. Run all NIH-LT experiments: `bash run_nih-lt_experiments.sh` (changing the `--data_dir` argument to your NIH ChestXRay14 path). | ||
4. Run all NIH-LT experiments: `bash run_nih-cxr-lt_experiments.sh` (changing the `--data_dir` argument to your NIH ChestXRay14 path). | ||
|
||
Labels for the MIMIC-CXR-LT benchmark presented in this paper can be found in the `labels/` directory. Labels for NIH-LT are readily available upon request; for access, please email Dr. Ronald Summers ([email protected]) and Greg Holste ([email protected]). All experiments were conducted on a single NVIDIA RTX A6000 GPU. | ||
|
||
|
File renamed without changes.