Update README.md
Browse files
README.md
CHANGED
|
@@ -6,30 +6,8 @@ tags:
|
|
| 6 |
|
| 7 |
## Model description
|
| 8 |
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
More information needed
|
| 14 |
-
|
| 15 |
-
## Training and evaluation data
|
| 16 |
-
|
| 17 |
-
More information needed
|
| 18 |
-
|
| 19 |
-
## Training procedure
|
| 20 |
-
|
| 21 |
-
### Training hyperparameters
|
| 22 |
-
|
| 23 |
-
The following hyperparameters were used during training:
|
| 24 |
-
|
| 25 |
-
| Hyperparameters | Value |
|
| 26 |
-
| :-- | :-- |
|
| 27 |
-
| name | Adam |
|
| 28 |
-
| learning_rate | 0.001 |
|
| 29 |
-
| decay | 0.0 |
|
| 30 |
-
| beta_1 | 0.9 |
|
| 31 |
-
| beta_2 | 0.999 |
|
| 32 |
-
| epsilon | 1e-07 |
|
| 33 |
-
| amsgrad | False |
|
| 34 |
-
| training_precision | float32 |
|
| 35 |
|
|
|
|
|
|
| 6 |
|
| 7 |
## Model description
|
| 8 |
|
| 9 |
+
We present this model in relation to previous work by [Oakden-Rayner et al.](https://arxiv.org/abs/1909.12475) and [Jiménez-Sánchez et al.](https://arxiv.org/abs/2211.04279), discussing the issue of hidden features in medical images which tend to confuse machine learning models on the task of pathology detection.
|
| 10 |
+
They studied model performance on pneumothorax cases, which they find to depend on the presence of chest drains that are used for treatment of the pathology.
|
| 11 |
+
We developed this model as a *chest drain detector* in chest X-rays.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
|
| 13 |
+
More details on our study as well as code and data is available [here](https://github.com/trer2547/DrainDetection_ChestXrays).
|