-
This is malignancy classification using simple deep learning method in LIDC-IDRI dataset.
-
Malignancy (True): 633, Benign (False): 1992
-
The dataset can download the below link.
-> Dataset (LIDC-IDRI) : https://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=1966254
-
Contact
The code is based on
- torch == '2.0.0'
- timm == '0.6.12'
- pylidc == '0.2.3'
- torchmetrics == '0.9.3'
- sklean == '1.0.2'
- kornia == '0.6.8'
<ROC analysis>
Acc: 81.09, Balenced Acc.:75.21, spec.:0.8109, pre.:0.7964, rec.:0.8109, F1: 0.7957 ROC AUC: 0.7671, PR AUC: 0.8106
Confusion matrix
| False | True | |
|---|---|---|
| False | 557 | 41 |
| True | 108 | 82 |
Analysis #1
| precision | recall | f1-score | support | |
|---|---|---|---|---|
| False | 0.8376 | 0.9314 | 0.8820 | 598 |
| True | 0.6667 | 0.4316 | 0.5240 | 190 |
| accuracy | 0.8109 | 788 | ||
| macro avg | 0.7521 | 0.6815 | 0.7030 | 788 |
| weighted avg | 0.7964 | 0.8109 | 0.7957 | 788 |
Analysis #2
| Sensitivity | Specificity | Precision | ACC | |
|---|---|---|---|---|
| False | 0.931438 | 0.431579 | 0.837594 | 0.810914 |
| True | 0.431579 | 0.931438 | 0.666667 | 0.810914 |