pytorch-covid19

This is merely an experiment done on a few images and has not been validated/checked by external health organizations or doctors. No clinical studies have been performed based on the approach which can validate it. This model has been done as a P.O.C. and nothing can be concluded/inferred from this result.

Directory structure

.
├── data
│   ├── external
│   ├── interim
│   ├── processed
│   ├── raw
│   │   ├── covid
│   │   ├── normal
│   │   └── pneumonia
│   └── raw.csv
├── models
│   └── checkpoint.pth
├── reports
│   ├── architecture.csv
│   └── figures
├── scripts
│   ├── activationmap.py
│   ├── architectures.py
│   ├── datagen.py
│   ├── __init__.py
│   ├── test.py
│   ├── train.py
│   └── utils.py
├── makedataset.py
└── trainer.py
├── evaluate.py
├── README.md

Dataset

1. covid-chestxray-dataset : dataset of chest X-ray and CT images of patients which are positive or suspected of COVID-19 or other viral and bacterial pneumonias (MERS, SARS, and ARDS.).

2. chest-xray-pneumonia : dataset of chest X-ray images of normal patients and infected with Pneumonia ( bacterial and viral )

CNN Model

base model : RestNet50, input_shape=(256, 256), pretrained=True with modified fc_layer

Scripts

1. makedataset.py

create csv file with LABEL and IMAGE_PATH

path = "./data"
sample_per_category = 500
seed = 24
split_frac = 0.20

output -

./data/2_class_test_df.csv
./data/2_class_train_df.csv
./data/3_class_test_df.csv
./data/3_class_train_df.csv
./data/raw.csv

---

2. trainer.py

train_file = "data/3_class_train_df.csv"
num_workers = 2
val_split = 0.2
batch_size = 32
num_epochs = 20
input_shape = (3, 256, 256)
le = LabelEncoder()

output -

./models/checkpoint.pth

---

3. evaluate.py

test_file = "data/3_class_test_df.csv"
image_file = "data/raw/covid/covid_001.jpg"
num_workers = 2
batch_size = 1
input_shape = (256, 256)
le = LabelEncoder()

• test_model(model,testloader,device,encoder=None)

• test_image(model,image,in_shape,transform,device,labelencoder=None,cam=None)

---

Sample Model Results

[phase: test] total: 240, correct: 112, acc: 46.667

            precision    recall  f1-score   support

        0       0.00      0.00      0.00        38
        1       0.69      0.17      0.28       104
        2       0.44      0.96      0.60        98

    accuracy                           0.47       240
macro avg       0.38      0.38      0.29       240
weighted avg       0.48      0.47      0.37       240

[phase: test] confusion matrix

Predicted  0   1    2  All
Actual
0          0   4   34   38
1          1  18   85  104
2          0   4   94   98
All        1  26  213  240

{0: 'covid', 1: 'normal', 2: 'pneumonia'}

---

Sample Image Results

1. Normal X-Ray

2. Covid19