Application of ML with radiological images for COVID-19 diagnosis.

import pandas as pd
from tqdm import tqdm
train_set = pd.read_csv("covid_data/train.csv")
train_set.head()

import seaborn as sns
import matplotlib.pyplot as plt
sns.countplot(x="label", data=train_set)
plt.show()

from pathlib import Path
from skimage import io
img_path = Path("covid_data/data")
images = pd.DataFrame(
    [{"image_id": img.name, "image": io.imread(str(img))} for img in tqdm(img_path.glob("*"))]
)
images.head()

def abnormal_image():
    abnormal_image = {
        index: img.image.shape
        for index, img in images.iterrows()
        if img.image.shape != (256, 256, 3)
    }
    if abnormal_image:
        from collections import Counter
        print(f"Abnormal shape: {Counter(list(abnormal_image.values()))}")
    return abnormal_image

incorrect_image = abnormal_image()

import numpy as np
def compare_images(*images: dict):
    n = len(images)
    fig, axes = plt.subplots(1, n, figsize=(10*n, 5))
    for i, image in enumerate(images):
        ax = axes[i]
        ax.imshow(image['img'].astype("float"), cmap="gray")
        ax.set_title(f"{image['title']}")
    plt.show()

from skimage import transform
target_size = (256, 256, 3)
img = images.loc[2641, "image"] / 255
img_remove = img[:, :, :3] 
img_resize = transform.resize(img, target_size)
compare_images(
    {"img": img, "title": "original image"},
    {"img": img_remove, "title": "removed image"},
    {"img": img_resize, "title": "resized image"},
)

from tqdm import tqdm
for index in tqdm(incorrect_image.keys()):
    image = images.loc[index, "image"]
    resized_image = transform.resize(image, target_size)
    images.loc[index, "image"] = resized_image
    del resized_image