I base on classical image classification pipeline

1. Input image
   • Visualization
   • Categorization
2. Data Preprocessing
   • Standartization
   • Augmentation (if needed)
3. Features extracion
4. Classification model
   • Choosing
     • network arhitecture
     • loss functions
     • evaluation metrics
   • Train, vlidation, test
5. Testing the results with statistical methods

• Writing technical requirements for the internship
• Get familiar with working data
• Read about histogram matching

1. Problem statement
2. Working data
    a. ChestX-ray 
    b. CheXpert
3. Methodology adopted
4. Results
5. References

There are two datasets to deal with chest x-ray images:

• ChestX-ray
• CheXpert

Eight visual examples of common thorax diseases

ChestX-ray dataset comprises 112,120 frontal-view X-ray images of 30,805 unique patients from with the text-mined fourteen disease image labels:

• Atelectasis
• Cardiomegaly
• Consolidation
• Edema
• Effusion
• Emphysema
• Fibrosis
• Hernia
• Infiltration
• Nodule
• Mass
• Pleural_thickening
• Pneumonia
• Pneumothorax

The dataset were presented by the National Institutes of Health Clinical Center with the next material:

• 112.120 frontal-view chest X-ray PNG images in 1024x1024 resolution
• Meta data for all images (Data_Entry_2017.csv):
  • Image Index
  • Finding Labels
  • Follow-up number
  • Patient ID
  • Patient Age
  • Patient Gender
  • View Position
  • Original Image
  • Pixel Spacing
• Bounding boxes for ~1000 images (BBox_List_2017.csv):
  • Image Index, Finding Label,
  • Bbox [x, y, w, h]
    • [x y] are coordinates of each box's topleft corner.
    • [w h] represent the width and height of each box.
• Two data split files. All studies from the same patient will only appear in either training/validation or testing set:
  • train_val_list.txt
  • test_list.txt

CheXpert is a large public dataset for chest radiograph interpretation, consisting of 224,316 chest radiographs of 65,240 patients with fourteen labels:

• Atelectasis
• Cardiomegaly
• Consolidation
• Edema
• Enlarged Cardiomyopathy
• Fracture
• Lung Lesion
• Lung Opacity
• No Finding
• Pleural Effusion
• Pleural Other
• Pneumonia
• Pneumothorax
• Support Devices

The dataset were presented by Stanford Hospital with the next material:

• 224,316 images in 320x320 resolution
• Data split files:
  • train.csv
  • valid.csv
  • test.csv

Histogram matching or histogram specification is the transformation of an image so that its histogram matches a specified histogram

Example (later there will be a chest x-ray image example)

Source image	Image to be adjusted

Source image grey levels	Image to be adjusted grey levels

Source image

Image to be adjusted

Source image histogram	Image to be adjusted histogram

Finally:

Source image	Matched image